Heliyon 8 (2022) e11941

Contents lists available at ScienceDirect

Heliyon
journal homepage: www.cell.com/heliyon

Research article

Assessing the economic effects of drought using Positive Mathematical

Planning model under climate change scenarios
Ali Ghaffari a, *, Mohsen Nasseri b, Abulfazl Pasebani Someeh c
a
Department of Civil and Environmental Engineering, Lehigh University, Bethlehem, PA 18015, United States
b
School of Civil Engineering, College of Engineering, University of Tehran, Tehran, Iran
c
ACECR, Tehran, Iran

H I G H L I G H T S G R A P H I C A L A B S T R A C T

Different types of droughts can have

complex economic effects on the culti-
vation regime.

Using SPI, water cost coefficient is
developed to translate the water defi-
ciency caused by future droughts into
added costs.

Positive Mathematical Planning (PMP)
model is used to calculate the future
optimal cultivation area for various
crops.

Future cultivation regime is affected by
water deficiency, estimated value of the
crops, and availability of resources.

A R T I C L E I N F O A B S T R A C T

Keywords: In recent decades, regions all around the world have experienced severe droughts adversely affecting their
Agroeconomic model agricultural production. Climate change, along with limited access to water will alter future production and
Meteorological drought agricultural development. The purpose of this study is to provide a perspective for the future cultivation regime in
Agricultural drought
the Divandarre region in the Sepidrood catchment in Iran, using historical climatic, agricultural, and economic
Supply-demand
information. Future precipitation values are determined for three climate scenarios, then downscaled and con-
Downscaling
MLS verted to pixel-based precipitation maps using the Moving Least Squares method. Future droughts are identified
SPI using the Standardized Precipitation Index at 3, 6, and 9-month intervals based on precipitation values and the
PMP relationship between different types of droughts (meteorological, agricultural and hydrological). We introduce a
new coefficient, the water cost coefficient, derived from drought characteristics that captures the added irrigation
cost in drought years because of increased water price. Using the Positive Mathematical Planning method and
considering limited land and water, predicted future prices and costs based on a linear regression of supply-
demand, and the annual water cost coefficient values, an agroeconomic model is built. After prediction of
future price and cost based on historical data from 2005 to 2018, we run future scenarios based on various price
and cost values to determine the optimal annual cultivation area for each crop from 2020 to 2040. All scenarios
indicate a decline in cultivation area for all crops making agriculture less beneficial in the future. The cultivation
regime moves away from more water-consuming products with less economic value (e.g. watermelon) toward less
water-consuming, more expensive products (e.g. lentils). The findings of this model along with expert economic

* Corresponding author.
E-mail address: [email protected] (A. Ghaffari).

https://doi.org/10.1016/j.heliyon.2022.e11941
Received 22 July 2022; Received in revised form 30 September 2022; Accepted 21 November 2022
2405-8440/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
A. Ghaffari et al. Heliyon 8 (2022) e11941

judgments help determine the economic effects of climate change on irrigation, farmers' decisions, and water
policies, including water markets, and improving irrigation efficiency. Authorities and farmers could adapt to
drought shocks and changes in the market while experiencing less revenue loss.

