Hydrology 3

UNIVERSITT STUTTGART
INSTITUTE OF HYDRAULIC ENGINEERING CHAIR OF HYDROLOGY AND GEOHYDROLOGY
Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy
Hydrology III
October 2003
Institute of Hydraulic Engineering, Universitt Stuttgart, Germany Pfaffenwaldring 61 * D-70550 Stuttgart Phone: 0711/685-4679 * Fax: 0711/685-4681 * e-mail: [email protected]
Hydrology III - I
Table of contents
1 FUNDAMENTAL PRINCIPLES OF RIVER BASIN MODELING .......................... 1
1.1 Scope ................................................................................................................................................ 1
1.2 Principal methods of river basin modeling ..................................................................................... 2 1.2.1 Statistical methods ..................................................................................................................... 2 1.2.2 Deterministic methods ............................................................................................................... 2 1.2.3 Combined methods..................................................................................................................... 3 1.3 Structure of river basin models ....................................................................................................... 3 1.3.1 Partial models ............................................................................................................................ 3 1.3.2 Drainage basin models ............................................................................................................... 4 1.3.3 Streamflow models..................................................................................................................... 6 1.3.4 Complex river basin model......................................................................................................... 8 1.4 Model approaches............................................................................................................................ 8
1.5 Analysis and synthesis ................................................................................................................... 11 1.5.1 Calibration of the model against observed in- and output values............................................... 11 1.5.2 Synthetic streamflow hydrographs............................................................................................ 11
STRUCTURE OF DRAINAGE BASIN MODELS................................................ 13

2.1 2.2 2.3 Formation of runoff and runoff concentration.............................................................................. 13 Formation of outflow and runoff concentration in simple drainage basin models....................... 15 Base flow ........................................................................................................................................ 16
MODELS OF RUNOFF FORMATION ................................................................ 17

3.1 Runoff coefficient........................................................................................................................... 17 3.1.1 Overall runoff coefficient for single rainfall-runoff events ........................................................ 17 3.1.2 Antecedent precipitation index and coaxial graphical plot ........................................................ 18 3.1.3 The SCS approach .......................................................................................................................... 21 3.2 Models to compute effective rainfall ............................................................................................. 24 3.2.1 Model requirements ................................................................................................................. 24 3.2.2 Runoff coefficient method............................................................................................................... 26 3.2.3 Index approaches, -index............................................................................................................. 26
BASIS AND METHODS OF SYSTEMS HYDROLOGY ..................................... 28

4.1 Definition of system properties ........................................................................................................... 28 4.2 4.3 Unit hydrograph ............................................................................................................................ 31 Analysis and synthesis of the unit hydrograph by the black box method..................................... 32
Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy 4.4 4.5
Hydrology III - II
System operation and instantaneous unit hydrograph.................................................................. 36 Computation of the unit hydrograph from the t-weighting function ......................................... 41
CONCEPTS OF HYDROLOGIC MODELS ........................................................ 43

5.1 Translation and retention.............................................................................................................. 43
5.2 Translation models......................................................................................................................... 43 5.2.1 Linear translation, linear channel............................................................................................. 43 5.2.2 Time of concentration .............................................................................................................. 45 5.2.3 Floodplan method .................................................................................................................... 47 5.2.4 Time-area diagram................................................................................................................... 51 5.3 Reservoir routing models............................................................................................................... 55 5.3.1 Linear reservoir........................................................................................................................ 55 5.3.2 Non-linear, exponential reservoir ............................................................................................. 58 5.3.3 Linear reservoir cascade........................................................................................................... 58 5.4 Parameter estimation for simple conceptual models..................................................................... 61 5.4.1 Moment method for linear model concepts ............................................................................... 61 5.4.2 Storage-outflow relation of single reservoir models .................................................................. 65 5.4.3 Outflow recession curve of the linear, single reservoir .............................................................. 67
COMBINATION OF MODEL CONCEPTS IN DRAINAGE BASIN MODELS ..... 70

6.1 One-component models for direct runoff ...................................................................................... 70 6.1.1 Clark model ............................................................................................................................. 70 6.1.2 Two-reservoir-model (Singh's model)....................................................................................... 71 6.1.3 Influence of precipitation on the model concept........................................................................ 71 6.2 Multi-component models, parallel reservoir cascades .................................................................. 71
FLOOD ROUTING MODELS ............................................................................. 74

7.1 Flood routing.................................................................................................................................. 74
7.2 Simple flood forecasting methods .................................................................................................. 77 7.2.1 Gage relation curve .................................................................................................................. 77 7.2.2 Travel time curve ..................................................................................................................... 79 7.2.3 Prediction of discharge changes ............................................................................................... 80 7.3 Hydraulic approaches to instationary flow ................................................................................... 82 Continuity equation ............................................................................................................................. 82
7.4 Hydrologic flood routing concepts................................................................................................. 88 7.4.1 Basic principles of hydrologic flood routing ............................................................................. 88 7.4.2 Muskingum-model ................................................................................................................... 88 7.4.3 Kalinin-Miljukov method, basic principles............................................................................... 93 7.4.4 Kalinin-Miljukov method, linear reservoir cascade................................................................. 103
PHYSICALLY BASED HYDROLOGICAL MODELS........................................ 106
Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy 8.1
Hydrology III - III
Fundamental principles ............................................................................................................... 106
8.2 Spatial model extension ............................................................................................................... 107 8.2.1 One-dimensional models........................................................................................................ 107 8.2.2 Two-dimensional models ....................................................................................................... 108 8.2.3 Three-dimensional models ..................................................................................................... 109 8.3 Temporal and spatial model resolution ....................................................................................... 110 8.3.1 Temporal resolution ............................................................................................................... 110 8.3.2 Spatial resolution ................................................................................................................... 111 8.4 Modeling of single processes........................................................................................................ 113 8.4.1 Infiltration .................................................................................................................................... 113 8.4.2 Evaporation............................................................................................................................ 114 8.5 Model parameters........................................................................................................................ 115 8.5.1 Parameter estimation.............................................................................................................. 115 8.5.2 Parameter variability.............................................................................................................. 116 8.5.2.1 Temporal variability........................................................................................................... 116
Hydrology III - 1
Fundamental principles of river basin modeling

Scope
1.1
The basis of rational water use and -management is understanding the temporal and spatial characteristics of water flow. Since transport- and transposition processes take place in the water, and large-scale input of man-made substances occurs, the description of water quality is closely related to the description of water flow. According to the actual problem different statements are required that origin from either statistical or deterministic approaches. The main tasks are: Computation of large-scale balances (provides basic information about the water regime and its spatial variations applying long time means and statistical approaches.). Design of water management structures (e. g. flood protection, river development, flood-control reservoirs, carryover storage) usually statistical approaches (e. g. HQ100, NQ10, MQ). Real-time forecasting (e. g. inflow for storage management, flood-forecast service, storage operation) statistical and/or deterministic approach (e. g. forecast by statistical time-series models, prediction of time and height of peak flow computations. Design and assessment of management measures and evaluation of alternatives (e. g. water body development, flood retention, storage operation) usually deterministic approaches (e. g. computation of retention effect of a flood-control reservoir. Process studies for better comprehension of complex hydrologic processes, predominantly deterministic approach.
Another possible subdivision derives from the examination time period: short-term medium-term long-term minutes, hours, days (e. g. flood events) weeks, months (e. g. low water, storage management) years (e. g. mean water, sizing of water power plants)
Hydrology III - 2
1.2
1.2.1
Principal methods of river basin modeling

Statistical methods
Statistical methods are exclusively based on the description of observed values without consideration of the underlying causes. Statistical methods are used for: theoretical probability distributions (distribution functions) of observed values at a certain river cross-section (e. g. extreme-value statistics of floods and low water), time-series analysis and -synthesis of outflow hydrographs, regionalization, Transformation of e. g. outflow characteristics applying regression/geostatistics based on typical features of the gaged and ungaged sites (e. g. size of drained area, inclination, geologic and morphologic features).
The meaningfulness of statistical investigations is dependent on the density of the gage network and the duration of observation. For more detailed information, see lecture Hydrological simulation techniques.
1.2.2
Deterministic methods
Deterministic methods investigate the correlation between cause and effect. It is essential to be able to quantify and mathematically describe the causes and the structure of the affiliated effects. These interrelations may derive from physical laws or from the analysis of short-term observations. Runoff may be attributed to various causes, therefore several different mathematical formulations and mathematical models may be applied. All natural outflow is primarily dependent on precipitation. Precipitation (e. g. rainfall-runoff models, drainage basin model); (see Chapter 2).
Additionally secondary effects of precipitation may be regarded as causes for surface runoff. Volume of groundwater storage (e. g. low-water models), Melting of snow and glaciers (e. g. equations of snow-melt), Discharge of the upper courses (e. g. streamflow models); (see Chapter 7), Operation and management (e. g. storage, discharge, withdrawal).
The results of the methods are as follows: simple outflow characteristics (e. g. time and magnitude of peak flow),
Hydrology III - 3
continuous outflow hydrographs (river basin model with temporal and areal resolution).
1.2.3
Combined methods
The combination of statistical and deterministic model calculations leads to combined methods. Often the natural impacts on a hydrologic system can only be described with statistics, whereas the effects of the system can be derived from physical laws. An example of this is the computation of extreme floods for an area that features only short or no outflow measurements. Therefore an extrapolation to determine rare peak discharges is impossible. Assuming that precipitation observations of sufficient duration at one or several representative gage sites are available, extreme flow may be computed from precipitation. This is accomplished by application of a mathematical discharge model which derives from short- term discharge measurements or physical approaches. The relevant input can be obtained from the statistical distribution of precipitation. In the case of a direct dependence the probability of the effect (discharge) equals the probability of the cause (precipitation). Prerequisite for this method is that the effect for the hydrologic system and the probability distribution of the input are known. The method provides the probability distribution of the output (see Figure 1.1).
Statistical technique (distribution)
Variable propability p1
Deterministic model
Result propability p1
e.g. Precipitation
e.g. Discharge
Figure 1.1:
Combination of statistical and deterministic methods
1.3
1.3.1
Structure of river basin models

Partial models
A hydrologic river basin model generates outflow according to the relevant hydrologic processes by transforming input (precipitation, meltwater supply, evaporation) into output (discharge at the outlet cross-section of the basin). Consequently the model describes the movement and the storage of the precipitated water on the land surface, subsurface and in the stream itself by partial models. The purpose of a mathematical river basin model is therefore the spatial and temporal reproduction of waterflow within a river basin. Thereby the basin is
Hydrology III - 4
broken down in the horizontal and vertical direction, however, the selection of the appropriate subdivision is always dependent on the actual problem. In principle a river basin model contains at least models to compute the soil moisture regime, groundwater and evaporation and reproduces the formation of outflow in the drainage basin, runoff concentration in the water-body system and the temporal course of discharge in the streams of the basin. The following three basic elements that contain the previously mentioned subdivisions are applied: precipitation drainage basin rainfall-runoff model, drainage basin model (see Chapter 2-6), river course streamflow models, flood-routing (see Chapter 5 and 7), natural or artificial storage structures (storage operation model).
The application of partial models based on physics is recommended if river basin models are applied to rarely gaged or ungaged areas, to assess human impact on the water cycle of an area or if models are coupled to evaluate water quality.
1.3.2
Drainage basin models
The drainage basin model serves the determination of discharge/streamflow caused by precipitation within the basin. The computation is related to a single cross-section of the receiving stream which can be considered the outlet cross-section for the drainage basin above it (see Figure 1.2). The actual size of the area is defined by the borders of the drainage basin. Subsequently precipitation is only considered in liquid matter, which means as rain. The conversion of snow melting is reproduced by a suitable model.
Hydrology III - 5
iN(t)
precipitation iN
time t areal precipitation iN Q(t) discharge Q drainage basin time t

discharge Q
Figure 1.2:
Principle of a drainage basin model, determination of streamflow from areal precipitation
First of all, the model must contain a method to convert the punctual precipitation measurements at gage sites to areal precipitation. areal precipitation temporal and spatial distribution of precipitation from the local data of the gage network (see lecture Hydrology I, Chapter 2.5).
The transformation of areal precipitation to streamflow takes place in two phases. Formation of outflow Transformation of precipitation considering evapotranspiration and the retention effect of the basin. Formation of runoff takes place at each point of the drainage basin. However, only a portion of precipitation is transformed into runoff. Concentration of outflow/ streamflow Concentration of the runoff in the outlet cross-section. In this regard it is important to determine the temporal distribution of outflow.
Hydrology III - 6
point model of areal precipitation precipitation model of discharge formation model of discharge concentration discharge at catchment outlet
Figure 1.3:
Basic elements of a drainage basin model
Usually drainage basin models are considered basic units that are not subdivided any further. Therefore each rainfall event must be spread evenly throughout the basin (block rain). Significant variations or partial rainfall is not permitted. Since natural precipitation may only on small-scale areas be considered evenly distributed, this provides the upper limit of the size of the model basin size. If the block rain assumption does not provide sufficient precision for the model, other vertical divisions must be found. Some areas always feature a typical areal rainfall distribution (e. g. mountain rims) that can replace block rain. Subdividing the area by hydrologic characteristics (hydrotopes) is often useful and easily applicable. However, since the complexity of the model increases with the horizontal division it is useful only up to a certain degree. The size of the drainage basin is a decisive factor for the design of the model. The respective topographic and orographic characteristics must be taken into consideration. For river basins in southgerman low mountain ranges models covering an area of up to 500 km are applied.
1.3.3
Streamflow models
Streamflow models reproduce the flow of flood waves in the streams which means instationary open channel flow. The river bed and its piedmonts constitute a retention space which holds the flood wave temporarily back. Continuous retention leads to a flattened flood wave ( wave distortion, see Chapter 7). The streamflow models usually applied in hydrology do not compute the streamflow all the way along the stream, the results are limited to a single control cross-section (outlet crosssection of the examined river section).
Hydrology III - 7
discharge QZ
time t
river course
discharge QA time t
QZ inflow
QA outflow
Figure 1.4:
Principle of a streamflow model, one tributary, (floodrouting)
Three tasks can be distinguished: Flow of a flood wave in stream without lateral inflow or withdrawal; this means one inflow QZ(t) and one outflow Q A(t) and therefore identical water volume (see Figure 1.4). Confluence of several flood waves from different streams; this means several inflows QZ,i(t) and one outflow Q A(t) (see Figure 1.5). Flow of a flood wave in an open channel with punctual or continuous lateral inflow from the traversed intermediate drainage basin. Combination of streamflow model in the open channel and rainfall-runoff model in the traversed drainage basin (see Figure 1.6).
Figure 1.5:
Principle of a streamflow model with several tributaries (flood forecasting)
Hydrology III - 8
precipitation i
time t discharge Q Z QZ inflow
areal precipitation i
time t intermediate catchment
discharge Q A
QA outflow
time t
Figure 1.6:
Principle of a streamflow model with intermediate drainage basin
1.3.4
Complex river basin model
From the previously introduced single components, a complex river basin model including storage spaces may be established. The interfaces of the model components must be selected in a way such that the structure of the model matches the natural formation of outflow (see Figure 1.8). Streamflow gages, water-level gages Confluence of tributaries Points of limited discharge capacity, control sites (bridges, villages, etc.) Points that offer management possibilities (e. g. barrages)
1.4
Model approaches
The character of the individual model is selected with the previous knowledge of the hydrologic system. According to the state of knowledge about the physical laws and the extend of required data the model is selected. Hydraulic mathematical models are based on physical laws (e. g. conservation of mass and energy, model with previous physical knowledge). The model is developed using detailed geometric and hydraulic measurements (e. g. channel cross-sections, bed slopes, roughness coefficients). To describe the complex spatial flow characteristics of a drainage basin hydraulic models are unsuited. The plurality of essential measurement values and the considerable amount of
Hydrology III - 9
calculations limit their application in hydrologic study. Therefore, hydraulic models are applied merely to compute instationary open channel flow. Model concepts are based on simplified physical concepts (e. g. continuity- or storage relations, translation). The complex physical transformation mechanisms are replaced by coarsened model assumptions. The model is defined by a number of parameters (as few as possible) that mostly are derived from only a few and not necessarily very precise geometric and/or hydraulic data or from calibrations against observed values. In this case there is no correspondence between the natural system and the model parameters, just a relation. The application of systemhydrologic models is therefore limited in the case of combined discharge-, transport- and quality analysis. Black-box models contain a merely mathematical description of the transformation characteristics according to systemtheoretical methods (input-output models, models without previous knowledge). Physical principles are completely disregarded. The model is defined by empirical system parameters (see Chapter 4). After the model has been defined the parameters are calibrated against observed outflow values (observed in- and output). Models that use previous knowledge explain the underlying physical processes, models that do not use previous knowledge only model the processes.
time t
time t
intersection
time t
water flow model
time t
rainfall - runoff model
time t
Figure 1.7:
Interfaces of a river basin model
Hydrology III - 10
rainfall-runoff-model catchment model Si intersections water flow model Ki gage site (discharge gage)
water flow mo including catch (lateral inflow)
Figure 1.8:
Structure of a complex river basin model
Many model concepts can be described by methods of system theory. The advantage is that both approaches are based on the same mathematical foundations. Furthermore, due to the connection a direct comparison of the transformation characteristics is possible. Another aspect for the arrangement of model approaches is the relation to the size and shape of the hydrologic system. Models that consider the size of the hydrologic system Defining the model considers the areal extension of flow. The parameters are assigned to spatial-, areal- or linear gridpoints (hydraulic- and some conceptual models). Concentrated models, block models A limited space is considered a hydrologic unit. Flow is therefore artificially concentrated at one point (black box- and most conceptual models).
Hydrology III - 11
1.5
1.5.1
Analysis and synthesis

