Hydrogeology Journal (2022) 30:1399–1415

https://doi.org/10.1007/s10040-022-02469-6

REPORT

Modelling groundwater flow in the Plaine du Nord–Massacre shallow

aquifer, Haiti
Luc Arnaud 1 & Alexis Gutierrez 1 & Inès Zegoulli 2 & Nyankona Gonomy 3

Received: 17 January 2021 / Accepted: 22 February 2022 / Published online: 21 April 2022
# The Author(s) 2022

Abstract
To assist in the sustainable management of water resources, a groundwater model of the shallow groundwater in the Plaine du
Nord–Massacre transboundary aquifer system was developed. The Plaine du Nord–Massacre Aquifer is one of the largest aquifer
systems of Haiti, supposedly bearing a high potential for water supply and irrigation. The shallower part is already tapped by
hundreds of domestic wells and hand-pumped public and private boreholes. In the framework of the AVANSE project (Appui à
la Valorisation du Potentiel Agricole du Nord, pour la Sécurité Économique et Environnementale), a USAID funded project
aiming at developing sustainable agriculture in the region, a groundwater model was built to assist in decision making. The finite-
difference-based MARTHE software was used for the model. Due to the lack of monitoring data, the calibration of the model
could only be carried out under steady-state hydraulic conditions. It should thus be considered as a first step towards a better
characterisation of the aquifer’s potential. However, the model computed a groundwater balance at the scale of the aquifer
showing that significant water resources can be easily mobilized if one carefully manages to rule out local operational difficulties.
The simulated groundwater drawdowns using transient simulation of irrigation scenarios are pluri-metric, reaching 6–7 m. New
irrigation wells must be deeper than the current average depth of wells to support agricultural development projects.

Keywords Haiti . Alluvial aquifer . Groundwater modelling . Groundwater development . Agriculture

Introduction as well as larger communities such as Cap-Haitien, the second

largest city of the country where an extension and rehabilita-
The Haiti alluvial plains system has been recognized for a long tion program of the water supply network is ongoing.
time as the country’s greatest potential for directly exploitable Hundreds of hand-pumps and thousands of domestic wells
groundwater resources (Adamson et al. 2016; Margat 1990; supply all other smaller communities from the upper part of
Martin and Moussu 1977). Stretching over the departments of the aquifer (Gutierrez 2011). The population of Cap-Haitien
North and North-East, the alluvial system of the Plaine du (Arrondissement) is estimated to be 357,000 inhabitants (IHSI
Nord/Massacre transboundary alluvial aquifer, hereby called 2015) and could reach nearly 600,000 by 2025 (Archambault
Plaine du Nord aquifer for brevity, is considered a major aqui- and Biscan 2018). The Plaine du Nord aquifer is also the main
fer (UNDP 1991), extending from the plain of the Limbé water source for industrial sites, power plants, and agricultural
River in the west to the Massacre River, whose transboundary irrigation. Indeed, the alluvial aquifer plays a major role in the
watershed extends into the Dominican Republic (ISARM and supply of water for the whole north and north east of Haiti.
UNESCO 2009). The alluvial aquifer system is used to supply In addition to growth in demand for water, climate change
water to smaller towns such as Fort Liberté and Terrier Rouge, in the northern Caribbean is expected to result in intensified

This article is part of the topical collection “Advancements in

hydrogeological knowledge of Haiti for recovery and development”

* Alexis Gutierrez
[email protected] 2
IZIC, 2044 Mornag, Tunisia
3
Département du Génie Rural de la FAMV /UEH, Haiti State
1
BRGM, BP 36009, Orléans Cedex 2, France University, Ambroise, Haiti
1400 Hydrogeology Journal (2022) 30:1399–1415

droughts due to rising temperatures and decreasing annual surface watersheds, from west to east: Haut du Cap, Grande
rainfall (Hall et al. 2013). A rise in sea level is also expected Rivière du Nord, Trou du Nord, Marion, and Jassa (Fig. 1).
in the region (Le Cozannet et al. 2020), which will exacerbate The watershed of the Jassa River is a tributary of the Massacre
saltwater intrusion in coastal aquifers (IPCC 2007). Adamson River which has a watershed that extends beyond the border
et al. (2016) and Gutierrez et al. (2012) have already docu- with the Dominican Republic. The average annual rainfall
mented saltwater intrusion in the Plaine du Nord aquifer. varies between 1,000 and 1,300 mm, with a winter dry season
Collectively, these threaten water security for all users. and a short dry season in July–August. The average annual
The Appui à la Valorisation du Potentiel Agricole du Nord, temperature varies between 22 and 27 °C.
pour la Sécurité Économique et Environnementale (AVANSE)
[Support for the Valuation of the Agricultural Potential of the Geological and hydrogeological setting
North, for Economic and Environmental Security] project be-
gan in 2013, funded by USAID and implemented by The northern plain corresponds to a filling of the collapsed north-
Development Alternatives Inc. (DAI). The principal aim of east slope of the northern massif by detrital materials. The geol-
the project is to promote and develop a sustainable approach ogy of the region is affected by the proximity of the northern edge
to agriculture within the region (DAI Global, LLC 2020). One of the Caribbean plate in collision-subduction with the North
of the objectives of the project is to increase irrigated area from American plate. The great northern fault (sinister strike-slip in
1,500 to 4,000 ha in a sustainable manner (Molnar et al. 2015). the direction of WNW–ESE) which affects the north of the
This expansion of irrigated land will assist in securing the pro- Dominican Republic passes offshore of the northern plain which
duction of the value chains of banana, rice, and cacao, which happens to be in a tectonic strip, limited in the south by the fault
are vital to the regional and national economy. To assist in the of la Española (Monthel et al. 2013). The nature of the deposited
sustainable management of water resources, a groundwater alluvium is very heterogeneous as evidenced by the high vari-
model of the shallow water in the Plaine du Nord was devel- ability of the available geological logs (Tomasi 1990a). Alluvium
oped to support the decision-making process. A thorough un- is made up of alternating sands, gravels and clays. Silty facies
derstanding of the water balance in the area, as well as the many over at least 6 m are found on the surface (Monthel et al. 2013).
different interactions between water sources and extractions The coarsest facies were found in the old beds of Grande Rivière
was necessary for the conceptual framework of the model. To du Nord and Trou du Nord. In contrast, clay facies seem to
develop a robust decision tool, it was necessary to search for predominate in Phaëton, Fort-Liberté, Terrier Rouge and
and compile available data, acquire relevant key additional da- Carrefour Oranger. The alluvium that make up the northern plain
ta, develop a conceptual and numerical model, calibrate the is thick, increasing from south to north, up to more than 300 m at
model in the first instance at a steady-state, and then develop the edge of the ocean in the central zone, a thickness estimated by
some prospective scenario simulations related to the extension geophysics (Radstake and Chery 1992; Miner et al, Northwater
of irrigated areas for banana crops. International (USA), ‘Geophysical reconnaissance of the western
The lack of functional climatic and water monitoring net- basin of the Plaine du Nord aquifer, Haiti’ (Hydrogeology
works (Adamson et al. 2016) was the most challenging hurdle Journal, this issue), 2022). Boreholes reaching the basement are
to overcome. A local team of researchers mobilised the vari- rare. Among the UNDP drilling campaign results in 1990, only
ous water stakeholders in the study area to make the best use two drillings reached the basement (Phaeton and Chabert, respec-
of historical data and current status of knowledge. Critical to tively at depths of 63 and 37 m). The local presence of the
the success of this study, and indeed to the furthering of basement at shallower depth reflects the existence of thresholds
hydrogeological knowledge in northern Haiti was the close between the Trou du Nord, Terrier Rouge and Fort-Liberté
collaboration between DAI, BRGM (French Geological basins, which could represent an obstacle to underground flows
Survey), and the Universities of Port-au-Prince and (Tomasi 1990b).
Limonade. Each milestone of the modelling process was sup- In the area of Cap-Haitien, the Balan municipal boreholes
ported by a training session in Port-au-Prince, bringing togeth- that serve Cap-Haitien intersected over a 100 m of alternations
er BRGM hydrogeologists, current graduate students, profes- of sand and clay which correlate quite poorly despite the prox-
sors, and DAI technicians and engineers. imity of the drillings (a few hundreds of meters) but in which
coarser layers (3–5 m thick) can be identified around the depths
of 10, 30, 45 m and sometimes at deeper levels but at varying
Materials and methods depths depending on the boreholes. The inadequacy of the po-
sition of the well screen and the filter beds led to the silting up of
Study area old boreholes and the construction of two new ones (F10 and
F11) in 2005 (BRGM-BRL-PC Consortium 2011). In 2020, the
The study area stretches across the North and North-East de- rehabilitation of two of the four boreholes and the drilling of two
partments of Haiti. The northern alluvial plain concerns five new ones in the same area was planned (Archambault and
Hydrogeology Journal (2022) 30:1399–1415 1401