1. Introduction livelihoods” and “organizing consultative meetings between experts and

farmers to combine indigenous and modern knowledge for increasing the
Drought is a common and recurring destructive natural phenomenon effectiveness of drought mitigation programs” are the most effective
that has socio-economic consequences by affecting agriculture. Drought approaches to deal with drought.
generally occurs when water resources are depleted to such an extent that After determining the effects of drought on the growth of agricultural
they can no longer meet the needs of the region. Recent studies have products, these effects are converted to an economic form known as
shown that drought has significantly reduced global ecosystem produc- profit and loss functions. There are several methods to determine the
tion; In general, droughts are classified into four categories: meteorolog- economic effects of drought, the most widely used of which are Market
ical, agricultural, hydrological, and socioeconomic (Dracup et al., 1980) Valuation Techniques, Input-Output Analysis, Computable General
(Wilhite and Glantz, 1985). Drought (and specifically agricultural Equilibrium Analysis (CGEA), Contingent Valuation Method (CVM),
drought) affects agricultural production more than any other natural Choice Experiments (CE), and determining the optimal point of hydraulic
phenomenon due to the gradual onset and duration of the long-term effects capital. Although many models have been developed using these
compared to other natural phenomena such as floods. Intergovernmental methods, there are limitations. For example, they require a lot of accurate
Panel of Climate Change (IPCC) has reported a significant decline in and detailed information to produce results that may not even be reliable
agricultural production in arid and semi-arid countries (such as Iran), due to a lack of time series, or rapid infrastructural changes in the
which indicates that food security conditions in these countries are worse development of the economic system. The accuracy of the results is also
than in previous years (P.R. Shukla et al., 2019). It is also predicted that strongly influenced by the structure of resources, environment, and
climate change will exacerbate these conditions by increasing the fre- policy constraints. In many cases, the cost and time of implementing
quency of drought recurrence mainly due to fluctuations in annual rainfall. these methods are so high that the results are not produced in the design
Many studies have tried to link hydrological drought to changes in planning horizon and as a result, they lose their practical value. For these
vegetation and, consequently, agricultural drought. Lin et al. (2017) tried reasons, the use of mathematical programming models to assess the ef-
to find a relationship between hydrological drought and water-climatic fects of climate change on agricultural economics has become wide-
variables, water resources performance, and vegetation in the Xijian spread today. The reasons for this include fewer empirical constraints in
River in China. The results of this study show while signal reservoir farm modeling, the possibility of defining nonlinear concepts with in-
operation can reduce drought severity in spring and summer, they formation, time and computational constraints, and considering endo-
worsen the autumn droughts. This shows the incapability of a sole water geneity of the final price and risk behavior in the objective function
resources management method in reducing drought effects. Sun et al. (Howitt, 1995). In this paper, the PMP method, one of the most widely
(2018) tried to determine the severity of droughts in the Yangtze River used mathematical methods in impact assessment, has been used to
Basin using WSDI (Water Storage Deficit Index) and information ob- determine the effects of drought. The selection of PMP is due to a lack of
tained from the GRACE satellite. WSDI can capture the historical drought long climatic and economic historical data in the study region, and also
events and therefore, could be a robust and reliable method for charac- large, rapid structural changes because of urbanization, boycotts, and
terization of future droughts. However, the length of data used may not ever-changing governmental policies that cause inconsistencies in data.
be available for smaller and less developed watersheds. Yu et al. (2014) PMP has proven to be a robust method in case of inconsistent and
used the DNDC (DeNitrification-DeComposition) model to simulate plant short-period data.
growth using the nitrogen uptake process by the plant, taking into ac- As discussed, there is a gap that needs to be addressed here. The case
count water and temperature stresses. In this model, plant growth is study here is the representative of croplands in developing countries that
modeled with PGI (Plant Growth Index), mainly focusing on the impact suffer from lack of continuous and reliable climatic and economic data.
of drought, its severity, and duration on the decline in the growth of Large infrastructural changes and abnormalities in historical trends make
maize crops in the United States. Although this model can provide it even more difficult to interprete this limited data or use more advanced
comprehensive yield information for various end-users, it requires very methods. Therefore, the purpose of this study is to prepare a future plan
strong and voluminous base data for accurate calibration. In addition, the for the cultivation regime in the Divandarre region in the Sepidrood
determination of the coefficients related to the desired plant (such as catchment in Iran, using limited historical climatic, agricultural, and
corn) is done through several experiments, which is very time-consuming economic information as inputs to a PMP model. This is done in five
and costly. Zargar et al. (2017) tried to determine the tolerance of plants steps: (1) Predict future precipitation values, (2) Determine future
to drought by examining the effect of drought on plant growth and more droughts and their properties, (3) Predict future economic crop data, (4)
precisely the process of photosynthesis in different plants. Changes in Model calibration, (5) Future simulation to get cultivation areas for
plants in dealing with drought include reduction of trigger pressure, different crops. Each steps is discussed in detail in the following sections.
closure of the stomata during the day, leaf rotation, etc. which depend on
the duration of the stress, the growth stage, the genetic potential, and the 2. Study area
surrounding environment. To determine these responses, as well as the
production of compounds in the plant structure, it is necessary to study The Great Sepidrood catchment is one of the largest water basins in
the basic metabolic processes of the plant in different stages which is Iran and a subset of the Caspian Sea catchment. The area is 59217 square
specific to the region and also costly. Drought is a multi-dimensional kilometers and the main river is Sepidrood. The study area, the Divan-
problem that requires the water sector and the social and agricultural darre region in Kurdistan (Figure 1), is part of the Sepidrood catchment,
sectors to adapt and plan accordingly to solve it. Savari and Shokati which according to the moisture regime, rainfed and irrigated cultivation
Amghani (2022) used the combined model SWOT-FAHP-TOWS to (with an area of 122865.5 ha and 5194.3 ha, respectively) is relevant and
develop 12 appropriate strategies to increase farmers' adaptation to the relationship between agricultural and meteorological drought will be
drought. According to the findings, the two strategies “supporting the meaningful. Agriculture is the main activity of the people and irrigated
development and establishment of microcredit organizations and funds and rainfed crops are both common with the predominant irrigated and
using the spirit of empathy and rich culture of farmers to diversify their rainfed crops in the region being alfalfa and wheat, respectively. In recent

2
A. Ghaffari et al. Heliyon 8 (2022) e11941

Figure 1. Geographical position and territory of Divandarre

years, the hydrological regime has become more complicated than ever well as base year data are obtained from the national statistics of the
with water shortages and frequent droughts that threaten the food se- Ministry of Agriculture.
curity of the region. According to local authorities, the Sepidrood basin The downscaling model used in this study is SDSM (Statistical
has experienced major droughts in the last 15 years which has aggravated DownScaling Model), a linking method between a climatic statistical
the conditions for various crops with the last one being rice. This problem generator and a multilinear fitting method that converts climatic vari-
can seriously threaten the food security in the region and the whole ables at the synoptic scale into local meteorological variables. To better
country since rice is the major source of food in Iran, mainly grown in the match the observed time series, statistical techniques are used to artifi-
Sepidrood basin. This problem is intensified in the Divandarre region due cially increase the variance of the downloaded weather time series. Here,
to very little access to groundwater making the local agro-economy SDSM takes daily precipitations at 64 stations from 1961 to 2006
dependent on rainfall and surface water. Following the main trend in (training: 1961 to 2000; calibration: 2000 to 2006) and after determining
semi-arid parts of Iran, temperature rise and fewer rainfall events have the optimal point carrying the largest amount of information, calculates
caused large losses to farmers. According to the national statistics orga- precipitation values for three future RCP scenarios (i.e., RCP 2.6, 4.5, and
nization of Iran, the total rainfall in Kurdistan in the first half of 2020 was 8.5) from 2006 to 2040. Although the oldest precipitation values are
73.4 mm with a 40% decrease compared to the last year with the least from 1961, they are very scattered and most of the stations (44 out of 64)
value recorded for Divandarre. All the aforementioned reasons make the have data starting from 1987 which makes it hard to choose a longer
Divandarre region a perfect fit for the case study. calibration period.