Calibration of the model against observed in- and output values
Models that either use very little or no previous knowledge at all (black box) obtain their transformation characteristics only from analysis of an output based on a known input. Consequently in- and output data such as precipitation-, in- and outflow hydrographs are required. However, the parameters that derive from characteristic values of the hydrologic system are subject to substantial uncertainties. Calibration against in- and outflow data on the other hand provides a means to suit the model better to the respective aim. This is also valid for hydraulic model approaches. The analysis compares the transformation characteristics of the natural hydrologic system and the model. For the same input the respective outputs should match as closely as possible. By specific optimization the model can be suited to the hydrologic system (see Figure 1.9). Thereby either the values of the hydrographs or characteristic hydrograph values such as the moments are compared. The data flow in the course of analysis and synthesis is displayed in Figure 1.10.
1.5.2
Synthetic streamflow hydrographs
To compute synthetic streamflow hydrographs the input values and the transformation characteristics of the model must be known (cause-transformation-effect). For drainage basin models the input is precipitation, for streamflow models it is inflow.
observed input natural hydrological system e.g. discharge, precipitation observed output e.g. discharge
calibration of comparison parameters
m athem atical model (parameters)
generated output
Figure 1.9:
Calibration of a mathematical hydrologic model
Hydrology III - 12
The synthesis of streamflow hydrographs can be split up into three tasks. Check of historic events If only the input (precipitation or inflow) or output (outflow) of a historic event are known, the unknown values may be found by applying a model. Thus short-term observation data and gaps in observation time-series may be augmented. Estimation of extreme flows For design reasons rare outflow magnitudes of small exceedence probability or probability that outflow falls below the respective value are required. Therefore the input must be connected to a corresponding statistical statement (e. g. 100-year exceedence precipitation as input for a drainage basin model). Methods to determine design precipitation are discussed in the lectures "Hydrology I, Chapter 2.6" and "Hydrologic simulation methods". Prediction of effects of water management projects The effect of water management structures (e. g. storage) can only be assessed applying model calculations. The model simulation is based on historic and/or synthetic outflows.
analysis known input model, analysis of parameters synthesis known input model, known parameters synthesis of output known output
Figure 1.10: Data flow in the course of analysis and synthesis
CHAIR OF HYDROLOGY AND GEOHYDROLOGY
Hydrology III - 13
2
2.1
Structure of drainage basin models

Formation of runoff and runoff concentration
Only a fraction of the precipitation that falls above a drainage basin eventually appears as runoff. Already through the course of the precipitation event evaporation returns a fraction of the water back to the atmosphere. The portion of precipitation that later appears as runoff (effective rainfall) infiltrates dependent on intensity and duration of the precipitation event into different stratums of the drainage basin. Usually flow is separated into three components of roughly uniform character (DIN 4049, Part 1, see Figures 2.1 and 2.2). Surface runoff The portion of flow that moves into the receiving stream on the surface. Interflow The portion of flow that flows through the subsurface towards the receiving stream. Interflow may be further subdivided into delayed and fast interflow (unsaturated soil zone). Groundwater flow The portion of flow that flows delayed towards the receiving stream from the groundwater body (saturated soil zone). The total of surface runoff and fast interflow is termed direct runoff. Base flow is formed from groundwater flow and delayed interflow. The formation of runoff is reproduced in the model as a two-phase process. Separation of precipitation into two parts: The first part, called net precipitation or effective rainfall contributes directly to the surface runoff. The other part is composed of losses to interception, evaporation, depression storage, and regional storage. Distribution of the effective rainfall into the three components of flow.
Hydrology III - 14
outflow Q [m3/s]
flood increase vertex flood hydrograph Q
flood decrease
surface runoff QO interflow QI
direct outflow Q
base flow assumed drought outflow hydrograph
time t [h]
Figure 2.1:
Separation of flow components of a flood wave
evaporation transpiration interception surface runoff QO areal precipitation iN effective precipitation iNe interflow QI direct outflow QD overall outflow Q
infiltration
base flow QB
formation of outflow
concentration of outflow
Figure 2.2:
Separation of areal precipitation in the course of flow formation components of runoff concentration
Hydrology III - 15
2.2
Formation of outflow and runoff concentration in simple drainage basin models
Simple drainage basin models only consider two flow components (see Figure 2.3). Direct runoff, QD Portion of the flood wave that arises directly and quickly from a precipitation event. Base flow, QB Portion of the outflow that is not directly concerned with the flood event. Base flow is a constant flow that changes only slowly.
Both components are connected to their causes and are treated separately in the setup of the model. The actual precipitation event causes direct runoff, whereas base flow is dependent on the regional soil moisture and the groundwater volume and pertains to the long-term precipitation history (previous precipitation). Simple models compute outflow by separating it into two components (see Figure 2.3). effective rainfall or net precipitation, iNe Precipitation that eventually appears as direct flow hydrograph at the basin outlet. losses, iV Combination of all components that are not included in the direct runoff (evaporation, regional storage, etc.).
formation of outflow concentration of outflow
precipitation losses iNV
areal precipitation iN
effective precipitation iNE
direct outflow QD overall outflow Q
precipitation history, initial soil moisture
base flow QB
Figure 2.3:
Flow formation and runoff concentration in simple drainage basin models
Hydrology III - 16
2.3
Base flow
Base flow is the portion of outflow that is not directly associated with the precipitation event. In comparison to direct runoff base flow shows only small magnitudes. Its effects on the flood wave, especially on the peak flow are only marginal. Assuming the base flow hydrograph as a straight line provides sufficient precision, it may even be regarded as constant. The separation is carried out graphically by a horizontal or slightly inclined straight line from the starting point of the flood wave (see Figure 2.4). The starting point is indicated by a recognizable increase of flow. Subtraction of the base flow QB from the overall flow Q provides the direct runoff QD: at the beginning and at the end the direct runoff hydrograph has a value of zero.
QD ( ti ) = Q ( ti ) QB ( ti )
(2.1)
discharge Q [m3/s]
start
time t [h]
Figure 2.4:
Separation of base flow by a a) horizontal or b) slightly inclined straight line
Computation of synthetic outflow is conducted separately for base flow and direct runoff. Adding the two components provides the overall outflow hydrograph. Considering accuracy, the base flow is only of minor importance. Assigning the mean outflow at dry-weather conditions to base flow is of sufficient precision. (For further investigations the analysis of the coaxial graphical plot is recommended, see Chapter 3.1.2).
Hydrology III - 17
3
3.1
3.1.1
Models of runoff formation

Runoff coefficient
Overall runoff coefficient for single rainfall-runoff events
The overall runoff coefficient is the volumetric ratio of direct runoff to areal precipitation. It is the fraction of a precipitation event that contributes to runoff.
VD hNe = VN hN
[-] [m3] [m3] [mm] [mm] overall runoff coefficient volume of direct runoff volume of precipitation overall effective depth of precipitation of the event overall depth of precipitation of the event
(3.1)
VD VN hNe hN
Usually the direct runoff hydrograph QD(ti) is plotted as a succession of linear interpolations between discrete values. The first (i = 0) and the last (i = k) always equals zero. Consequently the discrete integration is reduced to the trapezoidal algorithm.
VD = t 3600 QD ( ti )
i =1
k 1
(3.2)
QD 3600
[m3/s] [s/h]
direct runoff conversion factor
The volume of the observed areal precipitation is the product of the overall depth of precipitation and the size of the drainage basin. VN = 1000 AE hN AE hN 1000 [km2] size of the drainage basin [mm] overall depth of precipitation 3 2 [m /(km mm)] conversion factor (3.3)
Hydrology III - 18
3.1.2
Antecedent precipitation index and coaxial graphical plot
The overall runoff coefficient of a rainfall-runoff event is related to the duration of the precipitation event, the overall depth of precipitation, the soil moisture and the season. A measure for the initial soil moisture is the Antecedent Precipitation Index (API) hVN. The seasonal variation of evapotranspiration and the detention storage may be considered by a continuous array of numbers nW assigned to each week.
hNe = hN hNe = f ( hN , TN , hVN , nw )
(3.4)
=
hNv hN hNe TN hVN nW
hNe hN
[mm] [mm] [mm] [h] [mm] [-]
(3.5) precipitation losses overall depth of precipitation of a single event (e. g. design precipitation) effective depth of precipitation of a single event duration of the precipitation event antecedent precipitation index week number (of the year)
The antecedent precipitation index is based on the assumption that the soil moisture after a precipitation event decreases exponentially. The more time elapsed between precipitation events, the smaller is its impact and vice versa. The weighted daily depths of precipitation of a limited period of time preceding the actual event are taken into consideration. The individual weights are always smaller than 1.
Hydrology III - 19
precipitation hN [mm]
t0-n
antecedent prec. index hVN [mm]
t0-i
time t [d]
hVN (t0)
t0-n
t0-i
time t [d]
Figure 3.1:
Antecedent precipitation index
The antecedent precipitation index is computed as:
hVN ( t0 ) = hN ( t0i ) i
i =1
(3.6)
t0 hVN hN n
[d,h] [mm] [mm] [-] [-]
start of precipitation event, antecedent precipitation index, daily depth of precipitation, number of days preceding the event, empirical weighting factor < 1.
Usually the impact is limited to a time of 30 days. Preceding precipitation is not considered. Empirical investigations have suggested a weighing factor of = 0.9. The precipitation losses hNV are displayed in coaxial graphical form (see Figure 3.2).

week number
Hydrology III - 20
antecedent precipitation index hVN [mm]
depth of prec. losses hNV [mm] (calculated)
duration of precipitation TN [h]
depth of precipitation hN [h]
Figure 3.2:
Coaxial graphical plot of precipitation losses hNV for a given drainage basin, reading example
The interdependences may be found applying multiple non-linear regression or graphically by trial and error. The equation displayed below indicates a possible non-linear regression to represent the coaxial diagram as a formula, however, here instead of the antecedent precipitation index hVN the base specific discharge qB is used, and instead of the week number nW the month M is used to compute the precipitation losses hNV.
hNV
hN eC qB e DTN A + B sin ( M 4 ) 6 = hN + hN e E hN eC qB e DTN A + B sin ( M 4 ) 6

[mm] 2 [l/s/km ] [-] [h] precipitation losses (regional storage) base specific discharge at the beginning of the event month duration of precipitation
depth of prec. losses hNV [mm] (observed)
(3.7)
hNV qB M TN
Hydrology III - 21
A...E
[-]
parameters
3.1.3 The SCS approach

According to the DVWK (1984), for the estimation of effective or net precipitation in the case of rain storm events and small drainage basins the application of the SCS approach developed by the U.S. Soil Conservation Service is recommended. This method considers effective rainfall hNe as a function of the depth of precipitation hN and a curve number CN dependent on the drainage basin: 5080 + 50.8 hN CN hNe = 20320 203.2 hN + CN
2
(3.8)
The CN value again is a function of the soil type, land cover, cropping practice and the antecedent moisture condition, which is dependent on the antecedent precipitation of the preceding 5 days and the season. Table 3.1 displays CN-values for various soil types and land cover/ cropping practice for antecedent moisture condition II. From Table 3.2 the current antecedent moisture condition may be taken. In case it deviates from II, the final CN-value may be determined applying Figure 3.3.
Hydrology III - 22
Table 3.1: Land use
CN-Values for antecendent soil moisture condition II CN for hydrologic soil group A B 86 80 73 76 69 79 58 66 60 55 100 C 91 87 79 84 79 86 71 77 73 70 100 D 94 90 82 88 84 89 78 83 79 77 100
Bare soil Root crops, wine Wine (terraced) Corn, forage plants Pasture (normal) (barren) Meadow Forest (open) (medium) (dense) impervious areas Hydrologic soil group A: Hydrologic soil group B:
77 70 64 64 49 68 30 45 36 25 100
Soils with great infiltration potential, even after antecedent wetting (e. g. thick sand and gravel stratums) Soils with medium infiltration potential, thick and moderately thick stratums, fine or moderately coarse texture (e. g. moderately thick sand stratums, loess, loamy sands) Soils with low infiltration potential, sorts of fine or moderately coarse texture or with impervious layers (e. g. thin sand stratums, sandy loams) Soils with considerably low infiltration potential, clay, thin soil stratums overlying impervious layers, soils with constantly high groundwater stage.
Hydrologic soil group C:
Hydrologic soil group D:
Hydrology III - 23
Table 3.2:
Current antecedent moisture condition accumulated depth of precipitation within the preceding 5 days in unit [mm] vegetation period other < 15 15 - 30 > 30
antecedent moisture condition
I II III
< 30 30 - 50 > 50
CN for soil moisture class I, III
CN for soil moisture class II
Figure 3.3:
CN for antecedent moisture condition I and III cross-linked to antecedent moisture condition II
This method should only be applied for rain storm events. Experiences in the past have shown that for depth of precipitation lower than 50 mm the method underestimates effective rainfall. Modifications of equation (3.8) try to compensate this.
Hydrology III - 24
3.2
3.2.1
Models to compute effective rainfall

Model requirements
The current formation of outflow is dependent on the intensity iN, the duration tN and the variations (e. g. breaks) of precipitation. Due to depression storage and wetting the initial losses (or after breaks) are larger than during periods of intensive precipitation. In hydrologic practice usually two methods are applied. Runoff coefficient method In the course of a precipitation event only a portion of precipitation is transformed into direct runoff. The runoff coefficient is the ratio of effective or net precipitation iNe(t) to the observed precipitation iN(t).
iNe ( t ) = iN ( t )
(3.9)
In principle the runoff coefficient is a function of precipitation intensity and -duration.
= f ( iN ( t ) , t )
(3.10)
However, simple models consider the runoff coefficient as constant throughout the whole precipitation event (see Chapter 3.2.2). Index approaches Only the precipitation that is equal to or more than a certain infiltration capacity iv ( = losses) contributes to direct runoff.
Hydrology III - 25
i ( t ) iNv iNe ( t ) = N 0
prec. intensity i [mm/h]
for iN ( t ) iNv ( t )
for iN ( t ) > iNv ( t )
time t [h]
time t [h]
const. coeff. of discharge
phi - index method
loss variable with time
time t [h]
Figure 3.4:
Models to determine effective rainfall a) constant runoff coefficient, runoff coefficient method b) constant loss ratio, -index method c) loss ratio decreasing exponentially
base flow Q B (t)
overall outflow Q(t)
separation of base flow
direct outflow Q D (t)
areal precipitation iN (t)
formulation for effective prec.
effective precipitation iNe (t)
precipitation losses iNv (t)
Figure 3.5:
Determination of outflow formation for simple drainage basin models, sequence and data flow
Hydrology III - 26
The initial rate is always larger than at the end of a precipitation event. Usually the temporal development of losses can be represented by an exponentially decreasing function (see Figure 3.4c).
iNv ( t ) = f ( iN ( t ) , t )
(3.12)
The simplest model features a constant loss rate, which is only a very rough approximation (see Chapter 3.2.3). The analysis of rainfall-runoff events is carried out in established steps (see Figure 3.5). Separation of base flow (linear course) Computation of overall runoff coefficient from the volumes of direct runoff and areal precipitation Computation of effective rainfall (runoff coefficient method or index approaches)
3.2.2 Runoff coefficient method

The runoff coefficient remains constant throughout the entire course of the precipitation event (see Figure 3.4a). It corresponds to the overall of discharge (see Chapter 3.1). Consequently computation of effective rainfall is reduced to the simple formula displayed below.
iNe ( ti ) = iN ( ti )
(3.13) effective rainfall intensity in time interval t observed precipitation intensity in time interval t overall runoff coefficient
iNe(ti) iN(ti)
[mm/h] [mm/h] [1]
3.2.3 Index approaches, -index

A constant loss rate in the course of a precipitation event is referred to as -index.
iV ( t ) = const. =
(3.14) constant loss rate, -index
[mm/h]
The portion of precipitation that exceeds the -index is the effective rainfall.
Hydrology III - 27
i ( t ) iNe ( t ) = N 0
for iN ( t )
for iN ( t ) >
The -index must be determined by step-by-step iterations since negative precipitation is impossible. The iteration provides a constantly increasing -index; the procedure is repeated until it equals the known effective rainfall.
Hydrology III - 28
Basis and hydrology
methods
of
systems
4.1 Definition of system properties

