Content area
The degradation of water resources by diffuse pollution, mainly due to nitrate and pesticides, is an important matter for public health. Restoration of the quality of natural water catchments by focusing on their catchment areas is therefore a national priority in France. To consider catchment areas as homogeneous and to expend an equal effort on the entire area inevitably leads to a waste of time and money, and restorative actions may not be as efficient as intended. The variability of the pedological and geological properties of the area is actually an opportunity to invest effort on smaller areas, simply because every action is not equally efficient on every kind of pedological or geological surface. Using this approach, it is possible to invest in a few selected zones that will be efficient in terms of environmental results. The contributive hydraulic areas (CHA) concept is different from that of the catchment area. Because the transport of most of the mobile and persistent pollutants is primarily driven by water circulation, the concept of the CHA is based on the water pathway from the surface of the soil in the catchment area to the well. The method uses a three-dimensional hydrogeological model of surface and groundwater integrated with a geographic information system called Watermodel. The model calculates the contribution (m3/h or %) of each point of the soil to the total flow pumped in a well. Application of this model, partially funded by the Seine Normandy Basin Agency, to the catchment of the Dormelles Well in the Cretaceous chalk aquifer in the Orvanne valley, France (catchment area of 23,000 ha at Dormelles, county 77), shows that 95 % of the water pumped at the Dormelles Well comes from only 26 % of the total surface area of the catchment. Consequently, an action plan to protect the water resource will be targeted at the 93 farmers operating in this source area rather than the total number of farmers (250) across the entire 23,000 ha. Another model, developed from Epiclès© software, permits the calculation of the under-root nitrate concentrations for each field based on soil type, climate, and farming practices. When the Watermodel and Epiclès© are coupled, nitrate transfers from the soil to the catchment and the river can be modeled. In this study, the initial pollution due to the actual farming practices was simulated and we were also able to estimate the efficiency of the agronomic action plan by testing several scenarios and calculating the time needed to reach the target nitrate concentration in the well.
http://crossmark.crossref.org/dialog/?doi=10.1007/s11356-015-5459-6&domain=pdf
Web End = http://crossmark.crossref.org/dialog/?doi=10.1007/s11356-015-5459-6&domain=pdf
Web End = Environ Sci Pollut Res (2016) 23:1584115851 DOI 10.1007/s11356-015-5459-6
INTERNATIONAL CONFERENCE ON INTEGRATED MANAGEMENT OF THE ENVIRONMENT - ICIME 2014
Use of modeling to protect, plan, and manage water resources in catchment areas
Thibaut Constant1 & Sverine Charrire1 &
Abdejalil Lioeddine1 & Yves Emsellem2
Received: 31 August 2015 /Accepted: 17 September 2015 /Published online: 9 October 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract The degradation of water resources by diffuse pollution, mainly due to nitrate and pesticides, is an important matter for public health. Restoration of the quality of natural water catchments by focusing on their catchment areas is therefore a national priority in France. To consider catchment areas as homogeneous and to expend an equal effort on the entire area inevitably leads to a waste of time and money, and restorative actions may not be as efficient as intended. The variability of the pedological and geological properties of the area is actually an opportunity to invest effort on smaller areas, simply because every action is not equally efficient on every kind of pedological or geological surface. Using this approach, it is possible to invest in a few selected zones that will be efficient in terms of environmental results. The contributive hydraulic areas (CHA) concept is different from that of the catchment area. Because the transport of most of the mobile and persistent pollutants is primarily driven by water circulation, the concept of the CHA is based on the water pathway from the surface of the soil in the catchment area to the well. The method uses a three-dimensional hydrogeological model of surface and groundwater integrated
with a geographic information system called Watermodel. The model calculates the contribution (m3/h or %) of each point of the soil to the total flow pumped in a well. Application of this model, partially funded by the Seine Normandy Basin Agency, to the catchment of the Dormelles Well in the Cretaceous chalk aquifer in the Orvanne valley, France (catchment area of 23,000 ha at Dormelles, county 77), shows that 95 % of the water pumped at the Dormelles Well comes from only 26 % of the total surface area of the catchment. Consequently, an action plan to protect the water resource will be targeted at the 93 farmers operating in this source area rather than the total number of farmers (250) across the entire 23,000 ha. Another model, developed from Epicls software, permits the calculation of the under-root nitrate concentrations for each field based on soil type, climate, and farming practices. When the Watermodel and Epicls are coupled, nitrate transfers from the soil to the catchment and the river can be modeled. In this study, the initial pollution due to the actual farming practices was simulated and we were also able to estimate the efficiency of the agronomic action plan by testing several scenarios and calculating the time needed to reach the target nitrate concentration in the well.