Fig. 1 Location map of the study area (SRTM90m produced by NASA)

Biscan 2018; Adamson and Miner 2018). These sandy-gravel downstream in the flat areas. The reported piezometry is
levels are framed by silty and clay formations which confine shown on the hydrogeological map in Fig. 2 (MARNDR
them. Some of the deeper boreholes, located in the downstream and UNDP 1990).
part, are even artesian (La Rue, Carrefour Oranger, Quartier Hydrochemical characterization of groundwater carried in
Morin); however, the water body is never far from the topo- the Cap Haitien region (Gutierrez et al. 2012) confirms that the
graphic surface. Between Cap Haitien and La Grande Rivière alluvial plain is a multilayer aquifer, the most superficial level
du Nord, an inventory undertaken by BRGM in 2011 shows a of which is in oxidizing conditions (unconfined water table).
first sandy-silty level tapped by hundreds of shallow domestic On the other hand, the semi-deep hand-pump wells (15–20 m
wells, 4 m deep on average (Gutierrez 2011). deep in this sector) tap deeper levels that are confined and
Borehole production as documented by Tomasi (1990b) are protected by clay or silty layers. The semi-deep (15–20 m)
on the order of 35–70 m3/h, reaching higher values in the old and deep (up to 100 m deep) wells benefit from good natural
river beds (Trou du Nord, Grande Rivière du Nord). Pumping protection and denitrification under reducing conditions.
rates of the Balan boreholes in 2018 were on the order of 100– The chemical analyses and piezometric measurements car-
180 m3/h—F8: 100–140 m3/h, F9: 180 m3/h, F10: 112 m3/h, ried out in October 2010 by BRGM on the sector of the
F1: 126 m3/h (Northwater International, unpublished data, Grande Rivière du Nord that crosses the plain on a north–
2018), slightly lower than in 2010 (BRGM-BRL-PC 2011). south axis show without ambiguity that the river strongly
Borehole F8 is now abandoned due to a break in the casing drains the aquifer upstream of Pont Parois (Gutierrez et al.
(J. Adamson, Northwater International, personal communica- 2012). The piezometric map in Fig. 2 also shows the same
tion, 2020). Well testing performed on the Balan boreholes by flow trend for all other rivers of the study area. On the other
Inter-American Development Bank in 2018 recorded discharge hand, on the upstream end of the plain, the valley bottoms are
at 284 m3/h with 43 m of drawdown at F9, and 205 m3/h with topographically higher than the water table; the flow of water-
36.5 m of drawdown at F8 (Adamson and Miner 2018). courses would thus participate in recharging the groundwater
To the south of the plain, the Massif du Nord is made up of by infiltration (Tomasi 1990c).
intrusive volcanic formations (rhyolites, dacites, andesites, di-
orites, gabbros, etc.), rocks that are by nature not very perme- Development of the groundwater flow model
able (secondary porosity). Many springs are exploited from
these rocks, with average flow rates between 0.18 and 3.6 m3/ Computer code
h (Tomasi 1990c). Large-diameter wells are also dug in the
weathered shallow horizon. The productivity of boreholes The finite-difference based MARTHE software was used for
intercepting the underlying fissured horizon remains limited the model (Arnaud and Amraoui 2018; Kloppmann et al. 2012;
(maximum 3.6 m3/h). The hypothesis of a largely dominant Thiéry 1993, 2010a, b, c; Thiéry et al. 2018, Thiéry and Picot-
runoff on the northern slope of the massif appears realistic. Colbeaux 2020), thus enabling three-dimensional (3D) simu-
This runoff contributes to the recharge of the groundwater lations of groundwater flow, mass and energy transport in po-
which flows from the south to the north with an average gra- rous media. The software allows one to model hydrosystems
dient of 0.45%, with a stronger gradient upstream an\d near by integrating groundwater flow, hydro-climatic budgets and
the border of the massifs (up to 2%) and weaker gradient surface-water flows. The underlying schemes can be simple or
1402 Hydrogeology Journal (2022) 30:1399–1415

Fig. 2 Extract from the

hydrogeological map of Haiti at
the scale of 1:250,000
(MARNDR and UNDP 1990)