3. Methodology
3.2. Map generation
The purpose of this study is to provide a perspective for the future
cultivation regime in the region using historical, climatic, agricultural, Precipitation values at stations need to be converted to pixel-based
and economic information. Historical precipitation data in this area were maps for spatial analysis. Moving Least Squares (MLS) is one of the
collected from 64 precipitation stations and after downscaling, they were most important computational frameworks, first proposed by Lancaster
converted into pixel-based precipitation maps with a spatial resolution of and Salkauskas (1981), for analyzing information based on local values,
1km by 1km using the MLS (Moving Least Squares) method. In this their behavior, and significance through the moving computational
process, three climate scenarios based on CanESM2 were used, which domain and kernel weighting functions. In MLS, the estimation is ob-
were developed in the framework of CMIP5. CanESM2 includes the tained using approximation (regression) in the points that are distributed
following GCMs: CanCM4, OGCM4, CMOC, and CTEM. irregularly as a constraint of distance. More specifically, each point re-
Using precipitation data, SPI (Standardized Precipitation Index) ceives a weight proportional to its distance so that neighboring points
values are calculated for three scenarios to determine future droughts have more participation in the estimation process. All points within the
and their characteristics (i.e., intensity, magnitude, and duration). Based [local] moving domain participate in the estimation process, while the
on drought characteristics, an additional cost parameter as the water cost rest are not considered. A comprehensive explanation of this method is
coefficient is calculated to capture the added cost of water under various given in Amini and Nasseri (2021). Details of the functions used in this
drought conditions. Converting qualitative drought properties into eco- article are discussed.
nomic parameters, PMP (Positive Mathematical Planning) model ana- Here, the base function is a second-order quadratic function of two
lyzes all possible cultivation options and chooses the best crop to be inputs DEM (longitude, latitude, and height) and climatic values (pre-
cultivated annually (Figure 2). The procedure can be broken down into cipitation). Also, r (effective fitting distance) is calculated as follows:
six steps: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u  2
 2  2
u
t x  xj 2 ð y  yj Þ H  Hj P  Pj
r¼ þ þ þ (1)
α γ ρ μ
3.1. Data collection and downscaling
Where x, y, H, and P are the longitude (degrees), latitude (degrees),
Inputs include rainfall, groundwater and land use maps, cost of height (meters), and precipitation (millimeters), respectively. α, γ, ρ and
cultivation and irrigation, crop price, yield, water demand, and cultiva- μ are distance parameters that need to be optimized. Because of higher
tion area. Economic data including the cost of production of agricultural convergence speed and efficient computational performance, the Shuffle-
products (2001–2018) and the on-farm price of products (2006–2018) as Complex Evolution (SCE) optimization approach is used here.

3
A. Ghaffari et al. Heliyon 8 (2022) e11941

Figure 2. Flowchart showing the methodology of the study.

3.2.1. SPI calculation II .Drought duration is divided by the maximum drought duration to
In this study, SPI drought index in 3, 6, and 9-month intervals are determine the relative duration of the drought. (fd )
used to examine the relationship between meteorological, agricultural, III. Drought magnitude is divided by the maximum drought magni-
and hydrological droughts. Using pixel-based precipitation data at a tude to determine the relative magnitude of the drought. (fm )
monthly scale, SPI values are calculated for the three climatic scenarios IV. Finally, the water cost coefficient is equal to:
for 2020 to 2040.
fw ¼ ðfi  fd  fm Þ þ 1
3.2.2. Modeling water deficiency 1 is added at the end to make all values greater than 1 (which is an
Droughts reduce the available water resources and thus increase the indicator of normal years).
price of water. Here, water cost coefficient introduces water deficiency to After determining the future drought events for the scenarios RCP 2.6,
the model to quantitatively assess various drought events and compare following the steps above, the water cost coefficients are calculated for
the profitability of different crops. The first step in this process is to each drought event and attributed to the corresponding year (see
calculate irrigation costs for future years as a part of the growing stage Table 2).
costs. Assuming irrigation costs change relative to growing costs, irri-
gation costs for the future are obtained using linear regression.
Deviation from the annual average rainfall cannot capture all drought 3.3. Economic modeling
characteristics. Therefore, the SPI-6 index (SPI in 6-month intervals) is
used to identify agricultural drought. Drought begins when the SPI index In this study, the PMP method is used to build an agricultural pro-
becomes negative, remains negative, and during the negative period, duction model using nonlinear production cost functions. The PMP uses
reaches values less than -1. Using this definition, droughts are identified base year yield data to provide self-calibrating models of agricultural
during the years 2020–2040. Drought characteristics are (1) Intensity:
the smallest SPI value in the desired period; (2) Duration: number of
months with SPI   1; 3) Size: Absolute value of SPI values during the Table 1. Quantification of drought intensity by the drought intensity factor.
drought period. Using these three parameters, the water cost coefficient
Qualitative intensity of SPI Drought intensity
is calculated as follows: water conditions factor (fi )
Almost normal -0.99 to 0 1
I. According to Table 1, for any SPI value, a number from 1 to 4 is Mildly dry -1.49 to -1.00 2
defined as the drought intensity factor (fi ) that shows the intensity Very dry -1.99 to -1.50 3
of a drought event. As shown in this table, the highest number (4)
Severely dry < -2.00 4
is for the most intense drought event.