Keywords Hydrogeology . Modeling . Water quality . Contribution . Watermodel . Epicls . Nitrate . Catchment
Background to the Orvanne basin
The Intercommunal Union of Drinking Water Supply of the Orvanne Valley provides more than 3000 inhabitants with drinking water. The drinking water resource is provided by a well located in Dormelles (Seine-et-Marne County). Water-quality-monitoring data shows occasional exceedances of the nitrate and pesticide limits (50 mg/l for nitrates, 0.1 g/l
Responsible editor: Philippe Garrigues
* Sverine Charrire [email protected]
Thibaut Constant [email protected]
Abdejalil Lioeddine [email protected]
Yves Emsellem [email protected]
1 InVivo AgroSolutions, 83 avenue de la Grande Arme, CEDEX 16,75782 Paris, France
2 Watermodel, 33 Boulevard Foch, 66000 Antibes, France
15842 Environ Sci Pollut Res (2016) 23:1584115851
for individual pesticides, and 0.5 g/l for the sum of pesticide concentrations (Parlement europen 1998). The Dormelles Well (19 m deep) is located 5 m from the river and drains to the Late Cretaceous unconfined chalk aquifer under the Orvanne alluvium. It has a flow rate of 200 m3/h. The catchment area (CA) of the well has been delimited to an area of 23,000 ha using the method recommended by Vernoux et al. (2007) (Fig. 1).
The nitrate concentrations in this area had been continuously increasing since the 1970s (Fig. 2). Even though the trend is now reversed and they have been decreasing slowly since 2004, the concentrations are still as high as 37.5 mg/l, which represents a threshold of action reinforced by the Seine Normandy Basin Agency.
The water samples were analyzed by either the Seine Normandy Basin Agency laboratory or the Regional Health Agency. The sampling frequency is chosen by the owner of the well (in this case, the municipality) and must comply with the minimum required by the law (Assemble nationale 2010).
Structure of the model
An integrated surface-/groundwater approach is required to obtain the most reliable quantitative water balance possible at the CA scale. Watermodel was chosen because it is an integrated surface-/groundwater model (Emsellem et al. 1994).
Based on the geology of the area, a multilayer system has been set up in Watermodel, which includes the following horizons:
& Layer 0: Soil& Layer 1: Superficial formations: alluvium and colluvium & Layer 2: Tertiary Oligocene and Eocene plateaus
& Layer 3: Sand, gravel, clay, and sandstone of the Ypresian clays and Silex.
& Layer 4: Chalk aquifer
The thickness of each layer has been calculated by interpolation of geological logs from the wells in the area, available from the national subsoil databank (BRGM 2009).
Grid construction
The area chosen for the modeling was larger than the CA area stated earlier to permit consideration of topographical and piezometric extensions of the basin. The area was then split into 200200 m cells and refined to 100100 m cells next to the river and to the wells (Fig. 3). The completed model comprised 46,671 grid cells.