A system is a distinguished arrangement of interrelated structures (DIN 19226). Each system features an entrance where the cause (input) ui affects the system, and an exit where the effect (output, system answer) vi occurs (see Figure 4.1). The interrelations of these values describe the system. The system-operation establishes a definite relation between input and output.
input output
system load result
Figure 4.1:
System with several in- and output variables (input vector ui(t) and output vector vi(t))
The simplest case is the definite relation between one output magnitude v and one input magnitude u (see Figure 4.2), e. g. effective rainfall - direct runoff. The mathematical relation between in- and output can be represented by
v ( t ) = {u ( t )}
where, u(t) v(t) time-dependent input signal, time-dependent output signal, system operator.
(4.1)
system
input variate output variate
Figure 4.2:
System with one input magnitude u(t) and one output magnitude v(t)
Hydrology III - 29
The system operation may be defined by certain regularities that allow a classification of the model systems. As an example these regularities are applied to the precipitation-runoffrelation in drainage basins. The input magnitude is the areal effective rainfall expressed as intensity in unit [mm/h], the output magnitude is the outflow at the basin outlet in unit [m3/s] (see Figure 4.3). It is essential to estimate effective rainfall correctly, as the system input affects the quality of the computed unit hydrograph. The drainage basin is an open, dynamic system.
A system is termed dynamic if at any time t1 the output signal v(t1) is not merely dependent on the input signal ui (t) at the same time t1 but also from preceding input signals u(t) for t < t1. In physical regard this feature corresponds to a temporal storage of the input magnitude which can be regarded as a system memory. Drainage basins, open channels and storage structures can be considered dynamic systems, because they temporarily store outflow and deliver it later and damped (retention). Figure 4.3a displays how a precipitation event of short duration TN is discharged as a flood wave of much longer duration Tb. Theory of proportionality Any input signal multiplied by a constant C produces an output signal multiplied by the same constant.
{C u ( t )} = C {u ( t )}
(4.2)
The effective rainfall displayed in Figure 4.3b is double the amount as in Figure 4.3a and produces a doubled outflow hydrograph. The duration Tb of the outflow hydrograph remains constant. Theory of superposition The system answer to accumulated input signals equals the total of the single output signals.
{u1 ( t ) + u2 ( t )} = {u1 ( t )} + {u2 ( t )}
(4.3)
Theory of linearity The combination of the theory of proportionality and superposition provides the theory of linearity.
{C1 u1 ( t ) + C2 u2 ( t )} = C1 {u1 ( t )} + C2 {u2 ( t )}
(4.4)
Theory of time invariance The system operation is not time-dependent. Shifting an input signal by the time interval T results in an output signal shifted by the same time interval without changing the signal itself (see Figure 4.4c). Tb is preserved.
Hydrology III - 30 (4.5)
{u ( t T )} = v ( t T )
Applying the theory of time-invariance and proportionality, consecutive input signals of different intensity can separately be assigned to individual output signals (see Figure 4.4d). The theory of superposition allows to overlay the individual signals to one.
intensity iNe [mm/h] intensity iNe [mm/h] unit hydrograph of discharge precipitation principle of linearity precipitation
discharge discharge QD [m3/s] discharge QD [m3/s]
discharge
time
time
Figure 4.3:
a) dynamic system operation, unit streamflow hydrograph b) theory of proportionality

intensity iNe [mm/h] intensity iNe [mm/h] principle of time invariance precipitation
Hydrology III - 31
principle of superposition precipitation
discharge
discharge
time
discharge QD [m3/s]
discharge QD [m3/s]
time
Figure 4.4: a) Theory of time-invariance b) Theory of superposition
4.2
Unit hydrograph
The unit hydrograph method is a linear, time-invariant model to determine outflow.
effective precipitation iNe*Ae input
drainage basin gE system (linear, dynamic, time-invariant)
direct outflow QD output
Figure 4.5:
Runoff concentration model as linear, dynamic, time-invariant system
To describe the system operation it is sufficient if the output function that pertains to one constant input signal is known. Using the theory of linearity and superposition the system answer to any series of discrete input signals can be determined. It is useful to relate the characteristic output function to a constant unit input of the duration t and the magnitude one (see Figure 4.3a). This function is referred to as discrete weighting function g(t,ti) with
Hydrology III - 32
reference interval t. Note that not the intensity, but the volume of the unit input possesses the value 1 and therefore the intensity is 1/t. The reference time interval is a defining feature of the weighting function g(t,t)i. For a linear and time-invariant drainage basin model the weighting function g(t,ti) is replaced by the unit hydrograph gE(t,ti) that considers the different dimensions of precipitation and outflow and the size of the drainage basin. The unit hydrograph describes the system operation of effective rainfall to direct runoff (see Figure 4.5). The unit hydrograph is a characteristic outflow hydrograph of a surface drainage basin that develops from constant effective rainfall of uniform distribution of 1 mm in depth and defined length (DIN 4049 Part 1). Thereby, effective rainfall is expressed by the depth, not by the intensity of precipitation.
4.3
Analysis and synthesis of the unit hydrograph by the black box method
The determination of the unit hydrograph for a system can be achieved directly by the analysis of the observed rainfall-runoff events. This approach ignores the physical structure of the system and applies only the system properties and is referred to as black-box. First, a reference time interval t is chosen that is valid for the discretion of all time-related data. When applying the computer program, a time interval that splits the flood hydrograph into 30-50 units is recommended. Separation of the base flow ( see Chapter 2.3) provides the system input, the direct runoff QD(ti). Subsequently the overall runoff coefficient and the effective rainfall hNe(ti) is computed (see Chapter.3.2). The flood hydrograph is composed of the hydrographs of the individual precipitation intervals (see Figure 4.6). Each time interval ti provides (assuming that the unit hydrograph is given) the direct runoff (synthesis). The overall outflow function that results from the overall inflow function may be developed by superposition of the hydrographs that pertain to the individual precipitation events. This procedure is referred to as superposition. Subsequently the equation system for a simple example enclosing n = 3 effective rainfall ordinates and m = 5 ordinates of the unit hydrograph is displayed:
Hydrology III - 33
Q1 = N1 G1 Q2 = N1 G2 + N 2 G1 Q3 = N1 G3 + N 2 G2 + N 3 G1 Q4 = N1 G4 + N 2 G3 + N 3 G2 Q5 = N1 G5 + N 2 G4 + N 3 G3 Q6 = Q7 = where Qi = QD(ti) Ni = hNe(ti) Gi = gE(t,ti) [m3/s] [mm] [m3/(smm)] direct runoff with time ti, effective rainfall in the interval between ti-1 and ti, unit hydrograph at time ti. N 2 G5 + N 3 G4 N 3 G5 (4.6)
The equation system may be reduced to a differential equation, the so-called discrete equation of superposition. For a given time ti:
QD ( ti ) = g E ( t , tk ) hNe ( ti k +1 )
k =1
(4.7)
Hydrology III - 34
areal and effective precipitation
precipitation intensity
time t [h] principle of superposition
direct runoff
time t [h
Figure 4.6: Theory of superposition of the unit hydrograph method
Since the initial and terminal value of the direct runoff hydrograph and the unit hydrograph always equal zero, these times are not considered. Therefore the number of discrete values and intervals are
o = m + n 1 m = o n +1
o n m
[1] [1] [1]
number of discrete values of the unit hydrograph method QD 0, number of effective rainfall intervals hNe, precipitation duration/t, number of discrete values of the unit hydrograph gE 0.
For the analysis the ordinates of the unit hydrograph must be computed from the linear equation system. For m + n - 1 equation and m unknown values the equation system is n - 1 times overdefined. The optimum solution provides computed runoff QD,ber as close to the observed runoff QD,gem as possible. This is accomplished by applying the method of the lowest square error.
m + n 1 i =1
(Q
D ,ber
( ti ) QD, gem ( ti ) )
= Minimum !
(4.9)
The solution of an overdefined equation system by the method of the lowest square error is available in closed form (BRONSTEIN-SEMENDJAJEV p.513-514). For a lower number of unknown values the solution may be found by trial and error. A test of plausibility derives from the definition of the unit hydrograph as the direct runoff hydrograph caused by the effective rainfall of 1 mm in depth within a time interval t. Therefore
3.6 t m g E ( t , ti ) = 1 AE i =1
AE 3.6 [km ]
2
(4.10) size of drainage basin, conversion factor, time interval.
mm km 2 s ] h m3 [h]
[
The runoff concentration within a drainage basin may only approximately be represented by a linear system. In the individual case the unit hydrograph method provides satisfactory results even though it features a certain variation when analyzing several rainfall-runoff events of same duration t. The generally applicable unit hydrograph is the mean of all obtained unit hydrographs while still considering the test of plausibility.
Hydrology III - 36
Oftentimes no suitable runoff- or precipitation measurements for the drainage basin are available. A number of methods exist to compute the unit hydrograph if measurements are unavailable; two of them are commonly used: Direct application of the unit hydrograph of a similar drainage basin. Selection according to the drainage basin properties. From similar, well-observed drainage basins the dimensionless system may be taken. The reference drainage basin is selected according to size, geology, slope, soil type, characteristic values of the receiving stream and land use (Literature: DVWK-Merkbltter 1982, 1988 catalogue of system operations). Formulation of a synthetic unit hydrograph by regionalization of the drainage basin properties: Relations between the parameters of a system operation and the properties of a drainage basin can be established. However, it is essential to examine a plurality of drainage basins and to apply weighting functions that can be described analytically (e. g. triangular hydrographs, hydrograph of constant rise and exponential recession or the gamma-function). Important parameters of the system operation are the time of rise, the peak and various recession parameters etc. The correlation of unit hydrograph and basin parameters can be described by e. g. regressions. The Geomorphologic Unit Hydrograph (GUH) is a physically based model. It makes use of stream network characteristics to determine the probability of occurrence of individual water particles at the basin outlet (Literature: SIVAPALAN et al. 1990).
4.4
System operation and instantaneous unit hydrograph
The system operation of linear, time-invariant systems, as outlined in Chapter 4.2, can be represented by a characteristic answer to a constant input signal of duration t. Other typical input signals exist that define the system by their affiliated output function. For several linear, time-invariant conceptual models (see Chapter 5) The output functions may be determined analytically. A special system input is the unit step function (t) (see Figure 4.7): 0 (t ) = 1 for t < 0 for t 0 unit step function (4.11)
(t)
[1]
The unit step function is equivalent to an on-switch at time t = 0. The affiliated system answer is termed s-curve h(t). For a dynamic system that features a storage effect, the system operation transforms the input to an s-shaped, lagged and dampened outflow hydrograph (see Figure 4.7). In practice, this may occur if a river or channel weir is suddenly opened. For
Hydrology III - 37
most hydrologic systems the system answer has the value zero at the time t = 0 which simplifies the subsequent formulas.
{ ( t )} = h ( t ) with h ( t ) = 0
h(t)
u E (t) 1
for t 0
(4.12)
[1]
system operation
Input u (t): unit jump E (t) 0 for t < 0 E (t) =
0 v h (t) 1
1 for t > 0 t Output v (t): system operation h (t) 0 for t < 0 0 - 1 for 0 <= t < infinity 1 for t = infinity
h (t) = 0 t
Figure 4.7:
unit step function (t) and system operation h(t)
In case the system operation h(t) is known, the output function for a constant input signal of the volume 1, duration t and the intensity 1/t may be computed. The jump function of the intensity 1/t is superpositioned by a negative jump function lagged by -1/t (see Figure 4.8a). This provides a constant pulse of the volume 1 and duration t.
u (t ) = 1 1 1 ( t ) ( t t ) = ( ( t ) ( t t ) ) t t t
(4.13)
Applying the theory of linearity and time-invariance the output function develops from the superposition of the respective system operations (see Figure 4.8a). The output flood related to a pulse of the volume 1 and duration t is the known t-weighting function.
Hydrology III - 38
v ( t ) = {u ( t )} =
1 1 h ( t ) h ( t t ) t t
1 = ( h ( t ) h ( t t ) ) = g ( t , ti ) t
g{t,ti) [1/h]
(4.14)
t-weighting function
For an infinitesimally small time t the input function becomes a theoretical function known as the needle pulse, instantaneous unit function or Diracs superposition pulse function and contains one unit volume of input. It corresponds to a generalized differentiation of the unit step function (see Figure 4.8b). According to the common rules of differentiation, single steps cannot be differentiated, hence the generalized differentiation for technical systems is introduced (The expression is not a numeric value and hence according to the common rules of differentiation no limit exists).
( t ) = lim
0 ( t ) = 0
1 d ( t ) ( t ) ) = ( t ) ( t 0 t dt
for t 0 for t = 0 for t 0
(4.15)
(4.16)
(t)
[1/h]
needle pulse function
Hydrology III - 39
pulse
dt pulse magnitude 1
dt weighting function
needle pulse
weighting function for h (t = 0) = 0 :
Figure 4.8:
a) development of the t-pulse and the t-weighting function g(t,t) b) needle pulse function (t) and weighting function (0,t)
The affiliated output function derives in the same way from the system operation by a limit approach.
v ( t ) = { ( t )} = lim 1 ( h ( t ) h ( t ) ) = g ( 0, t ) t 0 t
for h ( +0 ) = 0 , or h ( t ) = g ( 0, t ) dt
0 t
d g ( 0, t ) = h ( t ) dt
(4.17)
g(0,t)
[1/h]
weighting function
Hydrology III - 40
The system output that results from a needle pulse corresponds to the differentiation of the system operation h(t) and is termed the weighting function g(t) (with a reference time interval
t that approaches zero the expression g(0,t) appears, see Figure 4.8b).
The discrete finite difference form of the theory of superposition is transformed to the analytical integral of superposition (integral of convolution, or, in another expression also termed Duhamel integral). Discrete superposition in finite difference form with the weighting function
g(t,ti) (u(ti) = const. Within time interval ti):
v ( ti ) = g ( t , tk ) u ( ti k +1 ) t
k =1 i
(4.18)
= g ( t , ti k +1 ) u ( tk ) t
k =1
where u(ti) = 0 for ti < 0. Analytical superposition with the weighing function g(0,t):
v ( t ) = g ( 0, t ) u ( t t ) dt
0 t
(4.19)
= g ( 0, t t ) u ( t ) dt
0
where u(t) = 0 for t < 0. Model concepts try to represent these processes in the natural system by simple mathematical models. The system is identified. The parameters are determined to approximate the model input as closely as possible to the represented process. Most model concepts assign linear, time-invariant system operation to the hydrologic system or the individual system components. The system operation h(t) or the weighting function g(0,t) may derive from the model approach (see Chapter 5). Since usually discrete in- and output data are available, the superposition must be accomplished with the differential equation and discrete values. Also the t-weighting function g(t,ti) must be determined.
Hydrology III - 41
g ( t , ti ) =
1 ( h ( ti ) h ( ti 1 ) ) t t ti 1 1 i = g ( 0, t ) dt g ( 0, t ) dt t 0 0 1 i = g ( 0, t ) dt t ti1
t
(4.20)
g(t,ti) [1/h] t,t [h]
t-weighting function auxiliary variable for the integration
4.5
Computation of the unit hydrograph from the tweighting function
the input variable u(t) on the drainage basin is the volume of the effective or net areal precipitation. The output variable v(t) is the direct runoff at the basin outlet.
u ( t ) = iNe ( t ) v ( t ) = QD ( t )
u(t) v(t) iNe(t) AE QD(t) 3.6 [m3/s] [m3/s] [mm/h] [km2] [m3/s] [
AE 3.6
(4.21)
system input system output intensity hydrograph of the effective rainfall size of the drainage basin direct runoff hydrograph
mm km 2 s ] conversion factor h m3
The in- and output variables, here direct runoff and effective rainfall, are affiliated by the equation of superposition.
Hydrology III - 42
v ( ti ) = g ( t , tk ) u ( ti k +1 ) t
k =1 i
QD ( ti ) = g ( t , tk )
k =1
AE iNe ( ti k +1 ) t 3.6
(4.22)
By definition of the unit hydrograph (according to DIN 4049 Part 1), the input is not the intensity of effective rainfall iNe(ti) but the effective depth of precipitation hNe(ti) that pertains to the respective interval. For reasons of clarification it can be termed hNe(t,ti).
hNe ( ti ) = iNe ( ti ) t
(4.23)
This formula and the t-weighting function provides the unit hydrograph that considers the size of the drainage basin and the finite difference dimensions. g E ( t , ti ) = g ( t , ti )
i
AE mm 1 3.6 t
k =1
(4.24)
QD(ti) hNe(ti)
[m3/s] [mm]
direct runoff at time ti effective rainfall within the time interval ti - 1 to ti unit hydrograph of the drainage basin at time ti
m3 gE(t,ti) [ ] s mm
Hydrology III - 43
5
5.1
Concepts of hydrologic models