4
A. Ghaffari et al. Heliyon 8 (2022) e11941

Table 2. Calculation of WCC (Water Cost Coefficient) based on droughts in 2020–2040 according to RCP 2.6 scenario.

Drought Start End Intensity Duration (months) Magnitude fi fd fm WCC

1 2020–10 2023–09 -2.03 36 28.88 4 0.52 0.43 1.90
2 2024–01 2028–04 -2.51 52 59.07 4 0.75 0.88 3.65
3 2028–06 2034–02 -2.97 69 67.29 4 1.00 1.00 5.00
4 2035–11 2036–04 -1.52 6 4.11 3 0.09 0.06 1.02
5 2037–01 2037–09 -1.59 9 8.69 3 0.13 0.13 1.05
6 2037–11 2040–09 -2.47 35 35.92 4 0.51 0.53 2.08

production and resource use, in line with microeconomic theory, 4. Results

considering the heterogeneity of land and livestock. PMP automatically
calibrates the data without using "flexibility constraints". The resulting 4.1. Average precipitation by climate scenarios
models are more flexible in their responses to policy changes, and the
underlying factors are recognizable in production variance or resource After running the meteorological model, the average rainfall values
elasticities. Numerous researchers have used PMP in their studies. For for the years 2020–2040, for three climate scenarios are 32.03 mm, 31.55
example, water pricing in China (Huang et al., 2016), evaluating the mm, and 33.00 mm for RCP 2.6, 4.5, and 8.5, respectively. Average
impacts of variations in crop profitability and market innovations on rainfall decreases from RCP 2.6 to RCP 4.5 to a small extent but it in-
farm profitability, land use and water consumption (Donati et al., 2013), creases from RCP 4.5 to RCP 8.5 which could be possible due to the
increasing the price and reducing the amount of water, and changing the change in hydrological regime in the region. However, a single value for
price of agricultural products to use deficit irrigation (Cortignani and average rainfall over the coming years cannot show the temporal dis-
Severini, 2009). The PMP method seeks to accurately calibrate area, tribution of precipitation and therefore, is not enough for drought
production, and price. Kasnakoglu & Bauer used the PMP method in one modeling.
of the sectional models introduced by Hazel and Norton (1986). The
results for the Turkish Agricultural Sector Model showed continuous 4.2. Drought index by climate scenarios
calibration over seven years (1998).
Here, PMP determines the optimal cultivation area of each crop under After obtaining pixel-based precipitation maps, SPI in three intervals
different economic and climatic scenarios. Model is formulated as: (i.e., 3, 6, and 9 months) is calculated (Figure 3). SPI changes almost
similarly in RCP 2.6 and RCP 4.5 although SPI magnitude in RCP 2.6 is
X
6 X
4
max ai ðpi  ci Þ þ ar ðpr  cr Þ larger which means more precipitation in wet years and less in dry years.
i¼1 r¼1 All scenarios show a decrease in precipitation indicating several
droughts. However, droughts and wet years lagged for 4–5 years in RCP
X
6 8.5 compared to RCP 2.6 and RCP 4.5 which could be due to different
s:t: ai  Ai temporal distribution of precipitation in RCP 8.5.
i¼1

X
4 4.3. Model calibration
ar  Ar
r¼1 In the original PMP model (Howitt, 1995) variables are bound to
change in the range of þ0.0001 and -0.0001 from the original values. In
ai ; ar  0 this study, however, the calibration is done without tight bounding
constraints of variables. By changing the objective function and modi-
Where: fying the costs (i.e., removing the land costs from the total costs), the
i and r refer to irrigated and rainfed crops, respectively; p, a, c, and A upper and lower limits are removed and the base year values are
are the price ($/hectare), cultivation area (hectares), cost ($/hectares), reproduced with acceptable accuracy which adds to the model flexibility.
and maximum available area, respectively. The results of this calibration along with the values of the base year by
type of cultivation are given in Table 3 and Table 4.
3.4. Model calibration
4.4. Prediction of crop prices and costs
Calibration is performed for six irrigated crops and four rainfed crops
based on the 2015–2016 water year; That is, the areas generated by the Crop prices and costs are the most important economic inputs of the
model should converge to the base year areas. In the original paper, PMP model. Five mathematical functions for extrapolating the price and
Howitt (1995) does this by using bounding constraints. In this paper, cost of production are defined in the form of five economic scenarios.
calibration is done without the need to define tight constraints bounding Prediction accuracy results based on various error indices are provided in
the variables, by softening the constraints and removing the land costs Table 5.
which makes sense since the majority of farmers are also land-owners in Results for crops’ price and cost values based on piece-wise linear
the study region. regression are represented in the following figures for both observation
and trend data sets (see Figures 4-6).
3.5. Future scenarios
4.5. Economic model results
After calibration, the model runs for future scenarios with price and
cost values determined by various prediction functions. Four different Droughts will limit access to water and therefore, change the water
functions are used to predict future price, cultivation cost, and irrigation price. This is an important input to determine the profitability of one crop
cost. Based on these functions, future scenarios include (1) linear, (2) over another in the cultivation pattern. In this part, the agro-economic
quadratic, (3) piece-wise linear (4) fixed inflation rate. model determines the optimum cultivation area for each crop in the