Data collection
The maps of transmissivity, vertical/horizontal permeability, matrix/fissure porosity, and storage capacity were determined from other existing hydrogeological studies (Mgnien 1970; Roux 2006). Piezometric data of the unconfined chalk aquifer in the area were obtained from previous studies (Mgnien 1970) and during two campaigns of water-table measurements at high and low water levels in 2010. During these campaigns, the piezometry was measured in 174 existing private wells in the chalk aquifer and perched water tables. These data were linked spatially and interpolated to produce two piezometric maps of the high and low water levels of 2010. These maps were then used to calibrate the model. Climatological data on a daily time-step for the last 21 years were taken directly from Meteo France. Topographical data were available at a scale of 5050 m from the National Geographic Institute (IGN). Soil rugosity (given
Fig. 1 Location of the Dormelles Well and its catchment area (CA)
Environ Sci Pollut Res (2016) 23:1584115851 15843
Fig. 2 Nitrate concentrations of several wells catching the Chalk aquifer in the Dormelles area since the 1970s (Cf. red dots in Fig. 1)
as a maximum speed against runoff in m/h1) was defined by tables linked to the soil cover and the soil water reserve from soil maps.
Calibration of steady state and transient mode
The model was calibrated with 2010 as a reference because this year was representative of the average hydrological conditions of the last 10 years and there was good piezometric data available.
After the little-known parameters were calibrated (those based on the literature only), virtual piezometric maps were produced (Fig. 4), which, when compared, calibrated well with the real ones (174 were measured) in high and low levels (R2=99 % for each case).
The model was run several times in transient mode to refine the calibration and, in particular, the spatial distribution of the hydrodynamic parameters. The runs were made for the period from 1989 to 2010 using observed rainfall data. The calibration also appears suitable for transient mode because the
annual and seasonal values and variations are well transcribed by Watermodel (Fig. 5).The following histogram and graph (Fig. 6) show the frequency distribution of the residual errors. These residual errors are concentrated between 2.5 and +3 m, but the highest frequencies are distributed between 1 and +1 m. The mean value of the residual errors is +0.03 m, which is quite good considering the size of the area studied and that a perfect calibration would be +0 m.
The calibration of the river flow rate at the Blennes hydro-metric station on the Orvanne River is presented in Fig. 7. Watermodel faithfully reproduced the measured flow between 1989 and 2010 (Fig. 7); it only slightly underestimated the flow rates for the rainy years (for example, 20002001) and slightly overestimated the flow for the dry years (for example, from 1992 to 1993). It is noteworthy that the differences are very small, considering the scale of the flow rates (<2 m3/s).
The model will be validated by new field measurements (in particular, piezometry and nitrate concentrations) for the years after the calibration year, but the present calibration was considered sufficiently accurate to obtain the following results.
AlluviumLoessSandy colluvium Flint clay
Etampes limestone
Fontainebleau sand
Champigny limestone
Ypresian clay Ypresian quartzite
Ypresian sand Chalk
Alluvium
Ter ary limestones
Ypresian (clay and sand)
Cretaceous chalk
Fig. 3 Grid of the model and NE-SW geological cross-section of the CA
15844 Environ Sci Pollut Res (2016) 23:1584115851
Fig. 4 Modeled piezometric map of the chalk aquifer in a steady state2010 (legend in m NGF)
scale
One of the ongoing aims of the study is to validate the model and to eventually recalibrate it regularly to get the best long-term results.
Contributive hydraulic areas
After the model was calibrated and, because of the extent of the CA, the contributive hydraulic area (CHA) was calculated, the next step was to determine the contribution of each soil cell to the flow rate at the Dormelles Well. The result is presented in Fig. 8.
The calculations for areas including 50, 70, 80, 90, and 95 % of the total measured flow (200 m3/h at the Dormelles Well) are presented in Fig. 9.
Therefore, modeling results show that 95 % of the flow at the Dormelles Well comes from 26 % of the CA. Only 93 farmers are included in this area, rather than the total number of farmers (250) in the CA.