complex (unsaturated flow, multiphase flow, density- uses two storage coefficients: the storage coefficient in an
dependent fluid flow, consideration of vegetation, aquifer– unconfined state (equivalent to the effective porosity) when
river interactions, etc.). It is also possible to make a reactive the cell is not fully saturated with water, and the confined
transport calculation using a coupling with the PHREEQC storage coefficient when the cell is saturated with water.
code (Thiéry 2015). The MARTHE software includes a package that fully cou-
The software uses a finite-difference numerical scheme ples the groundwater flow calculation with the river network,
using irregular parallelepiped or rectangular stencils. This can as well as lakes and wetlands. Depending upon the relative
represent complex multi-layer aquifer systems, with pinchouts position of the water level in the river and of the aquifer’s
and the occurrence of free surfaces in any layer. MARTHE groundwater level, water flows can be simulated from the
addresses modelling of regional multi-layer systems covering river reach to the aquifer, or, inversely, from the aquifer to
several thousand square kilometres, as well as local-scale the river network. Such flow exchanges are controlled, in a
modelling of porous media systems down to a few cubic nonlinear manner, by linking riverbed and banks proportion-
centimetres, with millimetre-size cells. It is especially designed ally to their surface area, thickness and hydraulic conductivity.
to evaluate and manage groundwater resources. Finally, the Different configurations of exchanges between an aquifer and
MARTHE software presents all the functions necessary for a river can be simulated by MARTHE as briefly presented in
modelling of the alluvial groundwater of the North Plain. Fig. 3.
In the MARTHE code, there is no concept of “confined
layer”, “unconfined layer” or “convertible layer”. At any time, Grid design and geometry
any cell can become confined, unconfined or dewatered, de-
pending upon the current hydraulic head. The transmissivity is The model domain overlays a regular grid of 136 rows and
calculated from the hydraulic conductivity assigned to each 343 columns with 200-m2 cells. In total, the model is made up
mesh and the water height. In the transient mode, MARTHE 20,431 active cells, covering an area of 817 km2.
Hydrogeology Journal (2022) 30:1399–1415 1403

Fig. 3 Different configurations of

exchanges between an aquifer and
a river with MARTHE and related
Eqs. (1), (2) and (3) (see text)

Given the very high heterogeneity of the facies involved Hydraulic boundary conditions
(clay, sand, gravel) and the rapid lateral variations of facies
(lenticular geometries), it was impossible to produce a 3D The boundary conditions were defined according to the
geological model of the alluvium due to the lack of localised hydrogeological configuration of the Plaine du Nord as follows:
geological information. Therefore, after performing a sensitiv-
ity analysis, it was decided that the alluvial aquifer would be – The groundwater level was fixed to zero along the coast-
equally well represented by a single-layer geological model. line to represent the sea boundary condition (389 cells
This analysis consisted of a preliminary simulation of different concerned).
aquifer structures—one layer, two layers, and three layers – The groundwater level was fixed along the natural limits
with two aquifers and one aquitard. of alluvial deposits and limestone or along the limits of
To the north, the conceptual model is limited by the natural volcanic formations to represent the inflow from the latter
limits of the alluvial outcrops and the coastline, while towards (599 cells concerned). The imposed hydraulic heads are
the south, the model extends to the foot of the slopes of the first reconstructed from the linear relationship between
Massif du Nord, and towards the east, the model was extended the groundwater levels and the topography, then adjusted
over the administrative border with the Dominican Republic during the calibration process. The linear regression re-
in order to include the entire watershed of the Jassa and sults from observations made by DAI on 792 wells (pie-
Massacre rivers. zometric level = 0.919 × elevation – 1.615, R2 = 0.97).
A 5-m resolution digital elevation model derived from In the transient state, the imposed hydraulic heads are
LiDAR was used to represent the topographic surface; it was replaced by imposed flows.
provided by CNIGS (Centre National de l’Information Géo- – No flux (Neumann-type boundary conditions) was fixed in
Spatiale). Within the modelled area, elevations vary between the lateral boundaries to represent groundwater divide lines.
0.15 and 72 m, with a mean value of 19 m above sea level. The – For the five main rivers, the exchange between ground-
geophysical surveys carried out by BRGM in 1980 (588 ver- water and surface water was represented by river limits
tical electrical soundings), coupled with the drilling data, according to Eqs. (1), (2) and (3) (Fig. 3).
allowed the estimation of a depth of the alluvial bedrock
(Radstake and Chery 1992). The thickness of the alluvium
varies between 36 and 300 m, with a mean value of 141 m
ðH R −H b Þ
(Fig. 4). QPercol ¼ AREAExch  K R  ð1Þ
Thickn
1404 Hydrogeology Journal (2022) 30:1399–1415

Fig. 4 Map of estimated

thickness of the alluvial aquifer of
the Plaine du Nord

ðH R −H G Þ achieved with a water balance calculation, and applies a coef-

QExch ¼ AREAExch  K R  ð2Þ
Thickn ficient based on country-wide GIS layers that include soil and


H G −H Topo geologic permeability, vegetation cover, relief, and
QOverfl ¼ ðAREAcell −L  W Þ  K R  ð3Þ karstification.
Thickn
Based on the work of Adamson et al. (2016), the modelled
where the terms are defined as follows: domain was finally divided into five homogeneous recharge
L Length of the river reach zones (Fig. 5), in which each mesh cell receives an identical
W Width of the river reach recharge flow. The average value, calculated from the grid at
KR Hydraulic conductivity of the river bed and river 30-m intervals, was initially assigned to each of the zones, i.e.
banks 100 mm/year for zones 1 and 2, 170 mm/year for zone 3,
Thickn Thickness of the river bed (and the river banks) 200 mm/year for zone 4, and 250 mm/year for zone 5.
HR Absolute elevation of the free water surface in These values were subsequently adjusted during the calibra-
the river tion of the model.
HG Hydraulic head in the underlying groundwater
Hb Elevation of the bottom of the river bed Pumping wells
HTopo Elevation of the topographic surface of the cell
containing the reach In terms of current use of the wells for groundwater abstrac-
AREACell Horizontal area of the cell in which the reach is tion (Fig. 6), groundwater uses were grouped in three catego-
located ries: drinking water supply, irrigation use, and industrial uses
AREAExch Area of the groundwater-river exchange (such as Caracol industrial park, Larue distillery, ice factories,
=L x [W + 2 x (HR − Hb)] water purifiers, etc.). Drinking water is supplied through a few
QExch Flow of the river exchange towards groundwater deep boreholes and a large number of semi-deep boreholes
QPercol Flow of the river percolation towards (hand pumps). For the deep boreholes supplying the munici-
disconnected groundwater palities of Cap Haitien and Ouanaminthe, the abstracted vol-
QOverfl Complementary overflow of groundwater to the umes were collected from DINEPA (DIrection Nationale de
river l’Eau Potable et de l’Assainissement).
For hand-pump wells, the estimation of the quantities re-
quired one to cross check three main sources of information to
Groundwater recharge arrive at a reliable average estimation of pumped water volumes.
These three sources are: the current distribution of population by
Due to the lack of recent and homogenous climatic data, it was municipality, the average water consumption per capita, and the
not possible to include the calculation of hydro-climatic bal- pumped water volumes extracted from the wells identified under
ances within the model to establish the groundwater recharge the ‘Haiti Outreach’ database (Haiti Outreach 2018).
potential within the study area. It was decided to rely on the For the average water consumption per capita, a distinction
national recharge map established by Miner and Adamson was made between rural areas (20 L/day per inhabitant) and
(2017). This geographic information system (GIS) model is urban areas (70 L/day per inhabitant); these estimations were
Hydrogeology Journal (2022) 30:1399–1415 1405