5
A. Ghaffari et al. Heliyon 8 (2022) e11941

Figure 3. SPI values in 2040–2020 based on three climate scenarios.

future considering water and land availability as well as the crop prices differ according to crop category and whether the system is rainfed or
and costs and also supply and demand relationships determined by linear irrigated. In this study, irrigated crops could take advantage of the price
regression models. The PMP model runs four times (number of price and effect due to quantity effect losses in rainfed crops.
costs prediction scenarios) for each of the three climate change scenarios In order to run future simulations, crops price and cost values need to
(RCP 2.6, 4.5, and 8.5). The results are provided in Figures 7 and 8 for be determined for the modeling priod (2020–2040) using historical data.
irrigated and rainfed crops, respectively. However, there is a very large jump in the historical cost data between
the years 2014 and 2015 (largely due to changes in government policies
5. Discussion towards fuel prices and also boycotts) such that values increase more
than 10 times on average. Due to this jump, the linear function is unable
Identifying future droughts is an important step in this study that was to predict future prices accurately. The quadratic function can capture
done using SPI values in 3-, 6-, and 9-month time intervals. In short, SPI-6 this jump, however, due to the sharp increase in exponential function
best captures agricultural droughts. In Figure 3, it can be seen that slope after passing the jump, it can lead to extremely high values for price
intense droughts occur in years with low SPI values but the continuation in years to come. The piece-wise linear function can bring the best of both
of a negative SPI could also aggravate the conditions. In the years worlds by capturing the large jump in the data and yielding fairly rational
2020–2026, the rainfall does not further decrease, but it stays constantly price values for the modeling period. In this method, two functions are
low and causes a major drought, especially in RCP 2.6 and RCP 8.5. This used to fit two separate sections of the time series. The separation point is
result is only obtainable from the SPI diagram based on the duration defined so that the variance difference between the two sections is
concept. A similar case can be observed in the years 2034–2040. SPI-6 maximized. The fourth scenario is defined by multiplying the present
results (agricultural drought) are only valid if the general trend is in values by a fixed value (called the inflation rate). Here, the inflation rate
accordance with SPI-9 results (hydrological drought). In other words, in (IR) is applied annually from 2020 to 2040 to price (IR ¼ 0.2), cultivation
the long run, agricultural drought leads to the drying of natural water cost (IR ¼ 0.3), and irrigation cost (IR ¼ 0.4) to get a time series of values
resources causing a severe hydrological drought from 2024 to 2033. throughout the modeling period.
There are similar drought events in the early and final years of the In summary, the best prediction function is piece-wise linear regres-
modeling period but there is a major drought in the years 2030–2034 sion (Figures 4 and 5) because it can capture the large jump in the data
with the drought peak in 2032. To further validate the data, SPI-3 is used set with rational future price values. For a more accurate price and cost
to indicate short droughts known as meteorological droughts that are prediction, we need to conduct economic studies on the supply-demand
used to characterize the beginning and also the most stressful years
during a drought period (in here, the years 2022, 2028, and 2039). We
can also see a chronological order of meteorological, agricultural, and
Table 4. PMP model calibration results along with base year (2015) cultivation
hydrological droughts from 2025 to 2039. Continuation of hydrological areas for rainfed products.
droughts and failure to make appropriate adaptations can lead to socio-
No Product Area in 2015 (ha) Area from PMP (ha) Error (%)
economic drought. Espinosa-Taso
n et al. (2022) studied the economic
1 Cereals 91881.5 91882.27 0.0008
impact of multiyear droughts by applying the economic surplus in
Andalusia and found that the effects of hydrological drought on farmers 2 Legumes 30963.5 30963.74 0.0008
3 Industrial products 10.6 9.60 9.4348
4 Cucurbitaceae 9.9 9.90 0.0503
Table 3. PMP model calibration results along with base year (2015) cultivation
areas for irrigated products.
No. Product Area in 2015 (ha) Area from PMP (ha) Error (%) Table 5. Error indices for 5 economic scenarios; There is no error for Scenario 4
1 Cereals 570.0 570.03 0.0044 (Fixed Inflation Rate) because no reference data is available for this method.
2 Legumes 16.1 16.10 0.0310 Scenario R-Squared RMSRE MARE R-Squared RMSRE MARE
3 Industrial products 8.9 8.89 0.0562 Cost Cost Cost Price Price Price
4 Cucurbitaceae 15.4 15.39 0.0325 1 0.60 27.10 15.95 0.80 0.52 0.37
5 Vegetables 64.7 64.70 0.0077 2 0.89 13.80 8.14 0.88 0.31 0.22
6 Forage plants 4519.2 4519.19 0.0001 3 0.87 1.54 0.65 0.93 0.36 0.25

6
A. Ghaffari et al. Heliyon 8 (2022) e11941

Figure 4. Observed cost versus fitted cost for different crops (Part 1). The separation point is obvious in all graphs mainly due to large infrastructural changes,
sanctions, and the removal of subsidies on necessary raw materials.