Coupling Watermodel with the nitrate leaching model Epicls
Epicls: modeling nitrate leaching
The quantity of nitrogen in the soil at the beginning of the drainage period (when the rainfall is higher than the evapotranspiration and the soil water capacity is full) can be considered as the pool of nitrogen that is available for leaching in the winter period (which is between October and March for this case study). Modeling the quantity of nitrogen in the soil is an indicator of the impact of farming practices on each field. This indicator is linked to the excess of nitrogen used compared with the yield of the previous crop, the presence of cover crops, organic manure use in autumn, and the soil type.
Epicls was used to calculate this indicator. It is a decision support tool for fertilization that is used on 2 million ha in France and is validated each year by comparing the modeled
Fig. 5 Modeled piezometry (red valley context, green plateaus context) compared with the real pluviometry between January 2001 and January 2010
Environ Sci Pollut Res (2016) 23:1584115851 15845
Fig. 6 Histogram of the residual errors of the piezometric calibration and graph of the correlation between the measured and modeled piezometry in April 2010
and measured nitrogen pools in soil. In the Orvanne study, the use of Epicls was based on 4 years of agricultural data; this time period was chosen to permit comprehensive consideration of the crop rotation in the basin.
Epicls used the nitrogen balance method to calculate the quantities of nitrogen at the rooting depth of crops (Fig. 10) for each soil type. This method takes into account the following range of nitrogen inputs or outputs over time:
& Main inputs:
Mineral nitrogen and organic manure inputs Ground mineralization Mineralization of the previous crop straws
& Main outputs:
The nitrogen exported by the crop at harvest The nitrogen taken up by cover crops
At the crop rotation scale, the mean nitrogen quantities after harvest and before leaching (NBL; nitrogen quantity in the soil under-root depth given in kg/ha) in the Dormelles CA (agronomic diagnosis for 20082011) are presented in Fig. 11.
The mean nitrogen balance varied annually and ranged from 28 kg N/ha (good harvest in 2009) to 55 kg N/ha (bad harvest in 2006). Fertilization management at the field scale can be evaluated using this indicator.
The mean NBL for all crops (except meadows and fallow) ranged from 48 to 80 kg N/ha, depending on the year. The mean interannual NBL was estimated at around 60 units of nitrogen.
It is noteworthy that the nitrogen balance was lower in 2008 when the previous crop was harvested; the increase between the nitrogen quantity at harvest and the NBL can be explained by an increase in the area fertilized with manure during autumn 2008.
The quantity of nitrogen in soils estimated by Epicls is validated every year by sampling soils at a range of points in the areas of the 40 cooperatives that use Epicls in France (Fig. 12). Further, for this particular study, local validation has been carried out on 100 fields of the Dormelles CA in partnership with the Seine Normandy Basin Agency (Fig. 13).
There was a good correlation between the measured and modeled results: the mean difference between the measured (25 kg/ha) and modeled (28 kg/ha) NAL values was only 3 kg/ ha. Further, the distribution of the values (standard deviation) was similar and the uncertainty of the modeled results was actually much lower than the uncertainty of the field measurements.
Simulation of the nitrate transfer
The field map of the mean nitrate leaching concentrations produced by Epicls was plugged into the soil layer of Watermodel (Fig. 14).
Fig. 7 Modeled and real flow rates of the Orvanne River between 1990 and 2010
15846 Environ Sci Pollut Res (2016) 23:1584115851
Fig. 8 Contributions to the Dormelles Well from the soil layer for a flow rate of 200 m3/h, calculated with Watermodel
Contribution in m /h
(the sum is 200 m /h, which is the flow rate at the Dormelles well)
scale
Results show that the coupling of the two models gave reasonable estimations of the concentrations measured at the Dormelles Well. The modeled concentrations for 2010 were36.6 mg/l, while the measured concentrations were 37 mg/l
(analysis carried out by either the Seine Normandy Basin Agency or the Regional Health Agency). This result confirms that coupling of the two models seems to be helpful for understanding nitrate contamination of the aquifer.