Translation and retention
The complex processes that occur in the course of runoff concentration and instationary open channel flow are in hydrologic practice replaced by simple model concepts. The two main groups are: translation and retention.
The application of an individual model is usually not satisfactory in terms of accuracy. Only the combination of translation and retention yields results that are close enough to the natural conditions (see Chapter 6).
5.2
5.2.1
Translation models
Linear translation, linear channel
Translation is a shifting in time. The shape of the input signal u(t) reappears lagged by the translation time Tt as the output signal v(t). If the translation time Tt is a constant and thus not dependent on the magnitude and the temporal occurrence of the input signal (time-invariant), the system operation is called a linear translation (see Figure 5.1).
v ( t ) = u ( t Tt )
(5.1) time-related input signal time-related output signal translation time
u(t) v(t) Tt
[div.] [div.] [h]
Hydrology III - 44
u, v
u (t)
v(t) = u(t-Tt)
u1
v1=u1
dt Tt
dt dt
time t
Figure 5.1:
Input and output of a linear translation system
In the finite difference form Tt should be an integer multiple of the time interval t.
v ( ti ) = u ( ti Tt ) with Tt = m t
(5.2)
The linear channel is a flow model where the flow velocity v is constant. The effects of instationary channel flow (flow retention v(t) and local hydraulic conditions (cross-section A(x), roughness kSt(x)) are ignored.
v ( A ( x ) , kSt ( x ) , h ( x ) , h ( t ) , t ) = v = const.
v A kSt x h t [m/s] [m2] [m1/3/s] [m,km] [m] [h] flow velocity cross-section coefficient of roughness stream coordinate depth of water time
(5.3)
The time of flow Tt is the time required for a water particle to move a certain distance lF (DlN 4049 Part 1). For a linear channel the time of flow equals a constant translation time.
Tt = 1 lF 3.6 v
(5.4) time of flow flow distance, channel length constant flow velocity
Tt lF v
[h] [km] [m/s]
Hydrology III - 45
3.6
kms ] mh
conversion factor
The system operation of linear channels is a linear translation. Therefore it is a linear, timeinvariant model system where the weighting function g(0,t) is known.
g ( 0, t ) = ( t Tt )
(5.5) weighting function of the linear translation, needle pulse function (see Chapter 4.4).
g(0,t) (t)
[1/h] [1/h]
The outflow QA(t) of a linear channel is the inflow QZ(t) lagged by the time of flow Tt.
QA ( t ) = QZ ( t Tt )
(5.6) inflow of a linear channel, outflow of a linear channel.
QZ QA
[m3/s] [m3/s]
The linear channel is only a very rough approximation of the instationary outflow since it does not consider retention. Hence it may only be applied as a component of a combined model.
5.2.2
Time of concentration
The water particles that fall as precipitation in a drainage basin require different time to reach the basin outlet dependent on their position in the watershed. The time of concentration Tc is defined as the travel time of a water particle from the hydraulically most remote point in the basin to the outflow location, or, in other words, the time until the whole basin contributes to the outflow. From the mean flood flow velocity Specht suggested the rough estimation below.
TC =
vm Tc lF
1 h lF 2 km
[m/s] [h] [km]
to
1 h lF 3 km
means flow velocity from 0.6 to 0.8 m/s concentration time, length of the main feeder.
(5.7)
For mountainous regions, Kirpich established an empirical formula identical to the equation of the U.S.-Soil Conservation Service. A uniform procedure is applied to determine the mean inclination IF of the longitudinal section of the drainage basin. A straight line represents the slope of the basin in a way that the two regions A1 and A2 are of equal size (see Figure 5.2).
Hydrology III - 46
time of concentration according to Kirpich (coefficients not dimensionless)

l TC = 0.06625 F I F
0.77
IF =
hF lF 103
(5.8)
time of concentration according to the U.S.-Soil Conservation Service (coefficients not dimensionless)
l3 TC = 0.868 F hF
Tc hF lF IF [h] [m] [km] [1]
0.383
(5.9) time of concentration, relevant difference in elevation, relevant path of flow, mean inclination of the main channel or the longitudinal section of the basin
elevation drainage basin boundary elevation at basin boundary
drainage basin boundary outlet
outlet elevation at outlet length in cross-section
Figure 5.2:
Time of concentration, determination of mean inclination by equalizing the regions in the longitudinal section of the drainage basin (section that provides the longest extension lmax of the basin measured from the outlet)
Hydrology III - 47
5.2.3
Floodplan method
The floodplan method is a simple runoff concentration model based on linear translation. The water network of a drainage basin consists of linear channels of equal flow velocity. The drainage basin is approximated by a rectangular plain. The mean inclination IF and the time of concentration T c are determined according to Chapter 5.2.2. Precipitation is considered evenly distributed (block precipitation, see Figure 5.3). The drainage basin dewaters with a constant flow velocity vm., consequently a travel time Tt and an area At can be assigned to a point with the distance x from the outlet (Ai = Ai(Ti), see Figure 5.3). Assuming the precipitation being a jump function, as time elapses, progressively distant areas At contribute to the outflow (see Figure 5.4). In other words, up to a certain time t only the portion of precipitation that fell within the area A with a corresponding travel time t = Tt contributes to the outflow. As soon as the time of concentration Tc is exceeded (t > Tc) the whole basin AE dewaters and a constant outflow rate is acquired (see Figure 5.4). On the rising limb (t Tc) the area At(t) that contributes to the outflow is constantly increasing with time. t AE TC At ( t ) = AE At(t) AE [km2] [km2] for 0 t TC for t > TC area contributing to outflow, size of the drainage basin.
(5.10)
effective precipitation
drainage basin
Figure 5.3:
Floodplan-model, area At(Tt) that contributes to outflow at time Tt

precipitation direct runoff [m3/s] outflow
Hydrology III - 48
eff. prec. iNe [mm/h]
time t [h]
time t [h]
Figure 5.4:
System operation of the floodplan for precipitation in form of a jump function, temporal course of the area contributing to outflow At(t), system operation function
Since the floodplan method is based on linear translation it is a linear, time-invariant system. The function of the ratio of the contributing area to the size of the drainage basin At(t)/AE represents the system operation function h(t) of the floodplan method (see Figure 5.4). t for 0 t TC At ( t ) = h ( t ) = TC AE 1 for t > T C h(t) [1] system operation function of the floodplan method
normalized contributing area
contributing area At (t) [km2]
(5.11)
From the system operation function the unit hydrograph of the floodplan method can be determined (see Chapter 4.4). However, it is only representative if the concentration time Tc is a integer multiple of the time interval t. 1 for 0 t TC d g ( 0, t ) = h ( t ) = TC dt 0 for t > T C
g ( t , ti ) = 1 ( h ( ti ) h ( ti t ) ) t
with h ( +0 ) = 0
(5.12)
(5.13)
g E ( t , ti ) =
1 1 AE ( h ( ti ) h ( ti t ) ) 3.6 t 1 = AE g ( t , ti ) 3.6
weighting function of the floodplan,
(5.14)
g(0,t)
[1/h]
Hydrology III - 49
g(t,ti) [1/h] gE(t,t) [

m ] smm
3
t-weighting function of the floodplan, unit hydrograph of the floodplan.
For a constant time interval t the hydrograph of the direct runoff QD(ti) is computed using the effective rainfall hNe(t) in the equation of superposition. For the time interval t an integer fraction of the concentration time Tc is recommended.
k =1
(5.15)
If several time intervals ti of different duration are applied, for each precipitation interval the assigned hydrograph can be computed by adding all partial hydrographs at the discrete values. For the direct runoff QD that pertains to one precipitation interval of the duration t and the effective depth of precipitation hNe (intensity iNe = hNe/t):
QD ( t ) =
hNe 1 AE ( h ( t ) h ( t t ) ) t 3.6 1 = iNe AE ( h ( t ) h ( t t ) ) 3.6
(5.16)
The ratio (t/Tc) of the interval t to the concentration time TC governs the course of the weighting function and the partial outflow hydrographs. Figure 5.5 displays the construction of the t-weighting function for different time intervals. In both cases the depth of precipitation equals one. For a precipitation interval t shorter than the concentration time Tc (Figure 5.5d) only a fraction of the drainage basin contributes to outflow.
Hydrology III - 50
dt pulse input and normalized contributing area
dt pulse input
time t [h] dt pulse function = dt weighting function and normalized contributing area
time t [h]
time t [h]
time t [h]
Figure 5.5:
Effect of the time interval on the course of the t-weighting function (the pulse volume equals in both cases one a) t > TC b) t < TC
Figure 5.6 contains the governing characteristic values for the variations of t/ TC.
normalized area
normalized area
Hydrology III - 51
Figure 5.6:
Characteristic values of the t-weighting function or the partial outflow hydrographs for different ratios t/TC
t > TC
T1 = T3 TC Tges X1 for h(t) - h(t-t) X1 for g(t,t) TC
t = TC
TC 0 2TC = 2t 1 1/t
t < TC t
TC - t TC + t
t - TC t + TC
1 1/t
t/TC
1/TC
5.2.4
Time-area diagram
The term time-area diagram denominates a runoff concentration model that is founded on linear translation, but, in contrast to the floodplan method, also considers the topography of the drainage basin. In an arbitrary shaped topographic drainage basin the overland flow velocity is not constant, but depends on the actual inclination. The travel time from any point within the basin to the outlet is lower in steep parts (high flow velocity) than in less inclined areas (low flow velocity). Equal travel time Tt results in different distances from the outlet. Those distances are displayed as isochrones in a map. Isochrones are lines of equal travel time to an outlet (see Figure 5.7). Applying the area-related Kirpich formula adapts the travel time to the inclination of the drainage basin, therefore a topographic map is needed. In steep areas the isochrones are further apart than in low-angled areas (steep areas = contour lines are close = high flow velocity = short travel time = isochrones far apart). The time-lag between two neighboring isochrones is equated with the time interval of the precipitation- and outflow hydrographs.
Hydrology III - 52
Figure 5.7:
Isochrone plan (lines of equal travel time), areas enclosed by isochrones At,i (5.17) lag time of neighboring isochrones time interval of precipitation- and outflow hydrographs discrete hydrograph time interval discrete isochrone time interval
Tt = t , ti = Tt ,i
Tt t ti Tt,i
[h] [h] [h] [h]
For precipitation in the form of a jump function the areas At,i enclosed by the isochrones contribute successively to outflow (see Figure 5.8a). In-between the isochrones, similar to the floodplan method, a continuously rising participating area is assumed. The function At(t) represents the temporal development of the areal runoff contribution. Its course is, in contrast to the floodplan method, not linear but considers the actual shape of the drainage basin. At ( ti ) = At ,i
k =1 i
for ti TC
(5.18)
At ( t ) = AE
for ti > TC area between two neighboring isochrones, time-area function, time of concentration.
at,i At(t) Tc
[km2] [km2] [h]
Hydrology III - 53
normalizedarea contributing to outflow
system answer to jump function
area contributing to outflow
direct runoff
system answer to dt weighting function
dt unit hydrograph
time t [h] system answer to needle pulse time - area diagram W(t)
weighting function g(0,t) time - area diagram
time t [h]
Figure 5.8: a) System operation of the time-area diagram in case of jump function precipitation, temporal development of the contributing area At(t), system operation function
Hydrology III - 54
The function At(t)/AE corresponds to the system operation h(t) of the time-area diagram. According to the correlation discussed in Chapter 4, the t-weighting function and the unit hydrograph can be computed. h (t ) = At ( t ) AE (5.19)
g E ( t , ti ) = g E ( t , ti )
AE 1 At ( ti ) = 3.6 3.6 t
(5.20)
g ( t , ti ) =
1 ( h ( ti ) h ( ti t ) ) t 1 At ( ti ) At ( ti 1 ) 1 At ( ti ) = = AE t AE t
(5.21)
h(t) [1] g(t,ti) [1/h] gE(t,ti) [m3/(smm)]
system operation of the time-area diagram t-weighting function of the time-area diagram unit hydrograph of the time-area diagram
The unit hydrograph gE(t,ti) is a discrete function with linear interpolation between the discrete values (see Figure 5.8b). The weighting function g(0,t) (instantaneous unit hydrograph) is here designated time-area diagram w(t). The differentiation of the system operation function is a histogram-shaped continuous curve (continuous terraced line, see Figure 5.8c). Applying h( + 0) = 0:
w ( t ) = g ( 0, t ) =
d 1 d h (t ) = At ( t ) dt AE dt
(5.22)
and from the isochrone plan

At (Tt ,i ) At (Tt ,i 1 ) A d At ( t ) = t ,i = dt Tt Tt
(5.23)
w(t) At(t) AE Tt
[1/h] [km2] [km2] [h]
time-area diagram (corresponding to the weighting function) area contributing to outflow size of the drainage basin time-lag between two neighboring isochrones (usually Ti = t)
Hydrology III - 55
In contrast to the t-weighting function (discrete values), the time-area diagram w(t) remains constant within the time intervals Tt,i-1 < t < Tt,i (histogram). The floodplan method can be regarded as a simplified version of the time-area diagram with parallel isochrones of equal distance. The term "simplified time-area diagram" is valid for all methods that approximate the drainage basin and the isochrone plan by basic geometric figures such as triangles, trapeziums or circle sectors.
5.3
5.3.1
Reservoir routing models

Linear reservoir
The linear reservoir is a fictitious reservoir, where outflow is proportional to the storage volume.
S = k QA
S QA k [m3] [m3/s] [s] storage volume outflow storage or retention constant
(5.24)
QZ
v=0 S QA
Figure 5.9:
Reservoir routing model, single reservoir and reservoir retention
The storage constant k usually obtains the same dimensions as the elapsing time t (for floods in unit hours).
S = 3600 k QA
k 3600 [h] [s/h] storage constant, conversion factor.
(5.25)
For stationary flow the constant k equals the detention period of a water particle in the reservoir and consequently the time through which the whole storage volume has been
Hydrology III - 56
exchanged once. The change in storage volume is computed with the methods of reservoir retention (see Hydrology II, Chapter 4.2).
dS d = k QA ( t ) = QZ ( t ) QA ( t ) dt dt d k QA ( t ) + QA ( t ) = QZ ( t ) dt
QZ [m3/s] inflow
(5.26)
The inhomogeneous differential equation of the linear reservoir can be rearranged to determine the outflow QA(t):
QA ( t ) = QA ( t0 ) e
( t t0 )
k
1 t + QZ ( t ) e k t0
( t t )
k
dt
(5.27)
applying the initial condition:
Q A ( t0 ) =
t QA(t0) S(t0) e
S ( t0 ) 3600 k
start, t > t0 outflow at time t0, initial condition storage volume at time t0 (initial storage volume S0) Eulers constant
[h] [m3/s] [m3] [2.7183]
If at the time t0 the reservoir is empty (S(t0) = 0) and therefore the outflow equals zero also (QA(t0) = 0), the solution of the differential equation is simplified and acquires the form of the integral of superposition (see Chapter 4.4).
1 QA ( t ) = e k 0
( t t )
k
QZ ( t ) dt
(5.28)
where t0 = 0, QA(t0) = 0 and QZ(t) = 0 for t < 0. The linear reservoir is a linear, time-invariant system that provides the weighting function g(0,t) (instantaneous unit hydrograph) as
g ( 0, t ) = 1 tk e k
(5.29) weighting function of the linear reservoir.
g(0,t)
[1/h]
Hydrology III - 57
Applying the system operation function h(t) provides the t-weighting function of the linear reservoir g(t,ti) (see Chapter 4.4).
h ( t ) = g ( 0, t ) dt = 1 e
0 t t k
(5.30)
g ( t , ti ) =
1 ( h ( ti ) h ( ti t ) ) t ( ti t ) 1 ti k = e k + e t
(5.31)
Determination of the outflow hydrograph by application of the discrete theory of superposition is only possible if the inflow hydrograph is available as a histogram (with constant values within each interval). If the reservoir inflow only exists as discrete values with linear interpolation, the reservoir retention may be resolved by the finite difference method.
S ( ti ) t
=k
Q A ( ti ) t
=k
QA ( ti ) QA ( ti 1 ) t
(5.32)
1 1 k QZ ( ti 1 ) + QZ ( ti ) ) = ( QA ( ti 1 ) + QA ( ti ) ) + ( QA ( ti ) QA ( ti 1 ) ) ( t 2 2
(5.33)
Hence the working equation of the linear reservoir in finite difference form is established (QZ and QA in discrete value form) as:
QA ( ti 1 ) QA ( ti ) = C1 ( QZ ( ti ) + QZ ( ti 1 ) ) + C2
with the coefficients C1 and C2:
(5.34)
C1 =
2 k+ t
, 2
= C2
k t k + t
2 2
(5.35)
and
t >> k =
S QA
In rearranged form and with C1 and C2:
QA ( ti ) = QA ( ti 1 ) + C1 ( QZ ( ti 1 ) QA ( ti 1 ) ) + C2 ( QZ ( ti ) QZ ( ti 1 ) )
C1 = t k + t C2 = 1 C1 k
, 2
(5.37)
If we assume the interpolation between the discrete values is linear when integrating the differential equation, an exact solution for C1 and C2 is obtained.
C1 = 1 e
t k
C2 = 1 C1 k
(5.38)
5.3.2
Non-linear, exponential reservoir
A cylindrical reservoir that contains an outlet or overfall shows a storage-outflow relation that can be expressed as an exponential equation.
m S = k QA
(5.39)
m k
parameter of the storage-outflow relation storage constant, dependent on m
The affiliated inhomogeneous differential equation of the reservoir retention cannot be resolved analytically. Therefore an approximation is determined applying the finite difference method. Integration is only possible in the case of a drainage process (with known initial storage volume at time t0, no inflow after t0, QZ(t) = 0 for t t0 homogeneous differential equation, m 1)
m 1 t m 1 m 1 QA ( t ) = QA ( t0 ) m k
t0 QA(t0) [h] 3 [m /s]
for m 1
(5.40)
start, t t0, outflow at time t0, initial condition.
For m = 1 we obtain the single, linear reservoir (see Chapter. 5.3.1).
5.3.3
Linear reservoir cascade
The reservoir cascade is a series connection of reservoirs. The simplest form is the linear cascade, a series connection of n equal linear reservoirs, each with the same storage constant k (see Figure 5.10). For the concentration of outflow, the linear reservoir cascade is also known as Nash-cascade (named after the English hydrologist Nash). The Sowjet scientists Kalinin
Hydrology III - 59
and Miljukov applied this concept to a streamflow model (in the general form the series connection of non-linear reservoirs, see Chapter 7.4.3). Outflow may be computed step by step by the use of the working equation of the finite difference method (linear single reservoir , see Chapter5.3.1).
1st linear reservoir time t [h] 2nd linear reservoir time t [h] 3d linear reservoir time t [h]
nth linear reservoir time t [h]
Figure 5.10: Linear reservoir cascade, development of the equilibrium equation in case of increasing number of reservoirs A more advanced method is based on the systemtheoretical approach. The series connection of linear, time-invariant partial systems leads to a single combined, linear, time-invariant system (see Chapter 5.3.1), the linear reservoir cascade features the same properties. Therefore a weighting function (instantaneous unit hydrograph) g(0,t) can be determined.
1 t g ( 0, t ) = k ( n 1) ! k
g(0,t) k n e [1/h] [h] [1] [2.7183]
n 1
(5.41)
weighting function of the linear reservoir cascade, storage constant, number of reservoirs, (n = 1 equals an single linear storage), Eulers constant.
Note that (0)! equals 1.
Hydrology III - 60
The weighting function g(0,t) of a linear reservoir acquires its peak at the time t = 0 and then falls exponentially (see Figure 5.10). With an increasing number of reservoirs the retention effect increases. The peak flow decreases and shifts further to the right. Since the procedure can be regarded as reservoir retention, the vertex or peak lies on the recession limb of the inflow hydrograph. Integration leads to a superposition system operation function and the tweighting function g(t,ti) (see Chapter 4.4).
k i ( ti t ) e g ( t , ti ) = n k t i =1 ( n 1)!
n
ti t k
n 1
k i ti n k t i =1 ( n 1) ! e
n
ti k
n 1
(5.42)
g(t,ti) [1/h]
t-weighting function of the linear reservoir cascade
The concepts of factorial, which initially was defined for integer, positive figures n, can be generalized for any real figure x by means of the gamma function.
( x 1)! = ( x )
(5.43)
Employing the gamma function in the weighting function of the linear reservoir cascade provides a model that allows non-integer numbers of reservoirs.
1 t g ( 0, t ) = k ( n) k
n g(0,t) [1] [1/h]
n 1
(5.44)
number of reservoirs, weighting function of the linear reservoir cascade.
In this form the weighting function of the linear reservoir cascade corresponds to the density function of two-parametric gamma-distribution (see Lecture "Hydrologic simulation methods"). Again, integration to determine the superposition system operation function and the t-weighting function is not possible. For this reason, either tables or computer routines of the "unfinished gamma-function" p(x) (1-parametric form) must be developed with substitution to a 2-parametric form. Substitution: x =
t
t k
n 1
h (t ) =
0
1 t k (n) k
t k
dt = p ( x )
(5.45)
where p(x) selected from tables:
Hydrology III - 61
p ( x) =
0
xn1 e x dx (n)
(5.46)
p(x)
integral of the unfinished gamma-function (probability, that x is not equaled or exceeded).
5.4
5.4.1
Parameter estimation for simple conceptual models