relationships and also determine the growth of the inflation rate in the increased joblessness rates. Several studies have pointed out the same
country due to presidential periods, sanctions, and the relative impor- result; Movahedi et al. (2021) used structural equation modeling to find
tance of different products over time. Implementation of such a model is the most important factors causing the farmers to abandon agricultural
out of the scope of this study. lands. The most effective cause of this phenomenon was the
Except for lentils (Figure 6), the separation point in price prediction managerial-legal factor including the problem of segmented farms, and
happens at the beginning of the time series. Although government sub- not paying enough attention to establishing agricultural incentives. In
sidies are to prevent an increase in prices for strategic crops (cereals and another study, Boazar et al. (2019) mentions that government plans for
legumes), the results question the practicality of these subsidies. Obvi- water resources management were not efficient and they have switched
ously, the free market has not followed the desired path of the authorities from supply-side to demand-side management to prevent rice farmers
mainly due to in-between dealers abusing the market conditions in their from abandoning their farms or cultivating other crops. In another study,
own favor and leaving farmers in large losses. The primary results of this Neisi et al. (2020) analyzed farmers' drought risk management behavior
mismanagement include large immigration toward urban areas and by using Krejcie and Morgan's table method to sample 350 farmers. She

Figure 5. Observed cost versus fitted cost for different crops (Part 2). The graphs indicate a huge cost difference ($/ha) for Cucurbitaceae (watermelon and cucumber)
and vegetables (tomato and potato) in comparison to other products in Figure 7. This is because these products are not covered by government subsidies and are also
more water-consuming which adds to the costs.

7
A. Ghaffari et al. Heliyon 8 (2022) e11941

Figure 6. Observed price versus fitted price for major crops. Except for lentils, the separation point is placed at the beginning of the historical data which indicates a
constant price increase.

found that policies need to make farmers take their own risk management drought shocks on household behavior and found that farmers move
decisions and have access to a variety of tools and strategies. toward the implementation of coping mechanisms including
The last step of the modeling process involves future simulations for non-agricultural job opportunities in drought-affected areas of Fars, Iran.
predicted price and cost while taking into account various climatic sce- As the drought ends, the cultivation area for cereals increases but does
narios. In general, the cultivation area for all crops decreases in all not reach the base year value (Figure 7-Cereals). During drought years,
climate scenarios making it less beneficial to cultivate crops in the future. cereals are still one of the best crops to grow due to their large demand
Drought events are more severe and relatively long in the RCP 4.5 due to capacity and the potential for exportation but as the drought duration
less average rainfall leading to constant negative SPI values and there- increases, the profitability of cereals declines. Savari et al. (2022) studied
fore, a further decline in the cultivation area for more water-consuming the effect of drought severity on wheat crops and found that further
crops. This decline is exacerbated during the drought years for cereals in drought aggravation increases the farmers' vulnerability because of
all three climate scenarios with the RCP 8.5 scenario lagged. Khalili et al. higher vulnerability levels in regions with more critical conditions and
(2021) conducted a quantile regression model to find the effects of longer droughts.

Figure 7. Cultivation area for six groups of irrigated crops under three climate change scenarios for the modeling period (2020–2040). Except for cereals and legumes,
RCP 8.5 shows the same pattern as RCP 2.6 and RCP 4.5 with an average lag of 4–5 years because drought events happen later in the RCP 8.5 scenario.

8
A. Ghaffari et al. Heliyon 8 (2022) e11941

Figure 8. Cultivation area for four groups of rainfed crops for the modeling period (2020–2040). Note that vegetables and forage crops cannot be rainfed and
therefore, are not included in this figure.