Fig. 5 Homogeneous recharge

zones defined in the Plaine du
Nord model and initial values for
the steady sate conditions

extracted from a DINEPA technical leaflet (DINEPA 2013). abstraction points, related to the production of sachet drinking
Each well listed as functional in the Haiti Outreach database water. Once again, the information proved difficult to collect/
was assigned an average annual withdrawal volume—an ex- access and cross check. The estimation for the Caracol indus-
ample of the calculation is provided for the municipality of trial park is notably underestimated.
Fort-Liberté in Table 1. In total, 2109 abstraction wells were identified thanks to the
The same approach to estimate the withdrawals from irri- Haiti Outreach database and introduced in the modelled area.
gation wells proved to be difficult due to the lack of regular These wells represent an average annual withdrawal estimated
monitoring and reliable data. The decision was made to ignore at 9.5 Mm3/year, including 9.3 Mm3/year dedicated to drink-
their influence for the time being, with the exception of two ing water supply.
working wells listed in a BRGM study performed in 2010
(Gutierrez et al. 2012). These withdrawals for irrigation are Hydrodynamic parameters
unknown but represent small volumes (family farms, very
shallow wells). The aquifer of the Plaine du Nord has essentially been inves-
For the estimation of water extraction for industrial uses, tigated during the United Nations project at the end of the
the survey carried out by the local team identified only eight 1980s (Tomasi 1990b). Pumping tests performed allowed an

Fig. 6 Map of average annual groundwater withdrawals in the Plaine du Nord

1406 Hydrogeology Journal (2022) 30:1399–1415

Table 1 Method used to calculate

average annual withdrawals for Parameter Rural areas Urban areas
drinking water supply—example
of Fort-Liberté Average water consumption per capita (L/day per inhabitant) 20 70
Number of inhabitants 7,505 26,929
Number of wells 145 8
Total daily water consumption (L/day) 150,100 1,885,030
Operating rate of a well (L/day) 1,035 235,629
Operating rate of a well (m3/year) 377.8 86,004.5

overall assessment of the hydraulic transmissivity in different resistivity revealed by geophysical investigations performed
points of the aquifer. Most of tests concern deep wells (depth in 1990 (Radstake and Chery 1992). The areas identified as
> 40 m). The hydraulic transmissivity has been estimated to being more resistive correspond to coarser, and therefore more
range between 5.2 × 10−4 and 2.2 × 10−2 m2/s, with an permeable, alluvium, something that has been occasionally
average of 6.6 × 10−3 m2/s. For this study however, no infor- confirmed by various reconnaissance drilling endeavors.
mation was collected about the storage coefficient for the During the calibration process, the delimitation of these ho-
shallow aquifer in general. mogeneous permeability zones was adjusted in order to best
reproduce the observed piezometric levels. The aquifer is con-
Hydrographic network sidered isotropic. This study focused on uniform value ranges
rather than detailed reproduction of field data leading to a
The five main rivers crossing the alluvial plain—Haut du Cap, contrasting mosaic of values, because regional models can
Grande Rivière du Nord, Trou du Nord, Marion and Jassa— not consider heterogeneities observed at small scale. The dis-
were modelled explicitly with the software. For each tributary, tribution of hydraulic conductivity finally adjusted is shown in
the river data are defined by their section within the cell. The Fig. 7.
hydrological priorities are defined by a method of branching For the hydraulic heads on the fixed-head boundary, an
tree that arranges the tributaries’ connectivity and forms the adjustment on different segments was necessary in order to
modelled hydrographic network. A river section is described respect the consistency of the inflows into the alluvial plain
in the model by its geometry (length, width, and elevation of and the piezometric levels observed on the upstream edge of
the riverbed), by its physical properties (thickness and perme- the plain.
ability of the bed and the banks), by its source terms (pumping Regarding groundwater recharge, the zones and average
or injections) and by its state (height of water in the river). The annual values proposed by Adamson et al. (2016) were
geometric characteristics are taken from the field surveys car- retained (see Fig. 5), with the exception of the central zone,
ried out by DAI water technicians. where the value was lowered from 200 to 180 mm/year.
In addition to the river network within the study, an esti- The field observations available and used to constrain the
mation of the upstream inflows at the boundaries of the steady-state calibration process are—the piezometric map de-
modelled area was required for each main river. However, veloped by Tomasi (1990c) and revised within the framework
there are currently no flow-monitoring gauging stations up- of this project (Fig. 2), the piezometric levels noted by DAI
stream of the flat Plaine du Nord. The upstream river inflows water technicians and BRGM (506 measurement points), the
were therefore estimated by calculating the effective precipi- annual average flows at four historic stations, the gauge mea-
tation in the watershed of each upstream mesh using the surements carried out for the AVANSE project (DAI Global,
Thornthwaite method. LLC 2020) and maps of the wetlands and flood zones (Fig. 8).
The initial hydraulic heads are only given on the fixed head
Steady-state model calibration boundary. For the rest of the domain, the initial hydraulic
heads are fixed to 0 msl.
The three main parameters adjusted for the wedging were—
the permeability, and the hydraulic heads on the fixed-head Simulation scenarios
boundary and the layer of water infiltrated on the plain to a
lesser extent. For each parameter, a trial-and-error approach As previously noted, the lack of monitoring data (i.e. piezo-
was used. metric levels, river flows, withdrawals, rainfall and evapo-
A homogeneous initial permeability of 1e-5 m/s was cho- transpiration (ET)) represents a real challenge, and therefore
sen on the basis of the available data. The adjustment of per- does not allow for the calibration of the model for a transient
meabilities was then based on areas of high electrical state. However, the AVANSE project required an
Hydrogeology Journal (2022) 30:1399–1415 1407

Fig. 7 Spatial distribution of the

calibrated hydraulic conductivity

understanding of the seasonal fluctuations of the shallow wa- Assumptions for the theoretical transient model (scenario 1)
ter table, which are observed to be fundamental in the contin-
uous supply of water to the many hand pumps that are used on The results of the steady-state calibration for the piezometry
a daily basis. To this purpose, a theoretical transient-state and the flow rates in the five rivers were used as initial con-
model has been developed based on the following approach: ditions for the transient regime. The temporal variability of the
meteoric recharge at the level of the plain is taken into ac-
1. Development of a “theoretical” transient regime over the count, based on a calculation of effective rainfall at monthly
period 2000–2016, at monthly time steps, which repre- time steps over the period 2000–2016. The calculation was
sents the seasonal and annual fluctuations (scenario 1) performed based on monthly cumulative rainfall records for
2. Simulation of a theoretical scenario of irrigation sized by Cap Haitien and Terrier Rouge (source: USAID/MARNDR),
the AVANSE project agronomists, mainly based on the while the average monthly potential evaporation (PET) values
water needs of the banana crops according to the FAO were calculated using the FAO records. A correction coeffi-
CROPWAT irrigation demands 9 (scenario 2) cient was applied to get to the theoretical average monthly
3. Comparison of the irrigation scenario 2 with the reference recharge for the period 2000–2016. The correction coefficient
scenario 1 for the same period 2000–2016 was applied according to the following formula:

Fig. 8 Head and hydrometric data available for the steady-state model calibration
1408 Hydrogeology Journal (2022) 30:1399–1415

throughout the year. Finally, in the transient mode, a storage

RðiÞ ¼ PEFFðiÞ  PEFFavgð jÞ  Ravgð jÞ ð4Þ
coefficient value must be assigned to the model cells. An
average value of 2% was chosen for the unconfined state.
where the terms are defined as follows:
PEFF(i) Effective rainfall computed for month i with i =
Theoretical irrigation scenario (scenario 2)
1–204 (i.e. January 2000 to December 2016)
PEFFavg(j) Mean monthly effective rainfall calculated over
Eleven areas targeted for potential future development of ba-
2000–2016 (j = 1–12)
nana cultivation were studied. The water requirements were
Ravg(j) Mean monthly recharge used in the transient
estimated from FAO guides (CROPWAT 8.0 software). In
model (j = 1–12).
Haiti, the irrigation period lasts on average from June to
October (Table 2).
A recharge chronicle at monthly time steps (2000–2016) is
In accordance with AVANSE agronomists, the assumption
thus assigned to each of the five recharging zones of the mod-
was made that irrigation would be satisfied by setting up bore-
el. (Fig. 9). This seasonal recharge fluctuation was verified
holes, such that one borehole structure would satisfy the irri-
with some records for the same period collected at the
gation requirements for a plot of 10 ha. In the identified areas
AVANSE farms. The meteoric recharge introduced in the
for banana crop, an additional total of 270 boreholes were thus
model shows a large deficit during the drought period ob-
implemented in the modelled area to meet the water needs of
served between 2014 and 2016, with even zero recharge in
the 11 zones identified by AVANSE (Fig. 10).
some places.
Based on these assumptions, the unit production rate of an
Concerning the inflows coming into the model boundaries,
irrigation borehole varies between 2.2 m3/h (June) and
the lack of climatic data on the Massif du Nord for the period
13.6 m3/h (July). All uses combined, withdrawals are thus
2000–2016 was overcome by adopting an even more sche-
brought to a maximum of around 3.5 Mm3 in July, against
matic approach. A monthly distribution, identical for each
0.79 Mm3 for the reference model.
year, was defined from the calculation of the effective month-
In the scenario, agricultural boreholes were commissioned
ly average precipitation based on old rainfall and PET data
from 2007 onwards. In order to measure the impact of this
(MARNDR), representative of the Massif du Nord. For each
pumping scenario on the water resource of the alluvial aquifer,
cell concerned, the calculated inflow in steady state was
the reference simulation (scenario 1 without agricultural ab-
assigned a distribution coefficient.
straction) was compared with the prospective simulation (sce-
For the transient simulations, in absence of measurements,
nario 2 with agricultural pumping) for the period 2007–2016,
the hydraulic heads that have been fixed at the upstream limits
i.e. over a 10-year regular irrigation period.
are therefore replaced by fixed flows. This approach ensures
that the inflows simulated by the modelling software do not
increase unrealistically in the event that a significant draw-
down affects the modelled area (under the effect of pumping Results and discussion
near the limits, for example). All other boundary conditions
remain unchanged (with loads imposed on the coast, lateral Steady-state calibration
limits with zero flow, meshes with overflow over the entire
domain). Additionally, groundwater withdrawals concerning The simulated steady-state piezometric map is shown in Fig.
water supply and industrial uses are considered constant 11. The piezometric contours from the Haitian

Fig. 9 Evolution of annual

recharge in the transient model
(2000–2016)
Hydrogeology Journal (2022) 30:1399–1415 1409

Table 2 Water requirements of

banana crops (FAO) Month Water requirement (mm/month) Pumping rate for irrigated area of 10 ha (m3/h)

June 15.9 2.2

July 100.9 13.6
August 85.6 11.5
September 59.1 8.2
October 28.8 3.9

hydrogeological map (MARNDR and UNPD 1990) are also From a scientifical perspective, the calibration of the steady
shown for comparison. The results obtained from the river flow does not make sense; however, this was performed
modelled piezometric levels are satisfactory with well- to check the consistency of the simulated surface flows with
simulated flow directions and hydraulic gradients. The scatter the measured values. The only available historical hydromet-
diagram in Fig. 12 shows how close the simulated contours ric data span the period 1920–1940 for the main rivers, while
area aligned with the field measurements and the absence of more recent gauging data were collected through the
an underestimation or a systematic overestimation of the mod- AVANSE during 2017 (five locations with weekly measure-
el with deviations of less than 6 cm. The results for the simu- ments from January to July) from which an average measured
lation of the watersheds of Haut-du-Cap, Grande Rivière du flow was calculated (Table 3).
Nord and Trou du Nord are particularly good and match the Even if deviations exist for most of the measurement
measured piezometer levels of the 1990 hydrogeology map points, the orders of magnitude of the simulated flows are
well. The isocontour lines between 2 and 20 msl are almost acceptable (Table 3). Deviations are inevitable given different
superimposed. Towards the east, the simulation is less accu- measurement periods and uncertainties linked to the gauging
rate around Terrier Rouge and for the Marion and Jassa wa- operations on the one hand and the model on the other hand.
tersheds. The comparison of simulated and observed piezom- The absence of proper description of the hydrographic net-
eter levels in these areas is in any case less accurate since there work does affect the calculation of groundwater and river
were fewer field measurements undertaken in these areas, and exchanges and introduces uncertainties in the model results.
therefore the accuracy of the 1990 potentiometric map itself is The most important river parameters to survey are the stream
less certain. Only additional field measurements in those areas morphology and the geometry of the riverbed. A topographic
could help improve the calibration of the model in these survey campaign (cross sections) of the main five rivers was
basins. launched by the end of 2017 by the AVANSE team; however,
Although the model reproduces piezometer levels well, at the time of the assignment, the data were not ready. The
the scarcity of field observations on the upstream end of newly acquired data could be integrated into the model at a
the plain could be improved by undertaking additional later stage.
baseline data and by consolidating the steady-state piezo- The simulation of surface flows can only be consolidated
metric calibration. from a calibration of the model in the transient regime.

Fig. 10 Locations of the target

areas and boreholes implemented
for the theoretical irrigation
scenario
1410 Hydrogeology Journal (2022) 30:1399–1415

Fig. 11 Piezometric map calculated in steady-state conditions vs piezometric map produced by Tomasi 1990c

Therefore, the modelled exchange flows between the five locally exclude operational difficulties, in particular a risk of
main rivers and the alluvial groundwater should be considered saline intrusion. Indeed, the question of readily available wa-
with caution. ter volumes cannot be considered independently from spatial
(location of withdrawals) and temporal (high water/low water)
considerations. Most parts of the five modelled rivers are in-
Water balance deed draining groundwater (rivers are fed by the groundwa-
ter). The exchange flows from shallow groundwater to rivers
The diagram of Fig. 13 represents the main components of the represent 85% of the exchanged flows (Fig. 13).
hydrological balance of the alluvial aquifer at the scale of the The water contributions from areas outside of the modelled
model. The share of withdrawals remains moderate overall domain, mainly by water runoff on the Massif du Nord, are
compared to the flows contributing to the groundwater supply always difficult to quantify in the field. However, in order to
(infiltration + contributions from the Massif du Nord). At the validate the model’s calculations with regard to inflows from
level of the plain, it can therefore be concluded that the avail- upstream, the continuous acquisition of hydro-climatic data is
ability of groundwater resources is still significant. As will be essential, at least by establishing monitoring of—(1) the flows
shown later, this overall result at the level of the plain does not of the main rivers at their outlet into the plain, (2) piezometric
fluctuations upstream of the alluvial plain; (3) rainfall over the
massif. In summary of the overall water balance in the study
area, a better understanding of withdrawals, both in aquifers
and rivers, is essential. Some recent borehole measurements
could not be included in the inventory (Caracol industrial
park, large farms, etc.).