Moment method for linear model concepts
Moments are characteristic values of functions that are especially suitable for the description of time-dependent functions (see Lecture "Hydrologic simulation methods"). In the case of linear systems, certain parameters of the in- and output variables of the moments for the weighting function exist and may be used to estimate the model parameters. For the rth moment r,f of a time-dependent function f(t) related to zero (t0 = 0) (initial moment) the indefinite integral
r r , f = f ( t ) t dt 0
(5.47)
where
r,f f(t)
holds true.
rth moment of the function f(t) related to zero, any time-dependent function
The functions are limited to these for which: f(t) = 0 for t < 0 and f(t) 0 for t . The hydrographs of effective rainfall and direct runoff meet these requirements. The initial moment 0,f of the function f(t) represents the area below the function (e. g. the volume of effective rainfall or of the direct runoff).
0, f = f ( t ) t dt = f ( t ) dt
0 0 0
(5.48)
The center of mass ts,f of the function f(t) equals the ratio of the first and the initial moment.
ts , f =
f 1, 0, f
ts,f
[s,h,d]
temporal center of mass of the function f(t)
These moments are related to the center of mass ts = 1 /0 and are referred to as central moments r. The first central moment 1 is equal to zero by definition.
r, f =
0
1, f f (t ) t 0, f
dt
f =0 1,
(5.50)
r,f
central moment of the rth order of the function f(t) (related to the temporal center of mass)
The second central moment 2f derives from the initial moments.

2 2, f = 2, f 1, f
(5.51)
For linear system the following relations are valid. For r = 1: (relation of the initial moments) (5.52)
g 1, v 1, 1, = + u 0,v 0, g 0,u g 1, 1, = v + 1,u 0, g 0,v 0,u

ts , g = ts , v ts ,u
(5.53)
(5.54)
where the r initial moments of
th
'r,u 'r,v 'r,g
the input function u(t), the output function v(t) and the weighting function g(0,t).
The temporal centers of mass of ts,u ts,v ts,g [h] [h] [h] the input function u(t), the output function v(t) and the weighting function g(0,t).
The center of mass of the weighting function corresponds to the difference of the centers of mass for the in- and output functions.

th
Hydrology III - 63
The area below the weighting function i. e. the moment of the 0 order has the value one. Since the in- and output volumes are assumed equal, the results may be simplified further. Where: 0,g = 1 and 0,u = 0,v
g = 1,
1 v 1, u) ( 1, 0,v
(5.55)
For r = 2: (relation between the central moments)

2, g 2,v = 2,u + 0, v 0,u 0, g
(5.56)
Estimation of parameters according to the moment method assumes that the model is suited to the hydrologic system as soon as the empirical moments mr,g of the model correspond to the theoretical moments rg of the model. Moments are termed empirical if they derive from observed values.
r , g = mr , g or r , g = mr , g
(5.57)
r,g mr,g
theoretical moment of the model weighting function of rth order, empirical moment of the weighting function of rth order, computed from the moments of observed in- and output variables.
The number of required moment conditions corresponds to the number of model parameters. The theoretical moments of model concepts derive from the respective weighting functions. Therefore a relation between the parameters of the weighting function and the moments can be established. An essential disadvantage of the moment method is the sensitivity of the high order moments. In the equation that determines the moments, the term f(t)tr can be regarded as the weighting of the actual value of the function f(t) with the time-factor tr. As the time t elapses, the weighting of the function is constantly increasing, i. e. that the last part of the hydrograph is overemphasized in comparison to the beginning. This effect increases as the order of the moment increases. The recession limb of an outflow hydrograph, for example, has larger effect than the crest, whereas the aim of the model is more focused on the peak flow. Marginal variations of the observed values may cause substantial variations in the moments. This disadvantage of the moment method occurs from the second order on. However, this is managed quickly and simple, which suggests it is good as a rough parameter estimator. The method provides good results when only the first moments are required for the model.
Hydrology III - 64
The computation of empirical moments is dependent on the form of the discrete hydrograph, either in the form of a polygonal line or as a histogram. The subsequent approximations are usually of sufficient accuracy. Hydrograph in the form of polygonal lines with discrete values, the first and the last value equals zero (e. g. direct runoff hydrograph Qd(ti)).
f ( t0 ) = f ( tn ) = 0 , f = f ( ti ) tir t mr
i =1 n 1
or t
r
(5.58)
mr , f =
i =1
n 1
m f ( ti ) ti 1, f f m0,
Histogram (e. g. effective rainfall) , f = f ( ti ) ti t mr

i =1 n n
) t
r
or t
r
mr , f =
i =1
m f ( ti ) ti t 1, f 2 m 0, f
(5.59
mr,f f(ti) t n
[s,h,d] [1]
rth empirical moment of the discrete function f(ti), discrete, time-dependent function, time interval, number of intervals of the discrete function.
For the more important model concepts the relation between moments and parameters have been determined.
Hydrology III - 65
linear translation (see Chapter 5.2.1)

g = Tt 1,
(5.60)
linear single reservoir (see Chapter 5.3.1)

g =k 1,
(5.61)
linear reservoir cascade (see Chapter 5.3.3)
g = nk 1,
2 2, g = n k
1st central moment 2nd central moment

(5.62)
5.4.2
Storage-outflow relation of single reservoir models
If for an single reservoir the in- and outflow hydrographs QZ(t) and QA(t) are known, the storage hydrograph S(t) can be computed. S ( t ) = QZ ( t ) dt QA ( t ) dt + S ( t0 )
t0 t0 t t
(5.63)
= VZ ( t ) VA ( t ) + S ( t0 ) S QZ, QA t0 V Z, V A [m3] [m3/s] [h] [m3] storage volume, in- and outflow, start time, in- and outflow volumes.
If the outflow hydrograph is available in discrete form, the outflow volume V at the points of time ti may be determined. histogram (e. g. effective rainfall),
V ( ti ) = Q ( t k ) t
k =1
(5.64)
discrete values, trapezoidal approach (e. g. outflow hydrograph),
V ( ti ) =
k =1
Q ( tk 1 ) + Q ( tk ) 2
(5.65)
For a reservoir with an initial condition of S(t0) = 0 the discrete storage hydrograph S(ti) is:
S ( ti ) = VZ ( ti ) VA ( ti )
At any point of time ti the storage S t(t) can be plotted versus the reservoir outflow QA(ti) (see Figure 5.11).
storage content S [m3]
outflow QA [m3/s]
Figure 5.11: Determination of the storage constant k of a linear reservoir from the storageoutflow relation at the points of time ti If the mean variation is only marginal, the storage-outflow relation can be represented by an equalizing curve. For a linear reservoir, the relation is of a straight-line form and has the slope k.
S ( t i ) = k QA ( t i )
(5.67)
The fitting can either be accomplished graphically or by means of regression. This method to determine the parameters is best be applied for exponential reservoirs (see Chapter 5.3.2). In this case, there is a straight-line relation between the logarithmic values of in- and outflow (non-linear regression).

m S = k QA
Hydrology III - 67
or
ln ( S ( ti ) ) = ln k + ln ( QA ( ti ) ) m
ln S = ln k + ln QA m
(5.68)
5.4.3
Outflow recession curve of the linear, single reservoir
If the inflow ends at a given time t0, a drainage process results. With this type of reservoir the outflow features a specific hydrograph (recession curve). For the linear reservoir, the recession curve (see Chapter 5.3.1) is of an exponential form.
QA ( t ) = QA ( t0 ) e
t0 t k [h,d] [h,d] [h,d]
( t t0 )
(5.69)
time when inflow ends, time variable, in this case only time intervals, no inflow, t > t0 storage constant of the linear reservoir

areal precipitation iN (t) precipitation intensity iN [mm/h]
Hydrology III - 68
outflow Q (t) outflow Q [m3/s] recession curve 1
time t [h]
recession curve 2
outflow in logarithmic scale
time t [h] logarithmic outflow ln Q time t [h]
Figure 5.12: Determination of the storage constant of a linear reservoir from the logarithmic outflow curve. Here: Example of a rainfall-runoff system, hydrographs of the precipitation intensity, the outflow and the logarithmic outflow The logarithmic form provides a straight-line relation.
( t t 0 ) ( t t0 ) k ln QA ( t ) = ln QA ( t0 ) e = ln QA ( t0 ) k
outflow Q [m3/s]
(5.70)
Consequently the outflow recession curve of a linear reservoir plotted on logarithmic paper is a straight line (see Figure 5.12). This characteristic may be used to determine the storage
Hydrology III - 69
constant. The time intervals where no inflow occurs must be known. The storage constant k derives from the ratio
k=
t t0 ln QA ( t0 ) ln QA ( t )
(5.71)
(see Figure 5.12) For a rainfall-runoff system the precipitation intensity, the outflow hydrograph and the logarithmic outflow hydrograph are plotted (see Figure 5.12). Only if the logarithmic recession curve is of a straight line, relations can be established for parts of the hydrograph. Since the storage constant is determined by the recession curve, this method to estimate parameters is especially suitable for low-flow models that represent the development of flow during periods of drought.
Hydrology III - 70
Combination of model concepts in drainage basin models

One-component models for direct runoff
Clark model
6.1
6.1.1
Clark put the time-area diagram and the single linear reservoir in series. The outflow hydrograph of the time-area diagram (see Chapter 5.2.4) is the inflow hydrograph for the single linear reservoir (see Chapter 5.3.1). The model considers the translation - and the retention effect of a drainage basin (see Figure 6.1).
effective precipitation iNe * AE

Figure 6.1:
time time-area - area diagram diagram
linear reservoir
direct runoff QD
Clark model, series connection of the time-area diagram and linear reservoir
The weighting function of the Clark model g(0,t) derives from the superposition of the timearea diagram and the weighting function of the single linear reservoir. Integration is possible because w(t) remains constant through the time intervals (see Figure 5.8c).
g ( 0, t ) =
t 1 t k e w ( t ) e k dt k 0 t
(6.1)
w(t) k g(0,t)
[l/h] [h] [1/h]
time-area diagram storage constant weighting function of the Clark model
The Clark model contains two parameters, the travel time Tt (expressed in the isochrone plan) (see Figure 5.7) and the storage constant k. In addition to single linear reservoirs, non-linear reservoirs or reservoir cascades may be used.
Hydrology III - 71
6.1.2
Two-reservoir-model (Singhs model)
This model consists of two single linear reservoirs in series. The two storage constants k1 and k 2 characterize the detention storage and the retention effect within the receiving streams (see Figure 6.2).
effective precipitation iNe * AE

Figure 6.2:
linear reservoir k1
linear reservoir k2
direct runoff QD
Two-reservoir-model, series connection of two different linear reservoirs k1 and k2
Analogous to the linear reservoir cascade (see Chapter 5.3.3) a weighting function g(0,t) can be determined. 1 g ( 0, t ) = k1 k2
t kt k2 1 e e
(6.2)
6.1.3
Influence of precipitation on the model concept
Usually, a model concept that comprises of only one linear flow component provides too rough of an approximation of the hydrologic system. The outflow that pertains to a short and intensive precipitation event is more distinct than in the case of steady rain. The one-component model, with stated parameters, shows the same system operation for each precipitation event. To consider the changing outflow characteristics, the parameters are regarded as variables of precipitation and antecedent soil moisture. So coefficients and travel times may be related to precipitation intensity and -duration.
6.2
Multi-component models, parallel reservoir cascades
As outlined in Chapter 2.1 the runoff of a drainage basin is made up of several components: groundwater flow, interflow and surface runoff (see Figure 2.1). Surface runoff may again be subdivided according to its origins (paved or unpaved areas). A widely used model concept is the parallel connection of linear reservoir cascades. Each cascade represents one of the runoff components.
Hydrology III - 72
The model is linear if precipitation is distributed among the cascades at a constant ratio. Diskin introduced the model of two linear reservoir cascades and a constant distribution factor.
surface runoff
effective precipitation iNe AE
linear distribution alpha = constant
n1 reservoir intermediate flow
direct runoff QD
n2 reservoir
Figure 6.3:
Diskin model, parallel connection of two linear reservoir cascades and constant distribution factor
The model encloses five parameters n1, k1, n2, k2 that pertain to the linear reservoir cascades and the constant distribution factor . For 1 = and 2 = (1 - ) the weighting function is:
g ( 0, t ) =
k1,k2 n1,n2 1,2 [h] [1] [1] [1]
1 1 2 t t e k1 + k1 ( n1 ) k k2 ( n2 ) k
n 1
n2 1
t k2
(6.3)
storage constant of the reservoir cascades number of the reservoirs in each cascade constant distribution factor load factor of each reservoir cascade
Generally, the load of each component depends on the intensity and duration of the precipitation event and the antecedent soil moisture content. As a result, the model of runoff formation must precede the model of runoff concentration. Its task is to perform the separation into components in relation to precipitation. It is not necessary to separate direct runoff and base flow since the concept takes all runoff components into consideration. The input variable is therefore simply areal precipitation iN(t), the output variable the streamflow Q(t).
Hydrology III - 73
areal precipitation iN AE
distribution dependent on input
overall outflow Q
Figure 6.4:
Multi-component model, parallel connection of linear reservoir cascades and separation of precipitation according to precipitation
An example of a multi-component model is displayed in Figure 6.5. It is based on a exponentially falling distribution ratio that derives from the loss rate approach (see Figure 3.4c). Consequently the model is non- linear (see Figure 6.4) and weighting functions may only be determined for the single linear reservoir cascades of the individual flow components. In addition to the parameters of the single cascades ni and ki those of the distribution ratio must be determined. Calibration is only possible by means of non-linear optimization methods.
precipitation intensity i [mm/h]
iN (t)
surface runoff intermediate flow groundwater outflow
losses iV (t)
time t [h]
Figure 6.5:
Model of outflow formation, exponentially falling distribution ratio, threecomponent model
In general, the drainage basin model that contains linear reservoir cascades in parallel connection and non-linear component formation provides a very close approximation of the natural hydrologic system.
Hydrology III - 74
7
7.1
Flood routing models