Legumes are the second choice to go to due to the sharp price increase against drought. Proper management with an appropriate cultivation plan
but since this high price is not sustainable (because of low demand), the can change the drought threat into an opportunity for the farmers.
cultivation area for legumes decreases rapidly (Figure 7-Legumes). Musolino et al. (2018) conducted an empirical investigation on the
Farmers may not be able to benefit from this quick shift in the market due distributive effects of drought events in some areas of Southern Europe and
to their common fatalistic behavior and therefore, prior cultivation plans found that drought can create not only “losers”, but also “winners” and the
could help them adapt to the market earlier. Zarafshani et al. (2012) key factor to analyze the gains obtained by farmers in price.
mentioned this fatalistic behavior as one of the psychological factors that One important constraint is the capability of the land for crop rotation
increase farmers’ vulnerability to droughts because it directs them to- (for example, switching between forage crops and vegetables). There are
ward taking more passive coping strategies. two sides to this problem: (1) Physical constraint and plant physiology:
The area of industrial products decreases during the drought years The growing process is different for each crop but given the right ma-
and returns to the previous level at the end of the drought. Industrial chinery and tools, it is possible to implement crop rotation; (2) Economic
products are among the crops with moderate water demand and due to constraint: there must be sufficient demand for the increased supply of
the higher water cost, they are not the best choice in drought years crops which is addressed by linear supply-demand relationships here.
(Figure 7-Industrial Products). Due to the fixed supply-demand rela-
tionship, the area for these crops does not increase further than the base 6. Limitations
year value. Despite higher water demand, the higher relative profitability
increases the Cucurbitaceae area slightly during the drought years. As with any model, there are limitations to the model implemented in
Similar to industrial products, due to fixed supply-demand, the area this study. It has been addressed by several studies that identifying
returns to the base year values after the drought (Figure 7-Cucurbita- drought using only one variable may be subject to bias. In this study, SPI
ceae). The same procedure repeats for Vegetables (Figure 7-Vegetables). was used to determine droughts which is still the most commonly used
As for the forage crops, the area decreases during the drought years drought indicator along with PDSI. This is because the Divandarre region
since they are among the highest water-consuming crops that reduce the suffers from lack of continuous and reliable historical climatic data
relative profitability (Figure 9-Forage Crops). Forage crops are among the leaving the authors with precipitation as the only reliable historical data.
strategic crops due to their consumption by livestock. This pattern is also However, using more climatic variables (e.g., temperature, evapo-
mentioned in other studies in similar semi-arid regions. Mehdipour et al. transpitation, etc.) could lead to a more accurate drought identification
(2022) analyzed drought adaptation strategies to evaluate rural house- which will be a future consideration of this study.
hold liveability and found that increased prices in forage crops could Another limitation is the change in crop yield depending on climatic
damage the livestock population since farmers can no longer afford to and agricultural conditions. The best way to do this is to build an
feed them finally resulting in a loss of revenue and access to food. Having experimental crop yield model based on climatic and agricultural stress
an overview of the future problems could improve the policies to avoid tests. Such data is not available at the moment and the other methods
losses in interconnected sectors. (such as Stewart's equation) are either too simple or specific to a case
Rainfed crop cultivation area does also decrease in the long run study.
(Figure 8). However, due to the non-intervention of irrigation costs, the One of the most important inputs of the PMP model are future crops’
area decreases very slowly. The decreased area for industrial Cucurbi- price and cost values. Currently, these values are determined using piece-
taceae is even smaller because of the fixed supply-demand relationship. wise regression method. More elaborate economic models along with
Industrial products are no longer economically viable and therefore, are expert judgements could increase the accuracy of these predictions. Such
constantly zero. Savari and Moradi (2022) evaluated the effectiveness of models could also provide a more detailed relationship between supply
drought adaptation strategies and found that the most important adap- and demand. However, it should be noted that unstable economic con-
tation classes under drought conditions were farming strategies and crop ditions in Iran make it extremely difficult to devise reliable economic
management. This was mainly done by a reduction in rainfed wheat models. Surveys at farm level could be a beginning point for this purpose.
cultivation area in Fars simply because it was not economically viable to PMP is the best choice modeling tool given conditions in the study
harvest the crops. area. However, the model can be improved if more realistic constraints
The optimization problem is solved with inequality constraints are considered because conceptually, it is an optimization model. These
because, in reality, farmer prefers not to cultivate over not profitable constraints can include manpower, machinery, and movement of prod-
cultivation. This has led to widespread immigration of farmers to cities and ucts between states. This is only possible if detailed data is available of
the abandonment of agricultural lands. If labor forces are needed in other the import and export of crops at the country level and between states.
sectors (such as industry), this will lead to the optimal distribution of Physical constraints could also be employed to implement the idea of
manpower but in Iran, this will increase the unemployment rate. Nasrnia time lag between crop rotations.
and Ashktorab (2021) assessed the drought resilience patterns of rural
households. She found that governmental support for occupations such as 7. Conclusion
animal husbandry resulted in more engagement of residents in these ac-
tivities, leading to increased employment, reduced immigration to urban Droughts, as the most important threat to food security in an area,
areas, increased household incomes, and increased household resilience reduce the production and quality of agricultural products by changing