Fig. 9 Contributive hydraulic areas of Dormelles Well (m3/h)
Contribu ve Hydraulic Area (CHA)
Environ Sci Pollut Res (2016) 23:1584115851 15847
Soil mineraliza on
Nitrogen consump on
Quan ty of nitrogen at harvest
nitrogen quan ty before leaching (NBL)
nitrogen quan ty a er leaching (NAL)
+ Soil mineraliza on+ mineraliza on of the previous crop straws+ Rear eects of organic inputs+ Direct eect of nitrogen input- covercrops- nitrogen catched by the crop
Fig. 10 Parameters involved in the calculation of the nitrogen stock in the soil
Fig. 11 Nitrogen balance after harvesting and the nitrogen stock before the leaching period (NBL) on the CA
Nitrate concentra on
in the leaching water
The transient mode demands a lot of time to model the long time period needed for nitrate transfer in the unsaturated zone, so the climatic time series was replicated every 10 years to cover a simulation period of 90 years.
Simulations of nitrate concentrations in the Orvanne River
Nitrate concentrations (nitrate dosage) in the Orvanne River were measured in two campaigns during the CA study in April
and November. The measured concentrations were correlated with the results modeled by Watermodel (Fig. 15) and also proved the hypothesis that there was no denitrification (denitrification can be modeled in Watermodel if needed).
Simulations of prospective scenarios
Climate change has not been considered in this first test, but climatic data are needed to simulate future scenarios. Climate data from the Melun station data that span 21 years were
15848 Environ Sci Pollut Res (2016) 23:1584115851
80
nitrogen balance a er leaching (NBAL) measured (mean) in kgN/ha
Fig. 12 Comparison of the modeled (Epicls) and measured means for two agronomic situations of the nitrogen quantity in soils after leaching (NBAL) from a database of the nitrogen quantity in 4000 soils analyzed over 5 years at the Nouricia cooperative (Aube County, France)
70
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 nitrogen balance a er leaching (NBAL) by Epicls (mean) in kgN/ha
chosen for this study and will be replicated for a period of 70 years. Actually, nitrate transport modeling needs a long simulation period to take account into the thickness of the unsaturated zone and the slow motion of the vertical transfers in this zone (Lacherez-Bastin 2005). For example, the nitrate migration speed under 5 m of clay is about 0.6 m/year; in chalk under a few meters of silt, it is about 0.8 to 1 m/year, while in chalk under sand, it is about 1.25 m/year. These are average values, and it should be acknowledged that nitrate can be rapidly flushed through the different layers in very wet winters.
To forecast the ongoing evolution of the nitrate pollution from a range of farming scenarios, and to predict the scenario that will achieve the best outcome, several runs have been tested from an initial state (Fig. 16). The details of each scenario are as follows:
Scenario 1: no change in farming practices. The aim is to determine the impact of current practices on the water quality.
Scenario 2: application of an agronomic action plan over the whole CA that includes a decrease in the nitrogen export from the previous crop by setting a reasonable yield objective, modifying fertilizer applications using decision support tools, considering climatic and pedological characteristics, and optimizing the fertilization time and application rates during the season. Nitrogen export will be promoted in autumn/winter (seed and adaptable intermediate crops) and autumn soil mineralization will be limited by maintaining the C/N ratio of organic inputs at less than 15.
To simulate this scenario, the Epicls map of current farming practices was replaced in Watermodel by the Epicls map of current practices modified to take account of the action plan listed above. The simulation was run for the time period from 2012 to 2102.
Scenario 3: application of the same agronomic action plan but limited to 95 % of the CHA. The current farming practices remain the same on the rest of the CA. This scenario will test if the CHA are qualitatively and quantitatively relevant.