Flood routing
As a flood wave passes through a river reach, the peak of the outflow hydrograph is attenuated since the wave front proceeds faster than the recession limb due to the larger inclination of the water surface. The wave is attenuated and extended (see Figures 7.1 and 7.2). The riverbed constitutes a retention space that lags the flow of water by a certain time (river retention).
flood recession (low angle)
flood rise (high angle) drop of wave crest
water depth W [m]
r iv e r
be d
long
short
flow distance [km]
Figure 7.1:
Flood wave in a river, principal sketch
Hydrology III - 75
river kilometer [km]
time t [d] Isoplethe and line of temporal succession for Q = constant Line of temporal succession for successive max Q Line of temporal succession for successive max h
Figure 7.2:
Lines of equal flow, flood wave of the river Elbe in September 1890
For stationary flow, a straight water stage-outflow relation exists (discharge curve). In a uniform channel the water stage drop ISp corresponds to the riverbed slope ISo. In the case of a flood wave (instationary flow), however, the rising limb exceeds, and the recession limb slope is smaller than the slope of the riverbed (see Figure 7.3). For a given water level h different discharge magnitudes for rising and recession limb occur.
flood recession
flood rise
Figure 7.3:
Surface slope, flow velocities and the rising and recession limb of a flood wave
Hydrology III - 76
stationary flow, uniform river cross-section

I Sp = I So = I E = I stat
(7.1) surface slope riverbed slope slope of hydraulic gradient slope of hydraulic gradient for stationary flow
ISp ISo IE Istat
[1] [1] [1] [1]
instationary flow, rising limb (see Figure 7.3)

I Sp > I So
(7.2)
If the surface slope exceeds the bed slope, the flow accelerates.
vinst ( h ) > vstat ( h ) Qinst ( h ) > Qstat ( h )

v Q h stat inst [m/s] [m3/s] [m] stationary instationary
and
(7.3)
flow velocity outflow water depth
instationary flow, recession limb (see Figure 7.3)

I Sp < I So
(7.4)
At the recession limb the surface slope is lower than the riverbed slope and flow is attenuated.
vinst ( h ) < vstat ( h ) Qinst ( h ) < Qstat ( h )
and
(7.5)
The water stage-outflow relation for instationary flow is no longer definite. Instead of a uniform water stage-outflow curve (stationary flow) a loop is obtained (instationary flow, see Figure 7.4). The shape of the loop is dependent on the gradient of the individual flood wave. The change in outflow dQ/dt also varies from flood to flood. Especially for rising water stage, the difference to stationary flow is distinct. According to Figures 7.2 and 7.4, peak discharge occurs earlier than the maximum water stage.
Hydrology III - 77
outflow Q [m3/s] water depth W [cm]
water depth W [cm]
flood recession
flood rise stationary outflow curve time t [h] outflow Q [m3/s]
Figure 7.4:
Water stage-outflow relation in the course of a flood wave
In contrast to reservoir retention, in river retention the peak of the outflow hydrograph is not necessarily located on the recession limb of the inflow hydrograph. Reservoir retention is therefore only a special case of river retention.
7.2
Simple flood forecasting methods
Flood routing in river reaches intends to answer the question: to which extend do flood waves from drainage basins get attenuated by the storage effect of the reach. Therefore, methods must be determined, that allow flood forecasting in a longitudinal cross-section of the river, based on observed or computed discharge values at inflow gage sites. For flood forecasting, the time and the magnitude of the peak flood flow are of special interest. Observations at one or several gage sites upstream provide information and allow conclusions about the critical cross-section. All the methods described subsequently are easy to handle, but suffer the disadvantage of relatively inexact results.
7.2.1
Gage relation curve
This method determines the relation between the peak flow water stages at different river cross-sections. To apply the method, the following assumptions must be made: Only cross-sections at the same river may be compared, and they may not be located too far apart (causal connection). Only comparable water stages may be considered. Since there is no definite water stagedischarge connection for instationary flow, the method is limited to the peak discharge water stages (dW/dt = 0).
Hydrology III - 78
The analysis of numerous flood discharges eventually provides the gage relation curve (see Figure 7.5). For a given flood event, based on the observations made at gage site A predictions may be made for gage site B downstream. Since the method is very simple, it is especially suitable for approximate real time forecasting.
water depth at gage A WA [cm]
time t
water depth at gage B WB [cm]
time t
Figure 7.5:
Analysis of flood waves to determine a gage relation curve
HW at C MW at C NW at C
WAmin
WAmax
Figure 7.6:
Gage relation curve including the water stage at a tributary (gage site C)
The application of this method is limited to smaller distances. If no definite connection between the water stages at gage A and B can be determined, lateral inflow may be the cause. The water stages of a single tributary can be considered by a number of curves (see Figure 7.6). When plotting the chart, one must consider if several rivers contribute to the discharge of the observed river (see Figure 7.7).
Hydrology III - 79
cross arrows on this line
Figure 7.7:
Gage relation nomogram, prediction of peak flow water stage at gage B based on observations on several gage sites A, C, D
7.2.2
Travel time curve
The travel time curve provides the time of the peak discharge of an expected flood event. To determine the curve, representative flood discharges must be analyzed. To identify the travel time between two river cross-sections, the peak flow must be clearly visible. From the analysis of several flood events the average wave travel times are determined and plotted against the water stage of the observed gage site. (see Figures 7.8 and 7.9). The cross-sectional shape of the river strongly influences the shape of the travel time curve (see Figure 7.8). In principle, the travel time will decrease with increasing water stages due to greater flow velocities. If the cross-section is enlarged the storage effect increases, this again leads to smaller wave travel velocities. (see Figure 7.8, area ).
The travel velocity of flood peaks is higher than the flow velocity since they proceed with wave velocity.
water depth WA [cm]
cross-section between A and B
travel time Tl [h]
Figure 7.8:
Travel time curve, influence of the river cross-section on the travel time of the flood peak

gage reference curve WB (WA) travel time curve Tl (WA)
Hydrology III - 80
water depth WA [cm]
water depth WB [cm]
travel time Tl [h]
Figure 7.9:
Gage relation curve and travel time curve to predict magnitude and time of the peak water stage at a critical gage
7.2.3
Prediction of discharge changes
The conversion of water stages to discharge using the stationary discharge curve is not correct in the case of instationary flow. However, the error decreases if the discharge change Q is considered rather than the absolute discharge Q. For short-time forecasting and only small discharge changes the error remains within an acceptable range. The initial conditions for this method are similar to those of linear translation: The travel time Tt of the discharge change Q is considered constant between two crosssections. A discharge change Q at an upstream gage occurs in the same magnitude at a downstream gage.
The discharge changes at the critical gage result from the time lag of the discharge change at an observed gage, not from inflow along the reach (see Figure 7.10).
Qvor ( t1 + t ) = Qi ( t1 + t Tt ,i ) Qi ( t1 Tt ,i )
(7.6)
Qvor Qi t1 t t1+t
[m3/s] [m3/s] [h] [h] [h]
discharge change at the critical gage discharge at the observed gage i time of observation forecasting time period forecasting time
Hydrology III - 81
Tt,i
[h]
outflow at gage A QA [m3/s]
travel time from observed gage i to the critical gage
here
travel time Tt,A
outflow at gage C QC [m3/s]
travel time Tt,C
outflow at gage B QB [m3/s]
critical gage
observation time
prediction time
Figure 7.10: Forecasting based on discharge change (see ground plan Figure 7.5), critical gage B, observed gages A and C, lag of the discharge change QA and QC by the travel time Tt,A and Tt,C (here Tt,A = t and Tt,C > t) The more observed gages are taken into consideration, the more accurate the results get. The overall discharge change of the critical gage is the sum of the single discharge changes.
Qvor ( t1 + t ) = Qvor ( t1 ) + Qi ( t1 + t )
i =1
(7.7)
Qvor
[m3/s]
discharge at the critical gage
Hydrology III - 82
Qi
k
[m3/s] [1]
discharge change at the critical gage as a result of the discharge change at the observed gage i number of observed gages
7.3
Hydraulic approaches to instationary flow
From the hydraulic point of view, flood wave deformation is an instationary flow problem. The governing equations derive from two underlying physical laws: mass conservation (continuity equation), energy conservation (energy equation).
The usual approaches consider 1-dimensional open channel flow: along the river bed only gravity and friction forces are effective, no oscillating waves that transport energy or momentum occur, only mass-transport waves, evenly distributed flow velocity in each cross-section (1-dimensional flow), changes in the water stage as a function of time can be neglected, no vertical acceleration component, crossfall negligible, no acceleration in cross-direction, roughness coefficient kSt for stationary flow may be applied.
1-dimensional, instationary flow can be represented by two independent parameters, the time t and the flow length x. Dependent parameters are the water depth h(x,t) (or the water stage W(x,t)) and the flow velocity v(x,t) (or the discharge Q(x,t)). To derive the governing equations, we consider a channel sector of infinitesimal length dx (see Figures 7.11 and 7.12). Analyzing this element, the temporal changes d/dt and the changes parallel to the direction of flow d/dx must be taken into consideration. The shape of the channel is assumed to be prismatic Continuity equation The temporal change of volume within the element corresponds to the difference of the inand outflow.
dV = QZ QA dt
(7.8)
If the change of discharge along a longitudinal cross-section of the channel is denoted Q/x, for the in- and outflow of an element of length dx (see Figure 7.11a), it is
QZ = Q
QA = Q +
Q dx x
Equation (7.9) in equation (7.8) provides:

dV Q Q = Q Q + dx = dx dt x x
(7.10)
Figure 7.11: Principle of continuity at a channel cross-section of the length dx principle sketch a) longitudinal section, discharge change along the channel element b) cross-section in the middle of the channel element and temporal change of cross-section. The volume element dV is determined with the element length dx multiplied by its crosssectional area dA in the middle of the element (see Figure 7.11b):
dV = dA dx
(7.11)
Due to the temporal change of the cross-sectional area A/t, for the temporal volumetric change in an element of length dx
dV A = dx dt t
After canceling down dx and rearranging the formula with A = A(x,t) and Q = Q(x,t) :
(7.12)
A Q + =0 t x
Applying the continuity principle, the equation may be rearranged even further:
Q ( x, t ) = v ( x, t ) A ( x, t ) A ( v A) + =0 t x
(7.14)
According to the chain rule of differentiation and with A = A(x,t) and v = v(x,t):
A A v +v + A =0 t x x
A Q V bSp h t v x [m2] [m3/s] [m3] [m] [m] [s] [m/s] [m] cross-sectional area flow into the channel element ds volume water level width water depth time flow velocity coordinates in flow direction
(7.15)
In a prismatic channel, the cross-sectional area A and the water level width bSp are directly related to the water depth h:
A = A(h) and bSp = bSp ( h )
(7.16)
The differential cross-sectional area dA may now be computed from the water level width and the water depth (see Figure 7.11b):
dA = bSp ( h ) dh
(7.17)
After further rearrangements finally the continuity equation of 1-dimensional instationary flow is obtained. The water depth h(x,t) and the flow velocity v(x,t) are directly related to the flow coordinate x and the time t, the cross-sectional area A(h) and the water level width bSp(h) only indirectly by h.
h A v h + +v =0 t bSp x x
Oftentimes it is necessary to apply the water stage W related to a given geodetic datum instead of to the water depth h.
W ( x, t ) = h ( x, t ) + z ( x ) or h ( x, t ) = W ( x, t ) z ( x )
(7.19)
ISo W Z
[1] [m] [m]
riverbed slope water stage related to any geodetic datum elevation of the river bed
By partial differentiation the continuity equation of the related variables v(x,t) and W(x,t) can be obtained:
h W d z W = = + I So and x x d x x W A v W v + v I So + + =0 x bSp x t
h W = t t
(7.20)
Figure 7.12: Principal sketch to determine the energy balance at a river stretch element, length dx Energy balance between two river cross-sections at the distance dx: For free surface flow the principle of energy conservation can be applied (see Figure 7.12)
hS + h + hk + ha + hv = const. hS h hk ha hv [m] [m] [m] [m] [m] geodetic elevation of the riverbed, water depth, kinetic energy head, acceleration head (deceleration head), head loss due to friction.
I So dx + h +
v2 2g
left
side 1
(7.22)
v2 h v2 1 v 2g h+ dx + + dx + dx + I v dx x 2g x g t
right
side 2
Rearranging the equation under inclusion of the dependent variables v(x,t) und h(x,t) provides
v v h +v +g = g ( I So I v ) t x x
h v t g ISo Iv [m] [m/s] [s] [m/s2] [-] [-] water depth, flow velocity, time, gravitational acceleration, riverbed slope, slope of hydraulic gradient.
(7.23)
If the water depth is replaced by the water stage W, the energy equation with the dependent variables v(x,t) and W(x,t) becomes:
v v W +v +g = g Iv t x x
(7.24)
The slope of the hydraulic gradient can be computed according to Mannings equation. V and h provide only an indirect dependence.
Iv =
v v
2 k St rhy3 4
(7.25)
kSt rhy
[m1/3/s2] [m]
roughness coefficient according to Strickler hydraulic radius
Hydrology III - 87
The system of partial differential equations from the energy- and continuity principle are known as St. Venants equations. The equations are almost linear, inhomogeneous and of the hyperbolic type. The independent variables are x and t, the dependent variables are v(x,t), h(x,t) or W(x,t). For lateral, perpendicular inflow q(t) that remains constant along the river stretch, the equations of St. Venant are transformed to:
h A v h q + + = x bSp x t bSp
(7.26)
v v h v + v + g + q = g ( I So I v ) t x x A
q [m3/(ms)] lateral, perpendicular flow (mainly inflow), q(x) = const.
(7.27)
outflow Q [m3/s]
time t [h]
flo
c re di
n tio
m [k
Figure 7.13: Flow of a flood wave, finite difference grid of the temporal and spatial development, principle sketch The St. Venant equations can only be solved step by step with finite differences. The relevant period of time t and the river stretch x must be subdivided into elements of t and x size (finite difference method according to Lax). The discrete values are computed at the gridpoints of the finite difference grid (see Figure 7.13). This method requires extensive numerical computation and is therefore carried out best with a computer. In addition, exact geometric and hydraulic channel data must be available.
Hydrology III - 88
7.4
7.4.1
Hydrologic flood routing concepts

Basic principles of hydrologic flood routing
Hydrologic concepts usually take advantage of a special case of instationary open channel flow, the reservoir retention. This reduces the conditions that must be satisfied to the equation of continuity. The energy equation is considered in an implicit way by the design of the routing models, the storage function and the outflow retention. Computing instationary flow with the storage equation of reservoir retention offers two major advantages: Similar to stationary flow, a definite relation between discharge and storage capacity exists. The computational steps are relatively easy.
In this chapter, two model concepts will be introduced that differ among other things in their typical field of application. Muskingum-model (mountainous areas and lowlands) Kalinin-Miljukov-method (low mountain ranges and river deltas)
The models are defined by characteristic parameters for each river stretch section. It is assumed that the specific boundary conditions for each event do no influence the parameters. The models are calibrated against observed in- and outflow hydrographs (e. g. Muskingum method) or the parameters are obtained from roughly guessed geometric and hydraulic channel properties (e. g.Kalinin-Miljukov method). For composed channel cross-sections different parameters for the piedmonts and the main flow channel may be defined (different retention effects).
7.4.2
Muskingum-model
The Muskingum model represents the river stretch between two gages by a storage space. This storage space consists of two parts, the prism and the wedge storage (see Figure 7.14).
Hydrology III - 89
wedge storage S2 = k*x*(QZ - QA)
ch an ne
l s ect ion
channel section prism storage S1 = k * QA
Figure 7.14: Muskingum model, subdivision of a river stretch into prism- and wedge storage Prism storage S1 For stationary flow (QA = QZ) a definite relation between the water volume stored in the river stretch and outflow exists. The storage space may be approximated by a prism. A linear storage-outflow relation is assumed (linear storage).
S1 = K1 QA
S1 K1 QA [m3] [s] [m3/s] prism storage volume prism storage retention constant outflow from a river stretch
(7.28)
Wedge storage S2 The second storage space represents the momentary difference between in- and outflow. The water level is assumed to be linear in shape. Consequently, a wedge-shaped storage space is formed (see Figure 7.14). The wedge storage overlays the prism storage and acquires either a positive (QZ > QA,, flood rise) or a negative value (QZ < QA, flood recession). Again a linear storage-outflow relation is assumed.
S 2 = K 2 ( QZ QA )
(7.29) wedge storage volume wedge storage retention constant inflow in river stretch
S2 K2 QZ
[m3] [s] 3 [m /s]
The overall storage volume S is the total of the two storage volumes.
S = S1 + S2 = K1 QA + K 2 ( QZ QA )
In the commonly used form of the Muskingum equation, the parameters K1 and K2 are replaced by k and x, since these parameters are more meaningful. K1 = k K2 = x k S = k QA + x k ( QZ QA ) S = k ( x QZ + (1 x ) QA ) k x [s] [1] retention constant of the Muskingum method weighting factor of the Muskingum method (7.31)
The weighting factor x ranges from 0 to 1.