Results of simulation scenarios

Reference scenario

In a transient regime, the model allows one to simulate and

understand the dynamics of groundwater in relation to the
variable infiltration of precipitation on the plain and inputs
from the Massif du Nord. Annual cyclicality predominates
and recharge periods are short, spanning mostly 1–3 months,
Fig. 12 Simulated heads in steady-state conditions vs measured heads and generally occur between October and December, which is
Hydrogeology Journal (2022) 30:1399–1415 1411

Table 3 Simulated river flows in

steady-state conditions vs mea- Gauging station Monitoring period Observed flow (m3/s) Simulated flow (m3/s)
sured flow in 2017 (GRN: Grande (mean values) (steady-state model)
Rivière du Nord, HDC: Haut du
Cap river, TDN: Trou du Nord Historical river stations
river) GRN Parois 1920–1940 8.18 9.24
HDC Oranger 0.26 0.25
HDC Grison 0.82 0.36
TDN Chabert 0.98 0.57
Project gauges
HDC downstream January–July 2017 0.64 1.53
GRN downstream 4.21 9.2
TDN downstream 0.41 0.57
Marion downstream 0.44 2.13
Jassa downstream 0.31 1.49
TDN upstream 0.32 0.31
Marion upstream 0.53 1.91

then followed by a period of groundwater drainage that ex- Irrigation scenario

tends from January to September (Fig. 14).
At the beginning of 2015, the two successive dry years The impacts of agricultural pumping on groundwater levels
impact the water-resource balance, with the alluvial aquifer differ from one study area to another. The drawdowns appear
almost continuously draining (destorage) water out of the to be better controlled around the Cap Haitien and Limonade
aquifer, which results in noticeable absence of the usual rise sectors, with simulated maximum drawdowns of around 1 and
in piezometric levels (Fig. 14). 0.7 m, respectively (Fig. 16). In contrast, the piezometric
A further example is presented in Fig. 15, which presents drawdown is greater in the areas of Caracol, Terrier Rouge
the simulated piezometric map for the month of June 2016, and Ouanaminthe, with drawdowns that can exceed 4 m and
1 year in the drought period. It is interesting to note that, in a reach up to 6 m in Caracol (Fig. 16).
context of a recharge deficit, the piezometric depression of It is interesting to also note that the simulated piezometric
Cap Haitien is spreading, in particular at the mouth of the levels rise again as soon as the irrigation is stopped, to quickly
river, where simulated groundwater levels are below sea level. get back to the levels of the reference simulation (Fig. 16).
The model clearly shows that under very low water condi- This remains true during periods of pronounced low water
tions, as observed in 2016, the risk of saline intrusion is inev- such as during the drought of 2014–2016, with the exception
itably increased in the areas where the groundwater is of certain sectors located at the upstream end of the plain
unsustainably exploited. (Terrier Rouge). For the latter, the water inflow from Massif

Fig. 13 Water balance in steady-

state conditions (negative values
represent water leaving the aqui-
fer, whereas positive values rep-
resent water entering the aquifer).
MCM million cubic meters
1412 Hydrogeology Journal (2022) 30:1399–1415

Fig. 14 Transient groundwater dynamics in the Plaine du Nord during the period 2000 to 2016 with a variation of cumulative aquifer storage simulated
and b variation of simulated heads in Grande Rivière du Nord

du Nord does not appear to compensate for the additional to rely on shallow wells to meet crop water requirements.
agricultural pumping under low recharge conditions. According to the model, the most sensitive areas are
Finally, from an operational point of view, the two main around Caracol, Terrier Rouge and Ouanaminthe. The
operating constraints highlighted by the simulation of such a cumulative impact of intensive agricultural and industrial
pumping scenario are as follows: drawdown, combined with the potential effect of climate
change, could potentially lead to unsustainable use of
1. Along the coast, localised operating difficulties could be groundwater that would not allow for enough time for
expected due to a significant risk of seawater intrusion. recharge and a return to a safe water balance status. An
Indeed, under the effect of agricultural pumping, integrated use of water resources is therefore
upconing phenomena and/or reversal of groundwater recommended.
flow direction cannot be excluded. In particular, the areas
of Cap Haitien and Caracol that have simulated piezomet-
ric levels close to 0 msl, would be immediately impacted
by seawater intrusion when the water table drops below Conclusions
sea level. This conclusion is based on the simulation of
drawdown by the model. To develop this point, it might The Plaine du Nord aquifer represents a major water resource
be interesting in the future to introduce into the model for Haiti. Agricultural development projects combined with
variable-density groundwater flow (possibility provided urban development in the region require an integrated ap-
by the MARTHE software). proach to the management of the alluvial aquifer and the river
2. The simulated drawdowns amount to several meters in network that drains it. A stable hydrodynamic model was
most cases (without taking into account the quadratic developed based on the scarce data available and new data
head loss). It is therefore likely that it will not be possible acquired during the AVANSE project. Due to the lack of

Fig. 15 Piezometric map

calculated in transient-state con-
ditions (June 2016)
Hydrogeology Journal (2022) 30:1399–1415 1413

Fig. 16 Simulated drawdowns for the irrigation scenario (August 2016), with a variation of simulated heads in Cap Haitien area (for a 1-m diameter
well), b variation of simulated heads in Caracol area (for a 1-m diameter well), and c simulated drawdown map

monitoring of the groundwater, watercourses or even climate, can be easily mobilized at the level of the plain, if care-
the calibration of the model could only be carried out under fully managed to rule out localised operational
steady-state hydraulic conditions. The results, however, are difficulties.
promising and the work presented in this report is only the The potential for groundwater withdrawals is more limited
first stage of a much wider work package that could be for the Haut-du-Cap watershed due to an already higher pres-
complemented by the calibration of the transient state that sure for withdrawal to supply drinking water to Cap Haitien,
would allow one to better take into account the dynamics of and at the Marion River watershed, due to a lower recharge
storage/destorage of the aquifer in the various locations in the potential, whether by precipitation or by water flows from the
Plaine du Nord. Additional survey work should be considered, Massif du Nord. The simulated transient scenarios show that
and the new data acquired can easily be introduced into the the alluvial aquifer conditions allow for the commissioning of
model. agricultural irrigation in the identified areas by the AVANSE
The steady-state model has, however, already shed project. A rapid replenishment of the water resource was sim-
new light on the groundwater resources of the alluvial ulated after the irrigation campaign was stopped. This was still
plain. The development of the hydraulic balance of the the case in times of drought, with the exception being made of
groundwater is showing that significant water resources the areas upstream of the Plaine du Nord.
1414 Hydrogeology Journal (2022) 30:1399–1415