9
A. Ghaffari et al. Heliyon 8 (2022) e11941

the amount and frequency of rainfall. Identifying future droughts can References
provide a cultivation plan in the modeling period to guide policy-makers
and practitioners in mitigating drought effects by optimal use of limited Amini, Y., Nasseri, M., 2021. Improving spatial estimation of hydrologic attributes via
optimized moving search strategies. Arabian J. Geosci. 14 (8), 723.
land and water. In this paper, future droughts along with their charac- Boazar, M., Yazdanpanah, M., Abdeshahi, A., 2019. Response to water crisis: how do
teristics (i.e., intensity, duration, and magnitude) are determined by SPI Iranian farmers think about and intent in relation to switching from rice to less water-
values based on predicted precipitations for three climatic scenarios (i.e., dependent crops? J. Hydrol. 570, 523–530.
Cortignani, R., Severini, S., 2009. Modeling farm-level adoption of deficit irrigation using
RCP 2.6, 4.5, and 8.5). To simulate water deficiency, we introduce a new Positive Mathematical Programming. Agric. Water Manag. 96 (12), 1785–1791.
factor, the water cost coefficient, that shows the actual price of water in Donati, M., Bodini, D., Arfini, F., Zezza, A., 2013. An integrated PMP model to assess the
drought years. Using the PMP method, an agroeconomic model de- development of agro-energy crops and the effect on water requirements. Bio base
Appl. Econ. 2, 301–321.
termines the best cultivation area for each crop given future price and Dracup, J.A., Eong, K.$, Paulson, E.G., 1980. On the definition of droughts. In: WATER
costs based on linear supply-demand relationships. Results of this model RESOURCES RESEARCH, 16. Issue 2.
could provide the authorities and the farmers with a generic cultivation Espinosa-Tas
on, J., Berbel, J., Guti
errez-Martín, C., Musolino, D.A., 2022. Socioeconomic
impact of 2005–2008 drought in Andalusian agriculture. Sci. Total Environ. 826,
plan for the modeling period (2020–2040).
154148.
Here, linear supply-demand relationships are used to address this Hazel, E.B.R., Norton, R.D., 1986. Mathematical Programming for Economic Analysis in
problem but more elaborate economic models could yield more accurate Agriculture. Macmillan.
prices by considering transportation, import-export, sanctions, etc. The Howitt, R.E., 1995. Positive mathematical programming. Am. J. Agric. Econ. 77 (2),
329–342.
quality of the proposed model could be improved with expert judgments Huang, Y.F., Ang, J.T., Tiong, Y.J., Mirzaei, M., Amin, M.Z.M., 2016. Drought forecasting
from economists and local authorities by adding more constraints to the using SPI and EDI under RCP-8.5 climate change scenarios for langat River Basin,
optimization problem. Another limitation of this study is that deficit Malaysia. Procedia Eng. 154, 710–717.
Kasnakoglu, H., Bauer, S., 1998. Concept and Application of an Agricultural Sector Model
irrigation (not enough rainfall) could lower the quality of crops and for Policy Analysis in Turkey. Agricultural Sector Modeling.
therefore, the price. This item is not considered in this study due to the Khalili, N., Arshad, M., K€achele, H., Farajzadeh, Z., Müller, K., 2021. Drought shocks and
lack of experimental data showing the relationship between water farm household consumption behaviour: insights from Fars province of Iran. Int. J.
Disaster Risk Reduc. 66, 102625.
amount and crop yield. There are other formulae connecting crop yield to Lancaster, P., Salkauskas, K., 1981. Surfaces generated by moving least squares methods.
evapotranspiration and precipitation but they are outdated and also not Math. Comput. 37 (155), 141–158.
available locally. Lin, Q., Wu, Z., Singh, V.P., Sadeghi, S.H.R., He, H., Lu, G., 2017. Correlation between
hydrological drought, climatic factors, reservoir operation, and vegetation cover in
the Xijiang Basin, South China. J. Hydrol. 549, 512–524.
Declarations Mehdipour, S., Nakhaee, N., Khankeh, H., Haghdoost, A.A., 2022. Impacts of drought on
health: a qualitative case study from Iran. Int. J. Disaster Risk Reduc. 76, 103007.
Movahedi, R., Jawanmardi, S., Azadi, H., Goli, I., Viira, A.H., Witlox, F., 2021. Why do
Author contribution statement
farmers abandon agricultural lands? The case of Western Iran. Land Use Pol. 108,
105588.
Ghaffari, Ali: Conceived and designed the experiments; Performed the Musolino, D.A., Massarutto, A., de Carli, A., 2018. Does drought always cause economic
experiments; Analyzed and interpreted the data; Contributed reagents, losses in agriculture? An empirical investigation on the distributive effects of drought
events in some areas of Southern Europe. Sci. Total Environ. 633, 1560–1570.
materials, analysis tools or data; Wrote the paper. Nasrnia, F., Ashktorab, N., 2021. Sustainable livelihood framework-based assessment of
Nasseri, Mohsen: Conceived and designed the experiments; Analyzed drought resilience patterns of rural households of Bakhtegan basin, Iran. Ecol.
and interpreted the data; Contributed reagents, materials, analysis tools Indicat. 128, 107817.
Neisi, M., Bijani, M., Abbasi, E., Mahmoudi, H., Azadi, H., 2020. Analyzing farmers’
or data. drought risk management behavior: evidence from Iran. J. Hydrol. 590, 125243.
Pasebani Someeh, Abulfazl: Conceived and designed the experiments; Shukla, P.R., Skea, J., Calvo Buendia, E., Masson-Delmotte, v., P€ ortner, H.-O., 2019. IPCC,
Analyzed and interpreted the data; Contributed reagents, materials, 2019: Climate Change and Land: an IPCC Special Report on Climate Change,
Desertification, Land Degradation, Sustainable Land Management, Food Security, and
analysis tools or data; Wrote the paper. Greenhouse Gas Fluxes in Terrestrial Ecosystems.
Savari, M., Eskandari Damaneh, H., Eskandari Damaneh, H., 2022. Drought vulnerability
Funding statement assessment: solution for risk alleviation and drought management among Iranian
farmers. Int. J. Disaster Risk Reduc. 67, 102654.
Savari, M., Moradi, M., 2022. The effectiveness of drought adaptation strategies in
This research did not receive any specific grant from funding agencies explaining the livability of Iranian rural households. Habitat Int. 124, 102560.
in the public, commercial, or not-for-profit sectors. Savari, M., Shokati Amghani, M., 2022. SWOT-FAHP-TOWS analysis for adaptation
strategies development among small-scale farmers in drought conditions. Int. J.
Disaster Risk Reduc. 67, 102695.
Data availability statement Sun, Z., Zhu, X., Pan, Y., Zhang, J., Liu, X., 2018. Drought evaluation using the GRACE
terrestrial water storage deficit over the Yangtze River Basin, China. Sci. Total
Data is publicly available from national data websites which are Environ. 634, 727–738.
Wilhite, D.A., Glantz, M.H., 1985. Understanding: the drought phenomenon: the role of
mentioned in the manuscript. definitions. Water Int. 10 (3), 111–120.
Yu, C., Li, C., Xin, Q., Chen, H., Zhang, J., Zhang, F., Li, X., Clinton, N., Huang, X., Yue, Y.,
Declaration of interest's statement Gong, P., 2014. Dynamic assessment of the impact of drought on agricultural yield
and scale-dependent return periods over large geographic regions. Environ. Model.
Software 62, 454–464.
The authors declare no competing interests. Zarafshani, K., Sharafi, L., Azadi, H., Hosseininia, G., de Maeyer, P., Witlox, F., 2012.
Drought vulnerability assessment: the case of wheat farmers in Western Iran. Global
Planet. Change 98–99, 122–130.
Additional information Zargar, S.M., Gupta, N., Nazir, M., Mahajan, R., Malik, F.A., Sofi, N.R., Shikari, A.B.,
Salgotra, R.K., 2017. Impact of drought on photosynthesis: molecular perspective.
No additional information is available for this paper. Plant Gene 11, 154–159.

10