0 x 1
(7.32)
For x = 0, the Muskingum-model corresponds to the linear, single reservoir. The case x = 1 must be excluded, since a storage volume S that only depends on the inflow QZ is unrealistic. For practical use, x usually acquires values between 0.2 and 0.4. The value is dependent on the ratio of instationary and stationary outflow depth in the inflow cross-section, or in other words, from the rate of inflow change. For triangular cross-sections, x 0.3 is proposed. For a given inflow hydrograph QZ(t), the solution of the reservoir retention differential equation is
t Q (t ) 1 x k (1 x ) QA ( t ) = A 0 e k (1 x ) + Q t e dt QZ ( t ) 2 Z ( ) 1 x 1 x k (1 x ) t0 t t t
(7.33)
The last term in the equation reduces for small values of t the outflow QA(t) to a magnitude lower than the initial value QA(t0). This means that the outflow hydrograph decreases at the beginning of a flood event and constitutes one of the disadvantages of this method. If a discrete inflow hydrograph QZ (ti )is given by discrete values, the finite difference approach is applied. If the in- and outflow hydrographs are known (point values and linear interpolation), the reservoir retention equation in finite difference form is obtained (trapezoidal approach). The time interval t must be substantially smaller than the retention constant k.
t << k
1 1 S ( ti ) S ( ti 1 ) = QZ ( ti ) + QZ ( ti 1 ) QA ( ti ) QA ( ti 1 ) 2 t 1 k ( x QZ ( ti ) + (1 x ) QA ( ti ) ) k ( x QZ ( ti 1 ) + (1 x ) QA ( ti 1 ) ) = t
(7.35)
The outflow is computed step-by step with the working equation of the Muskingum-model.
QZ ( ti 1 ) + C3 QA ( ti 1 ) QA ( ti ) = C1 QZ ( ti ) + C2
(7.36)
C1 = = C2 = C3
k x 0.5 t k (1 x ) + 0.5 t with :
k x + 0.5 t k (1 x ) + 0.5 t k k x 0.5 t k (1 x ) + 0.5 t
C = 1
i i =1
(7.37)
Rearrangement of the working equation allows a direct comparison with the working equation of linear storage (see Chapter 5.3.1).
QA ( ti ) = QA ( ti 1 ) + C1 ( QZ ( ti 1 ) QA ( ti 1 ) ) + C2 ( QZ ( ti ) QZ ( ti 1 ) ) C1 = C2 = t k (1 x ) + 0.5 t 0,5 t k x k = 1 C1 k (1 x ) + 0.5 t t

(7.38)
The denominator is the same for all coefficients C1, C2, C3 and C1, C2 The parameters can be calibrated by a simple, semigraphical method. However, for calibration in- and outflow data are required. The Muskingum formula is rearranged and the weighted discharge QA is introduced. S = k ( x QZ + (1 x ) QA ) = k Q A Q A = x QZ + (1 x ) Q A Q'A [m3/s] weighted discharge with : (7.39)
For x = 0, an extreme value, the single linear reservoir is obtained. The storage volume S is directly proportional to the weighted discharge QA. Since the weighting factor x is unknown
Hydrology III - 92
at the beginning, several different values x (usually in a range from x = 0 to 0.4) must be applied.
storage content S [m3] weighted discharge QA [m3/s] weighted discharge QA [m3/s] k = tan alpha for x = 0.2 storage content S [m3] weighted discharge QA [m3/s] weighted discharge QA [m3/s]
Figure 7.15: Relation between storage volume S and the weighted discharge QA for different weighting factors x For given in- and outflow hydrographs QZ(ti) and Q A( t i ) and an estimated weighting factor x, for any point of time t the storage content S( t i ) and the weighted discharge Q A( t i ) can be determined. The storage content S at a certain time ti (trapezoidal approach for hydrographs with discrete values) can be computed as:
S ( ti ) = S ( ti 1 ) + t
QZ ( ti ) + QZ ( ti 1 ) 2
QA ( ti ) + QA ( ti 1 ) 2
(7.40)
Weighted discharge QA at a point of time ti for an estimated value x:
Q A ( ti ) = x QZ ( ti ) + (1 x ) Q A ( ti )
The corresponding values S(ti) and QA(ti) are plotted in a diagram (see Figure 7.15). Connecting the points in their temporal succession, the resulting curve will generally take the form of a loop. The relation between the weighted discharge QA and the storage volume S must be as close and as linear as possible. Considering this, the best value of x from the investigated range can be chosen (in Figure 7.15, it is x = 0.2). The storage constant k corresponds to the gradient of the straight line that fits best to the QA - S loop. To assess the dependence in an analytical way, the magnitude of the correlation coefficient can be determined.
7.4.3
Kalinin-Miljukov method, basic principles
The Kalinin-Miljukov-method (KM-method) is based on the principle of the reservoir cascade. The river stretch is subdivided into characteristic sections that are considered as individual reservoirs (see Figure 7.16). The problem of instationary open channel flow is therefore reduced to the successive computation of reservoir retention. A special form of this model is a cascade of identical, linear reservoirs. (Nash-cascade, see Chapter 7.4.4 and 5.3.3).
river stretch
inflow
outflow
outflow Qz
time t reservoir cascade model
outflow Qz time t
Figure 7.16: Kalinin-Miljukov method, instationary open channel flow described by a reservoir cascade model
Hydrology III - 94
Kalinin and Miljukov established the theoretical basis and developed a method to derive the necessary parameters from coarse geometric and hydraulic channel data without in- or outflow measurements. The KM-method requires a certain homogeneity of the river stretch or at least of distinct parts of it. In addition, the prerequisites of the hydrodynamic equations in Chapter 7.3 must be fulfilled. hydraulically similar river cross-sections (similar shape and roughness throughout the whole section), uniform bed slope (no drops or water breaks).
The method derives from Mannings equation of discharge: Q = A kSt rhy3 I E 2

2 1
(7.42)
In the case of uniform, stationary flow the energy gradient IE corresponds to the river bed slope I So . The stationary slope is referred to as I stat . I stat = I E = I So (7.43)
Between the stationary discharge Qstat and the water depth h a direct relation exists, the socalled water stage -outflow relation Qstat(h).
2 Qstat ( h ) = A ( h ) k St rhy ( h ) 3 I stat 2 1
(7.44)
Qstat h kSt A rhy IE Istat Iso
[m3/s] [m] [m1/3/s] [m2] [m] [-] [-] [-]
stationary open channel outflow, water depth, Stricklers roughness coefficient, cross-sectional area, hydraulic radius, energy line gradient, energy line gradient for stationary flow, river bed slope.
Mannings equation provides only a rough estimation of the outflow hydrograph or the water depth -outflow relation. Preferably the outflow should be measured directly. For instationary open channel flow it can be assumed that for identical water depths the flow parameters remain the same. Therefore the only difference between stationary outflow Qstat and instationary outflow Q inst is the difference in the energy lines. In contrast to the uniform gradient of the stationary energy line, the gradient of the instationary energy line may change in the course of a flood.

2 Qinst ( h ) = A ( h ) k St rhy ( h ) 3 Iinst 2 1
Q Iinst
[m /s] [1]
instationary open channel flow, energy line gradient for instationary open channel flow.
Under the premise that even in the case of instationary flow the energy line is parallel to the water surface line, the gradient of the instationary energy line Iinst consists of a stationary part Istat and an additional part I (see Figure 7.17). I inst = I Sp I inst = I stat + I ISp [1] [1] even for instationary flow slope of water surface, gradient difference between stationary and instationary energy line gradient. (7.46)
Figure 7.17: Longitudinal river cross-section, comparison of stationary and instationary energy line gradient (Istat and Iinst) for flood rise Inserting this in the discharge equation provides: Qinst ( h ) = A ( h ) k St rhy ( h ) 3 I stat + I
2
(7.47)
The root on the equation can be rearranged and transformed to a power function. It can be proved that already the first two elements of the function provide sufficient precision because the slope angles are usually very low (the error is usually lower than 6 %).
Hydrology III - 96
I stat + I = I stat 1 + = I stat
I I stat
1 I 1 + + ... 2 I stat
(7.48)
The equation of instationary open channel flow consists now of two components. For a certain water depth h the instationary outflow Qinst consists of an stationary part Qstat and a certain differential value Q.
2 1 I Qinst ( h ) = A ( h ) k St rhy ( h ) 3 I stat 1 + 2 I stat
1 I Qinst ( h ) = Qstat ( h ) 1 + 2 I stat Qinst ( h ) = Qstat ( h ) + Q ( h )
(7.49)
For equal water depths h, Q depicts the difference between stationary and instationary outflow.
1 I Q ( h ) = Qinst ( h ) Qstat ( h ) = Qstat ( h ) 2 I stat
(7.50)
Q(h) [m3/s]
outflow difference between stationary and instationary outflow for equal water depth h
Considering now the relation between outflow Q and the assigned water depth h for instationary and stationary flow (see Figure 7.18), it can be stated that for a certain instationary outflow Q and the matching water depth h, the same outflow Q occurs in the stationary case with a water depth h + h that differs in height by h.
Qinst ( h ) = Qstat ( h + h )
(7.51)
[m]
difference between water depths for identical outflow in stationary and instationary case
Hydrology III - 97
water depth h [m]
flood recession
y onar e stati curv w o l f out
flood rise
outflow Q [m3/s]
Figure 7.18: Relation between water depth and outflow for stationary and instationary flow (Qstat(h) and Qinst(h)) Considering this, it becomes obvious that the water stage-outflow curve Qstat(h) may be applied to determine instationary outflow Qinst , if the water depth h is replaced by the sum of h + h (see Figure 7.18). The outflow difference Q can be determined from the water stage-outflow curve as the difference between the stationary outflows at the water depths h and h + h (see Figure 7.18).
Q ( h ) = Qinst ( h ) Qstat ( h ) = Qstat ( h + h ) Qstat ( h ) = Qstat ( h )
(7.52)
qstat
[m3/s]
outflow difference related to the stationary water stage - outflow curve
Another simplification the KM-method takes advantage of is the assumption of a linear water surface. If this assumption is also considered valid for instationary flow, then geometric relations between the water depths at the individual cross-sections can be determined (see Figure 7.19). Starting at a cross-section r with a water depth h, the same water depth can be found upstream in the cross-section m at the distance lb.
Hydrology III - 98
Figure 7.19: Longitudinal cross-section through a river stretch with the characteristic length lch, water surface and reference cross-sections r and m
h = I lb
lb [m]
or
I =
h lb
(7.53)
reference stretch in the channel
If lb is known, the instationary outflow in the cross-section can be correlated with a stationary water depth at the cross-section. In other words, the stationary water stage - outflow curve indicates even instationary outflow. Therefore water depth and outflow are related directly, with the only problem being that the relation occurs between two different cross-sections. The stationary water depth that is associated with an instationary outflow occurs simultaneously at a certain distance lb further upstream. The stationary water stage - outflow curve may for this reason be applied in the case of instationary flow also. Hence the difficulties with the instationary outflow loop can be ignored. This relation is only valid if the cross-sectional shape and the outflow curve do not change along the river stretch.
Qinst ( h ) = Qstat ( h + h )
(7.54)
cross-section r cross-section m The association of instationary outflow with stationary water depths is only possible if the specific distance lb is considered. For all subsequent derivations it is easier to replace the unknown distance lb by the doubled amount, which we refer to as characteristic length lch.
Hydrology III - 99
h 1 I 1 lb Qstat ( h ) = Qstat ( h ) = Qstat ( h ) 2 I stat 2 I stat
(7.55)
lch = 2 lb =
Qstat ( h ) I stat
h Qstat ( h )
(7.56)
where
Qstat ( h ) = Qstat ( h + h ) Qstat ( h )
(7.57)
lch
[m]
characteristic length
The term h/Q(h) corresponds to the gradient of the outflow curve for the given depth h in finite difference form (see Figure 7.18). Therefore it is only necessary to know the stationary outflow curve to determine the reference stretch lb or the characteristic length lch. In a river stretch with the characteristic length lch several simple geometric techniques can be used to determine the volume (see Figure 7.19). If the water surface is linear and similar along the stretch, the volume of the instationary outflow wedge ACF equals the volume of the prism BCFE. Hence the instationary outflow volume VADGF can be related directly to the stationary water depth in the middle of the reference stretch
VACF = VBCFE VADGF = VBDGE = A ( h + h ) lch = A ( hm ) lch
(7.58)
where hm = h + h A h [m2] [m] cross-sectional area, water depth in the middle of the reference river stretch. (7.59)
Subsequently, the results for the middle of a river stretch section with the characteristic length lch are listed: The one-to-one stationary outflow curve Qstat(hm) represents the instationary outflow QA (QA = Qinst, reservoir outflow). The water depth hm indicates the volume S (storage contents). Even for instationary flow a direct relation between storage contents and outflow S(QA ) exists.
Hydrology III - 100
The water depth hm in the middle of the selected river stretch constitutes the connection between the outflow QA and the storage content S. Applying QA = QA ( hm ) = Qstat ( hm ) and S = S ( hm ) = A ( hm ) lch provides S = S ( QA ) or
Q A = QA ( S ) .
QA S
[m3/s] [m3]
outflow from river stretch of the length lch, storage content of the river stretch, length lch.
A channel cross-section with the characteristic length lch may be regarded as a reservoir. Instationary flow can be calculated relatively easy with the help of the reservoir retention method. Long flow paths lFl can be represented by a series of reservoirs, each with the characteristic length lch. The only data required are the representative channel cross-section A(h) and the stationary outflow curve Qstat(h). Since the outflow curve is a one-to-one function, h = h(Qstat), lch can be regarded as a function of Qstat. Looking at Figure 7.20, it becomes obvious that the characteristic length lch is not a constant because the slope of the outflow curve is not linear.
lch = lch ( Qstat ) =
Qstat dh I stat dQstat
(7.60)
storage constant k (h)
characteristic length lch (m)
outflow Qstat (m3/s)
Figure 7.20: Dependency of the characteristic length lch and the storage constant on the stationary outflow Qstat, determination of the relevant flood range, mean values lch,m and km
Hydrology III - 101
In the range of the relevant flood flows, however, the characteristic length lch remains almost constant. Therefore a mean characteristic length lch,m, that is taken from the upper part of the outflow curve (from Qmin to Qmax) is applied (see Figure 7.20).
lch, m
lch,m Qmin Qmax
max 1 = lch ( Qstat ) dQstat Qmax Qmin Qmin
(7.61)
[m] 3 [m /s] 3 [m /s]
mean characteristic length lower and upper boundary of the relevant range of the outflow curve
water depth h (m)
outflow Qstat (m3/s)
Qmax
Figure 7.21: Stationary outflow curve Qstat(h), relevant flood range from Qmin to Qmax, subdivision in outflow intervals Q For practical use the finite difference equation of the characteristic length is more convenient. The relevant range of the outflow curve is subdivided into m intervals of uniform length Qi , the characteristic length lch,i is determined in each interval middle Qi (see Figure 7.21).
lch,i =
Qi hi I stst Qi
with
Qi = Qstat ,i
(7.62)
lch, m
1 m = lch ,i m i =1
(7.63)
Another problem is selecting the best number of characteristic river stretch sections. The total length of the examined river stretch lFl rarely coincides with an integer multiple of the
Hydrology III - 102
characteristic length. Lch. To solve this, the integer of the quotient is taken and the reservoirs are fitted to this value.
n=
lch, m lFl
lFl n
integer value
(7.64)
* lch =
(7.65)
* S ( h ) = A ( h ) lch
(7.66) integer number of reservoirs total length of the examined river stretch fitted characteristic length
n lFl l*ch
[1] [m] [m]
The cross-sectional area A(h) and the outflow curve Qstat(h) remain unchanged. The following steps must be taken: determination of a representative channel cross-section A(h), determination of the representative stationary outflow Qstat(h), determination of the relevant flood range from Qmin to Qmax, subdivision of the flood range into intervals of uniform length Qi, determination of the characteristic length lch,i in each interval, computation of the mean characteristic length lch,m, determination of the number of reservoirs n, fitting of the characteristic length lch to the total length of the river stretch lFl, determination of the storage content-outflow relation S(QA of the reservoirs).
The storage content-outflow relation S(QA ) has much more influence on the precision of the results than the characteristic length lch or the number of reservoirs n. Usually attempting to replace the whole river stretch by a number of equal reservoirs is best. If this is impossible due to inhomogeneities (e. g. if the bed slope or the cross-sectional area change suddenly or tributaries occur), the sections of equal conditions must be considered separately. This being possible is an advantage of this method; additionally it can be fitted easily to the natural conditions. As a result, even natural or artificial reservoirs such as lakes or dams can be taken into consideration. If continuous outflow measurements for the river stretch are available, usually the calibration based on the outflow hydrographs provides better results than the KM-method.
Hydrology III - 103
7.4.4
Kalinin-Miljukov method, linear reservoir cascade
Another widely used special case of the KM-method is a series of identical linear reservoirs (linear reservoir cascade, see Chapter 5.3.3). In this case the model is only dependent on two parameters, the number of reservoirs n and the storage constant k of the linear reservoirs. In a linear reservoir, the gradient of the storage content-outflow relation remains constant.
S = k QA dS = k = const. dQA
k [s,h] storage constant
(7.67)
(7.68)
In a channel section with the characteristic length lch ,the instationary outflow QA can be related to the stationary outflow curve Q stat(h) .
QA = Qstat ( h )
(7.69)
The differential storage content dS is the product of the section length lch, the water surface width bSP and the differential water depth dh.
dS = lch bSp ( h ) dh
(7.70) water surface width
bSp
[m]
Consequently, the storage constant k can be determined only by use of the stationary outflow curve Qstat(h) and the cross-sectional shape b Sp(h) .
k = lch bSp ( h )
dh dQstat
(7.71)
The storage constant k of the KM-method is, based on its physical meaning, often referred to as travel time parameter T. In the stationary case, k represents the travel time of a water particle through a reservoir and in the instationary case the mean travel time (distance from the center of gravity, see Chapter 5.3.1). The reservoir constant of a linear reservoir cascade is like the characteristic length not a constant, but dependent on the outflow Qstat. Since the outflow curve is a one-to-one relation, even the water surface width bSp can be regarded as a function of the stationary outflow. Where h = h ( Qstat ) and bSp = bSp ( Qstat )
Hydrology III - 104
k = k ( Qstat ) = lch ( Qstat ) bSp ( Qstat )
dh dQstat
(7.72)
The storage constant is represented by km as the mean value over the relevant flood range (Qmin to Qmax). To compute the linear reservoir cascade requires the preceding computation of the characteristic length (see Chapter 7.4.3). Representative cross-section A(h), bSp(h) and stationary outflow curve Q stat(h) , relevant flood range from Qmin to Qmax, outflow intervals Qi, characteristic length lch,i assigned to each outflow interval Qi, mean and fitted characteristic length lch,m and lch and the number of reservoirs n.
The number of reservoirs n is taken from the linear reservoir cascade. To derive the storage constant from the discrete outflow curve the finite difference form of the equation is applied.
ki = lch ,i bSp ( Qi )
h Qi
with
Qi = Qstat ,i
(7.73)
km =
ki km m
1 m ki m i =1
[s,h] [s,h] [-]
(7.74) storage constant according to the values of the discrete outflow curve Qstat,i mean storage constant number of outflow intervals Qstat,i
If only an integer number of reservoirs is permitted, the storage constant needs to be fitted to the given river stretch lFl . Note that in principle even fractions for the number of reservoirs are possible (see Chapter. 5.3.3).
k = km
* * lch
lch, m
(7.75) fitted storage constant
k*
[s,h]
The parameter estimation by means of the KM-method uses only coarse principles of channel hydraulics. If for a river stretch observed outflow curves are available, the model must be calibrated against the observed values (optimum parameter estimation or moment method, see Chapter 5.4.1). The computation of outflow hydrographs depends on the characteristics of the given inflow hydrographs:
Hydrology III - 105
inflow values averaged over each time interval: direct computation by means of the t-weighting function and the principle of superposition (see Chapter 5.3.3). discrete values and linear interpolation: Step-by-step computation (reservoir after reservoir) by means of the working equation of the linear reservoir (see Chapter 5.3.1).
Hydrology III - 106
8
8.1
Physically based hydrological models