The simulated groundwater drawdowns at the additional Hygiene dans les Zones urbaines, periurbaines et rurales de la
Région Nord d’Haïti (ha-l1135). Rapport final [Environmental and
pumping wells are pluri-metric, reaching 6–7 m. Given that
social analysis (AES) of the Water, Sanitation and Hygiene Project
the current wells are very shallow (3–6 m depth on average), in Cap-Haitien, North Department, Haiti. Water, Sanitation and
such drawdowns would result in the current wells becoming Hygiene Program in Urban, Peri-urban and Rural Areas of the
dry. The siting of deeper boreholes (>20 m) should be consid- Northern Region of Haiti (HA-l1135)]. Final report]. https://www.
dinepa.gouv.ht/wp-content/uploads/2018/ORE_Nord/AES%
ered, especially in a context of deficit recharging such as the
20Rural.pdf. Accessed Feb 2022
one observed during 2014 and 2016 to allow for a sustainable Arnaud L, Amraoui N (2018) Vers une optimisation de l’exploitation des
balance between extraction and recharge. Finally, agricultural eaux souterraines du bassin versant crayeux de l’Avre (Somme,
areas that are close to the town of Cap Haitien and Caracol are France), 22èmes journées techniques du Comité Français
d’Hydrogéologie de l’Association Internationale des
particularly vulnerable to saline-water intrusion from the
Hydrogéologues. « Hydrogéologie de la craie » Le Havre, mai
ocean. 2018 [Towards optimisation of groundwater exploitation in the
chalky watershed of the Avre (Somme, France), 22nd workshop of
the French Committee of Hydrogeology from the International
Acknowledgements The authors are grateful to DAI for supporting this Association of Hydrogeologists]. https://hal-brgm.archives-
study, as well as for the modelling and for providing the necessary tech- ouvertes.fr/hal-02614072. Accessed Feb 2022
nicians and overall logistics to undertake the necessary monitoring, BRGM-BRL-PC Consortium (2011) Etude de faisabilité pour la
checks and surveys. The authors also thank the students of Limonade réhabilitation, l’extension du système AEP, la réalisation du
and Port-au-Prince universities and their professors who participated in système d’assainissement et la supervision de travaux de la ville
the field operations as well as the development of the model. Open access de Cap Haïtien. Inventaire des points d’eau [Feasibility study for
availability of this article was made possible through support from the rehabilitation, extension of the AEP system, construction of the
HydroLOGICA. sanitation system and supervision of works in the city of Cap
Haitien. Inventory of water points]. Prêt 1010/SF-HA, AOI no.
Funding This study was part of the AVANSE Program, funded by the 26–02/10, BRGM, Orléans, France
U.S. Agency for International Development (USAID) and implemented DAI Global, LLC (2020) AVANSE final report. Appui à la Valorisation
by Development Alternatives Inc. (DAI) in collaboration with the Haitian du potentiel Agricole du Nord, pour la Sécurité économique et
Ministry of Agriculture. Environnementale (AVANSE) [Support to the valuation of agricul-
tural potential, for economic and environmental security, April
2013–January 2020]. DAI Global, Bethesda, MD
Declarations DINEPA (2013) Fascicule technique: conception des réseaux
d’adduction et des réseaux de distribution d’eau potable [Technical
Conflict of interest On behalf of all authors, the corresponding author leaflet: conception of drinking water supply networks]. https://
states that there is no conflict of interest. dinepa.gouv.ht/lereferentieltechnique/geniecivilethydraulique.html.
Accessed Feb 2022
Open Access This article is licensed under a Creative Commons Gutierrez A (2011) Aquifère de la Plaine du Nord. Inventaire des points
Attribution 4.0 International License, which permits use, sharing, adap- d’eau. Etude de faisabilité pour la réhabilitation, l’extension du
tation, distribution and reproduction in any medium or format, as long as système AEP, la réalisation du système d’assainissement des rejets
you give appropriate credit to the original author(s) and the source, pro- liquides et la supervision de travaux de la ville de Cap Haïtien [Map
vide a link to the Creative Commons licence, and indicate if changes were of the North aquifer: inventory of water points—feasibility study for
made. The images or other third party material in this article are included the rehabilitation, extension of the DWS system, construction of the
in the article's Creative Commons licence, unless indicated otherwise in a liquid waste sanitation system and supervision of works in the city of
credit line to the material. If material is not included in the article's Cap Haitien]. https://doi.org/10.13140/RG.2.2.22450.84164/1.
Creative Commons licence and your intended use is not permitted by Accessed Feb 2022
statutory regulation or exceeds the permitted use, you will need to obtain Gutierrez A, Gourcy L, Braibant G (2012) Etude de faisabilité pour la
permission directly from the copyright holder. To view a copy of this réhabilitation et l’extension du système AEP, la réalisation du
licence, visit http://creativecommons.org/licenses/by/4.0/. système d’assainissement des rejets liquides et la supervision des
travaux de la ville de Cap Haïtien: caractérisation géochimique
[Feasibility study for the rehabilitation and extension of the DWS
system, the realization of the liquid waste sanitation system and the
supervision of the works of the city of Cap Haitien: geochemical
characterization]. BRGM-BRL-PC. PRET 1010/SF-HA AOI no.
References 26–02/10. https://doi.org/10.13140/RG.2.2.19986.68801
Haiti Outreach (2018) https://www.haitioutreach.org. Accessed Jan 2018
Adamson JK, Miner WJ (2018) Well inspection and testing report, Hall TC, Sealy AM, Stephenson TS, Kusunoki S, Taylor MA, Chen A,
existing wells at Balan, Cap-Haitien, Haiti. Preliminary report, Kitoh A (2013) Future climate of the Caribbean from a super-high-
Inter-American Development Bank, Washington, DC resolution atmospheric general circulation model. Theor Appl
Adamson JK, Jean-Baptiste G, Miner WJ (2016) Summary of groundwa- Climatol 113:271–287. https://doi.org/10.1007/s00704-012-0779-7
ter resources in Haiti. In: Wessel GR, Greenberg JK (eds) IHSI, Institut Haïtien de Statistiques et d'Informatique (2015) Population
Geoscience for the public good and global development: toward a totale de 18 ans et plus. Ménages et densités estimés en 2015 [Total
sustainable future. Geol Soc Am Spec Pap 520. https://doi.org/10. population over 18 years old. Household and density estimations].
1130/2016.2520(14) Direction des statistiques démographiques et sociales (DSDS). Mars
Archambault A, Biscan B (2018) Analyse environnementale et sociale 2015
(AES) du Projet Eau, Assainissement et Hygiene à Cap-Haitien, IPCC (2007) Climate Change 2007: The physical science basis.
département du Nord, Haïti. Programme Eau, Assainissement et Contribution of Working Group I to the Fourth AssessmentReport
Hydrogeology Journal (2022) 30:1399–1415 1415