Fundamental principles
System hydrological models offer an opportunity to calculate rainfall-runoff events for drainage basins. They may be used as a basis for design in engineering practice. The problem with these models is the insufficient spatial resolution, and the parameters being based on systems theory. Therefore, the application of models on other drainage basins is difficult and unreliable. In addition, problems arising from variations in land use or the results of possible climatic changes cannot be predicted or simulated, and the models only provide information on the selected output, e. g. runoff. System hydrological models cannot be applied to determine the mass transport in a drainage basin. Therefore, they are not suitable for various problems related to environmental protection, such as groundwater protection or contamination problems. Physically based problems with high areal resolution are currently mainly developed and applied for research purposes. Their primary goal is to provide better understanding of the interactions between various processes. Thereby it is desired to derive the single processes and their interactions from physical principles. Usually, physically based models are applied on small-scale problems only, whereas systems hydrology is more suitable for large-scale applications. Presumably, physical models will grow in importance in the future. The principal differences between physically based modeling and systems hydrology are (with regards to physical models): each process is considered, high spatial resolution, physically based parameters, application on other areas easier possible.
A physically based hydrological model must be able to compute all relevant processes based on physical laws. Nevertheless, the complexity of the processes makes it impossible to do the modeling based entirely on physical laws. Single processes will therefore be modeled in a conceptual way and the complete model will be a combination of these refined single processes. In the course of this, the interactions of the processes must be taken into consideration. The mathematical formulation that results is still only an approximation of reality. Figure 8.1 displays the hydrological processes and their interactions in a rural drainage basin.
Hydrology III - 107
Actual evapotranspiration
Precipitation
Interception
Precipitation from plants, land surface
Plant Transpiration
Infiltration
Depression storage
Contributing areas
Soil evaporation
Plant withdrawal
Soil moisture in root zone
Surface runoff
Soil moisture in deeper layers: groundwater replenishment, capillary rise
Intermediate flow
Base flow
Discharge of receiving stream
Discharge at basin outlet
Figure 8.1:
Hydrological processes and their interactions in a rural drainage basin
8.2
8.2.1
Spatial model extension

One-dimensional models
One-dimensional models (column models) describe the soil moisture regime without consideration of lateral processes. The processes interception, evaporation, infiltration, movement of water in the soil stratum, interaction of soil and groundwater and their respective interrelations are represented in the model. Surface runoff may be computed as infiltration excess. Figure 8.2 shows the processes considered in a one-dimensional model.
Hydrology III - 108
(1) Precipitation from atmosphere (2) Precipitation from plants (3) Dripping precipitation (4) Soil water movement due to root withdrawal (5) Soil evaporation (6) Root withdrawal (7) Plant transpiration (8) Evaporation from plant surface (9) Macropore infiltration (10) Matrixinfiltration (11) Water movement in soil matrix (12) Interaction macroporesystem - soil matrix (13) Depression storage
Figure 8.2:
Processes considered in one-dimensional models
Models of this type may be applied to calculate balance variables in a fast and efficient way, but are rarely used for flood modeling purposes. Only a runoff coefficient determined based on physical laws may be used as input for systems models. The modeling of the single processes differs only marginally from the approaches applied in other spatial resolutions and will be explained in detail in Chapters 8.3.1 - 8.3.6.
8.2.2
Two-dimensional models
Two-dimensional, physically based modeling describes the relevant hydrological processes in a cross-section (generally slope cross-sections) in the vertical and horizontal directions. The processes that must be taken into consideration are: interception, evaporation, infiltration, movement of water in the soil stratum, interaction of soil and groundwater, surface runoff and their interactions. Two-dimensional models are therefore suitable to determine the temporal development of surface runoff. Furthermore, the horizontal movement of water is computed. In addition to infiltration due to precipitation the infiltration caused by surface runoff can be taken into consideration. Figure 8.3 displays the processes considered in twodimensional models.
Hydrology III - 109
Figure 8.3:
Processes considered in two-dimensional models
8.2.3
Three-dimensional models
Three-dimensional models simulate hydrological processes as they occur in nature. The processes considered are the same as in two-dimensional modeling: interception, evaporation, infiltration, movement of water in the soil stratum, interaction of soil and groundwater, surface runoff and their respective interactions. These models are able to compute flowprocesses in the receiving stream. These models simulate surface runoff and the movement of water in the soil stratum and the temporal development of runoff in the entire drainage basin closely to reality.
Hydrology III - 110
(1) Precipitation from atmosphere (2) Precipitation from plants (3) Infiltration in micro- and macroporesystem (4) Surface runoff (5) Intermediate flow (parallel to inclination, close to the surface) (6) Water movement in soil matrix (7) Interaction macromicroporesystem (8) Capillary rise or deep percolation (9) Return flow (10) Root withdrawal (11) Soil evaporation (12) Plant transpiration
Figure 8.4:
Processes considered in three-dimensional models
8.3
8.3.1
Temporal and spatial model resolution

Temporal resolution
Discrete models Discrete models are designed to compute single events. They are applied to compute extreme events (e. g. floods) or conditions (e. g. malfunctions). They are applied only on selected precipitation events and the maximum duration is usually about 2 weeks. The time intervals of discrete models usually range between several seconds and a few hours. When applying discrete models, it is important to properly represent the dynamic behavior of the processes involved, however, possible systematic mistakes can easily be corrected and are only of minor influence. Continuous models Continuous models are applied to compute the water regime components over extended periods of time, e. g. drought periods. Short-time changes are only of secondary importance. Therefore, the aim is not to completely simulate each event that occurs within the examined period of time, but to correctly represent the overall long-term balance. Consequently, the
Hydrology III - 111
components of a continuous model may differ from those of a discrete model. The time intervals of continuous models range according to the actual problem from several hours to several days.
8.3.2
Spatial resolution
For processing reasons, the examined area is subdivided into elements. Most of the time the elements are of uniform size and shape. In the case of one-dimensional modeling the soil stratum is subdivided into the soil types that can be distinguished. The typical resolution ranges from 1 to 50 cm, with increasing element magnitudes from surface to bottom. In the case of 2-dimensional modeling, the examined area is subdivided into single columns, thereby the horizontal resolution ranges from 1 to 200 m, and vertically from 1 to 50 cm. Figure 8.5 displays an area partitioned into elements.
Figure 8.5:
Two-dimensional slope split up into elements
Computing the vertical processes in two-dimensional models usually requires a simplified, horizontally terraced approximation of the slope. However, for the determination of surface runoff the real inclination of the elements must be used. Some two-dimensional models also consider the width of the slope. The connection of the elements in this case is dependent on the width (runoff correction). Three-dimensional models usually divide the drainage basin up into rectangular elements. Each rectangle represents a soil column and is vertically subdivided into several layers. The typical horizontal extension of the elements ranges from 10 up to 200 m. The vertical subdivision is not uniform, with the upper layers usually measuring only several centimeters, with increasing thickness of up to 1-2 m at greater depths. Figure 8.6 displays a possible way to discretize a drainage basin. Two points must be considered when subdividing a drainage basin into three-dimensional elements:
Hydrology III - 112
High resolution takes into account all details of the drainage basin, but parameter estimation may be difficult, hard to compute and oftentimes unstable. Coarse resolution neglects the details of the drainage basin, and it is questionable if the physical laws valid for small-scale areas are still applicable due to the variability within the element.
Figure 8.6:
Three-dimensional hydrological model split up into elements
In a grid model, one-dimensional features such as rivers, streets, trenches can only be considered indirectly by so-called mixed elements. Discretizing in temporal regard (time-steps) is not only dependent on the time frame, but also on the spatial resolution. The higher the spatial resolution, the smaller the appropriate timestep (e. g. the mean soil moisture content of a soil layer of 5 cm thickness changes much faster than that of a one-meter layer). A number of partially physically based models exist, that divide the drainage basin up into non-uniform parts, such as sub-catchment areas. This type of model considers single processes predominantly by application of conceptual systems approaches. Hence they constitute a model class in-between systems models and physically based models in hydrology.
Hydrology III - 113
8.4
Modeling of single processes
Modeling based on physical laws consists of two major parts, the modeling of single processes and subsequently their combination. Outlined below are only some selected process models.
8.4.1 Infiltration
Figure 8.7 displays some hydrological processes relevant for infiltration.
Macropore
Figure 8.7:
Hydrological processes relevant for infiltration 1. Precipitation 5. Matrix infiltration 6. Movement of water from the macropores into the soil matrix 7. Movement of water within the soil matrix 2. Surface runoff 3.-4. Macropore infiltration
These processes must be determined in quantitative terms. Infiltration into the soil matrix and the movement of water within the soil matrix can be expressed by Richards equation. Nevertheless, for macropore infiltration new model approaches must be developed. Macropores only contribute to infiltration under certain conditions. Low-intensity precipitation is completely absorbed by the soil matrix. As soon as the precipitation intensity exceeds the permeability of the soil, the water flows on the surface towards the macropores. The water encounters the macropores and eventually flows down. If the soil matrix is
Hydrology III - 114
unsaturated, the water may be absorbed by the soil matrix due to the suction head. The relation between the macropore flow q and the macropore water content can be represented by two parameters: q = (8.1)
The parameters describe the geometry of the macropore system. The water movement within the macropores can be described with the principle of the kinematic wave. The equation of continuity is:
s=
q + t z
(8.2)
S represents the flux between the macropores and the soil matrix. The movement of water within the macropore system may therefore be estimated by the three parameters macropore density, depth and permeability.
8.4.2
Evaporation
Potential evaporation may be calculated by application of the Penman-Monteith formula. The actual evaporation derives from soil evaporation and the transpiration of plants.
Soil moisture
Figure 8.8: Transpiration reduction factor as a function of the water content
Hydrology III - 115
8.5
8.5.1
Model parameters
Parameter estimation
Physically based models require a large number of parameters that need to be estimated from observations or calibrations or can be taken from literature. Usually only a limited number of measurements are available. Spatial information is predominantly available as: digital area models (DGM) in 50 x 50 m grid resolution, which are precise down to a distance of a few meters, satellite-based land use information in 30 x 30 m grid resolution. With from 16 up to 50 land use classes, soil type maps, radar precipitation measurements.
Spatial information may partly be used directly or must be converted to model parameters. For example the leaf area index (LAI) to determine interception and transpiration may be estimated from land use and the season. The soil depth penetrated by roots can be estimated similarly. Estimations are usually based on literature data. The so-called pedotransferfunctions allow the estimation of physical soil characteristics based on easy-to-obtain variables such as the grain size distribution. These conversions are uncertain and the error they cause can be hard to estimate.
Point data: soil characteristics: saturated conductivity, PF curves, thickness and stratification, porosity, meteorological data: depth of precipitation, temporal development of precipitation (recording rain gages), temperature, radiation, wind direction and -speed, relative humidity.
Point data must be transformed to spatial information. Observations that consist only of one or two values (even if they are located partly beyond the borders of the reference area ) are easy to convert. Measurements that consist of several point data require the application of geostatisical methods. The initial and boundary conditions of discrete models strongly influence the obtained results. It is crucial to estimate the initial soil moisture content correctly.
Hydrology III - 116
8.5.2
Parameter variability
When estimating parameters, the required variables are usually represented by the mean value as the least improbable value. In addition to the uncertainties mentioned above, the parameter variability plays an important role, too. 8.5.2.1 Temporal variability Temporal variability plays an important role especially for precipitation modeling. If the daily depth of precipitation Is assumed as evenly distributed over 24 hours, the flood runoff would be significantly underestimated. Even the soil moisture distribution is dependent on the temporal variability of precipitation due to the infiltration of surface runoff. Other meteorological data such as temperature or radiation are also time-dependent, but this may to the major part be taken into consideration by a daily cycle. Soil parameters are oftentimes considered as time-independent. Unfortunately, this assumption is not correct. Since the soil stratum is influenced by fauna and flora, its physical parameters change with time. Macropore systems are formed and changed by worms for instance. Heavy rainfalls that lead to erosion also considerably influence the infiltration capacity of the soil.

Hydrology 3

Uploaded by

Copyright:

Available Formats

Hydrology 3

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hydrology 3

Uploaded by

Copyright:

Available Formats

UNIVERSITT STUTTGART

INSTITUTE OF HYDRAULIC ENGINEERING CHAIR OF HYDROLOGY AND GEOHYDROLOGY

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

STRUCTURE OF DRAINAGE BASIN MODELS................................................ 13

MODELS OF RUNOFF FORMATION ................................................................ 17

BASIS AND METHODS OF SYSTEMS HYDROLOGY ..................................... 28

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy 4.4 4.5

CONCEPTS OF HYDROLOGIC MODELS ........................................................ 43

COMBINATION OF MODEL CONCEPTS IN DRAINAGE BASIN MODELS ..... 70

FLOOD ROUTING MODELS ............................................................................. 74

PHYSICALLY BASED HYDROLOGICAL MODELS........................................ 106

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy 8.1

Hydrology III - III

Fundamental principles ............................................................................................................... 106

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Fundamental principles of river basin modeling

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Principal methods of river basin modeling

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Statistical technique (distribution)

Combination of statistical and deterministic methods

Structure of river basin models

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Drainage basin models

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

time t areal precipitation iN Q(t) discharge Q drainage basin time t

Principle of a drainage basin model, determination of streamflow from areal precipitation

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Basic elements of a drainage basin model

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Principle of a streamflow model, one tributary, (floodrouting)

Principle of a streamflow model with several tributaries (flood forecasting)

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

time t discharge Q Z QZ inflow

time t intermediate catchment

Principle of a streamflow model with intermediate drainage basin

Complex river basin model

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

water flow model

rainfall - runoff model

Interfaces of a river basin model

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

water flow mo including catch (lateral inflow)

Structure of a complex river basin model

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Analysis and synthesis

Synthetic streamflow hydrographs

calibration of comparison parameters

m athem atical model (parameters)

Calibration of a mathematical hydrologic model

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Figure 1.10: Data flow in the course of analysis and synthesis

CHAIR OF HYDROLOGY AND GEOHYDROLOGY

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

Structure of drainage basin models

CHAIR OF HYDROLOGY AND GEOHYDROLOGY

Prof. Dr. rer. nat. Dr.-Ing. Andrs Brdossy

flood increase vertex flood hydrograph Q