of the Intergovernmental Panel on Climate Change [Solomon S, Qin Thiéry D (2010c) Interaction between surface and subsurface flows:
D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller Somme Basin. In: Tanguy JM (ed) Practical applications in engi-
HL (eds)]. Cambridge University Press, Cambridge, United neering, vol 4, chap 13. Environmental Hydraulics Series. Wiley/
Kingdom and New York, NY, USA, p 996 ISTE, London, pp 143-156
ISARM, UNESCO (2009) Atlas of transboundary aquifers. Global maps, Thiéry D (2015) Modélisation 3D du Transport Réactif avec le code de
regional cooperation and local inventories. UNESCO, Paris calcul MARTHE v7.5 couplé aux modules géochimiques de
Kloppmann W, Aharoni A, Chikurel H, Dillon P, Gaus I, Guttman J, PHREEQC [3D modeling of reactive transport with the MARTHE
Kaitzer T, Kremer S, Masciopinto C, Miotlinski K (2012) Use of v7.5 computer code coupled with the geochemical modules of
groundwater models for prediction and optimization of the behav- PHREEQC]. Report BRGM/RP-65010-FR, 172 pp. http://
iour of MAR sites. In: Kazner C, Wintgens T, Dillon P (eds) Water infoterre.brgm.fr/rapports/RP-65010-FR.pdf. Accessed Feb 2022
reclamation technologies for safe managed aquifer recharge. Thiéry D, Picot-Colbeaux G (2020) Guidelines for MARTHE v7.8 com-
International Water Association Publications, London, UK puter code for hydro-systems modelling (English version). Report
Le Cozannet G, Idier D, de Michele M, Legendre Y, Moisan M, Pedreros BRGM/RP-69660-FR, BRGM, Orléans, France, 246 pp
R, Thiéblemont R, Spada G, Raucoules D, de la Torre Y (2020) Thiéry D, Amraoui N, Noyer ML (2018) Modelling flow and heat trans-
Timescales of emergence of chronic nuisance flooding in the major fer through unsaturated chalk: validation with experimental data
economic centre of Guadeloupe. Nat Hazards Earth Syst Sci from the ground surface to the aquifer. J Hydrol 556:660–673.
Discuss. https://doi.org/10.5194/nhess-2020-178 https://doi.org/10.1016/j.jhydrol.2017.11.041
Margat J (1990) Développement et Gestion des ressources en eau d’Haïti. Tomasi B (1990a) Inventaire des points d’eau de la région Nord
Rapport de mission de consultant du 9 au 25 juillet 1990 [Inventory of water points in the North region). Programme des
[Development and Management of Haiti’s water resources. Nations Unies pour le Développement (PNUD)/Ministère de
Consultant mission report from July 9 to 25, 1990]. Nations l’Agriculture, des ressources naturelles et du développement
Unies/DTCD-PNUD Projet HAI/86/003, UN, New York rural.Report HAI/86/003 no. 31, Service Nationale des Ressources
Martin A, Moussu H (1977) Les Ressources en eau Souterraines de la en Eau, Port-au-Prince, Haiti
République d’Haiti: etat des connaissances [Groundwater resources Tomasi B (1990b) Forages de reconnaissance et pompages d’essai dans la
of the Republic of Haiti: state of knowledge]. Report, BRGM 77-
Plaine du Nord [Reconnaissance drilling and test pumping in the
AGE-027, BRGM, Orléans, France
Plaine du Nord]. Programme des Nations Unies pour le
Miner WJ, Adamson JK (2017) Modeling the spatial distribution of
Développement (PNUD)/Ministère de l’Agriculture, des ressources
groundwater recharge in Haiti using a GIS approach. Geological
naturelles et du développement rural. Rapport HAI/86/003 no. 32,
Society of America, Annual Meeting 2017, Seattle, WA. https://
Service Nationale des Ressources en Eau, Port-au-Prince, Haiti
doi.org/10.1130/abs/2017AM-297120
Tomasi B (1990c) Ressources et besoins en eau d’Haïti. Volume 1:
Molnar JJ, Kokoye S, Jolly C, Shannon DA, Huluka G (2015)
Région Nord [Haiti’s water resources and needs, vol 1: North re-
Agricultural development in northern Haiti: mechanisms and means
gion]. Programme des Nations Unies pour le Développement
for moving key crops forward in a changing climate. J Agric
(PNUD)/Ministère de l’Agriculture, des ressources naturelles et du
Environ Sci 4(2):25–41. https://doi.org/10.15640/jaes.v4n2a4
développement rural. Rapport HAI/86/003 no. 33, Service
Monthel J, Bialkowski A, Jean Ph, Jean-Baptiste M (2013) Cartographie
Nationale des Ressources en Eau, Port-au-Prince, Haiti
géologique des villes de Cap-Haïtien, Fort-Liberté et Ouanaminthe
(République d’Haïti) [Geological mapping of the cities of Cap- MARNDR and UNDP [Ministère de l'Agriculture, des Ressources
Haïtien, Fort-Liberté and Ouanaminthe (Republic of Haiti)]. Final Naturelles et du Développement Rural and United Nations
report, BRGM/RP-61951-FR, BRGM, Orléans, France, 55 pp Development Program] (1990) Carte Hydrogéologique République
Radstake F, Chery Y (1992) geophysical prospecting for groundwater d’Haiti [Hydrogeologic map of the Republic of Haiti]. Projet HAI/
exploration in Haiti. Hydrol Sci J 37(1):1–12. https://doi.org/10. 86/003. United Nations Development Program, New York, Scale 1:
1080/02626669209492557 250,000, 1 sheet
Thiéry D (1993) Modélisation des aquifères complexes: prise en compte UNDP [United Nations Development Program] (1991) Développement
de la zone non saturée et de la salinité—calcul des intervalles de et gestion des ressources en eau: disponibilité en eau et Adéquation
confiance [Modeling of complex aquifers: taking into account the aux besoins, rapport technique, volume II, Région Nord [Water
unsaturated zone and salinity—calculation of confidence intervals]. resources development and management: availability of water and
Rev Hydrogéol 4:325–336 adequacy to water needs, technical report, vol II, North region).
Thiéry D (2010a) Hydrogeologic models. In: Tanguy J-M (ed) Projet HAI/86/003. République d’Haïti, Programme des Nations
Mathematical models, environmental hydraulics, vol 2. Wiley, Unies pour le Développement, Département de la Coopération
London, pp 71–92 Technique pour le Développement, UN, New York
Thiéry D (2010b) Groundwater flow modeling in porous media using
MARTHE. In: Tanguy J-M (ed) Modeling software, environmental Publisher’s note Springer Nature remains neutral with regard to jurisdic-
hydraulics series, vol 5. Wiley, London, pp 45–60 tional claims in published maps and institutional affiliations.