Scenario 4: application of the same agronomic action plan, plus back to grass for several fields. The back-to-grass fields account for 70 % of the CHA and include those in which the nitrate concentration are high (case 1: still over 75 mg/l, case 2: still over 61 mg/l), even after the agronomic action plan is applied.
Fig. 13 Box plots of the measured and modeled NBAL for the fields of the Dormelles CA
Nitrogen quan ty a er the leaching period (kg/ha)
Measured
Modelled
Environ Sci Pollut Res (2016) 23:1584115851 15849
Fig. 14 Nitrate leaching concentration map produced by Epicls (right) and nitrate concentrations in the chalk aquifer produced by Watermodel (year 2010)
The next graph (Fig. 17) shows the results of the nitrate concentrations in the Dormelles Well under these five scenarios.
The results show that:
& Nitrate concentrations will continue to increase slightly if no action plan is applied, but will remain lower than the drinking water threshold of 50 mg/l.
& Scenarios 2 and 3 appear to have the same long-term result, which confirms the hypothesis that the CHA 95 % is
qualitatively relevant as a working area for water quality and quantity.& Scenarios 2 and 3 show that an agronomic action plan can induce a decrease in nitrate concentrations, in this case from 36.7 to 31 mg/l.& After 70 years, the back-to-grass scenarios allow a minimal gain in concentrations of between 1 and 2 mg/l compared with the agronomic plan alone.& Nitrate concentrations increase under scenario 1. For the other scenarios, there is a rapid phase (for each scenario)
Fig. 15 Snapshot of the Watermodel results for the nitrate concentrations in the Orvanne River (left) and comparison of the modeled and measured results (right)
15850 Environ Sci Pollut Res (2016) 23:1584115851
Fig. 16 Nitrate leaching concentration maps produced by Epicls and translated into Watermodel (layer 0) for four case studies of agronomic practices
Fig. 17 Nitrate concentration in the Dormelles Well for the five scenarios
of decrease for the first 6 years, probably due to rapid nitrate transfers from limited areas of the CA. This is followed by a period of almost stagnation over the next 6 years. After this, the concentrations continue to decrease
slowly until 2056, at which point a stagnation period begins.& There is repetition in the few visualized figures, because the climatic data are replicated every 21 years.
Environ Sci Pollut Res (2016) 23:1584115851 15851
Conclusions
The results of this integrated hydrogeological modeling (surface and groundwater) study of the CA of the Dormelles Well under Watermodel show that:
& Piezometry can be calibrated with data obtained in 2010 from 58 field measurements in the chalk aquifer and 13 measurements in the perched aquifers at high and low water levels.
& It was possible to calibrate the Orvanne River flow with data from the Blennes hydrometric station, which is located within the CA and has data from 1989. As the model is integrated, the calibration of the river flow is linked with the piezometry and vice versa.
& The model allowed refinement of the CA delimitation, taking into account seasonal and annual variations of the piezometry. The revised area of the whole CA is 24, 800 ha.
& Watermodel allowed the calculation of the CHA of the Dormelles Well. It was possible to determine the surface of the CA that contributed up to 50, 70, 80, 90, and 95 % to the measured flow. It is noteworthy that only 26 % of the whole CA actually contributed to 95 % of the total observed flow. This geographic information means that the CHA area can be split to focus on the agronomic action plan. These actions are effective in an agronomic way, and focus on the areas that contribute most to the pumped flow. This method also allows for time and money to be saved during the management of the action plan by focusing both on efficient actions that are agronomically effective and on the areas that contribute most to the catched flow
& Coupling of Epicls and Watermodel permits simulation of variations in the nitrate concentrations in the aquifers, the rivers, and in the drinking water catchments.
& This coupling allows the ongoing development of these nitrate concentrations to be forecast in time and space, depending on the scenarios. Therefore, it is possible to predict if the objectives of the Seine Normandy Basin Agency can be achieved.
& The back-to-grass scenarios do not seem to be more efficient than the agronomic action plan.
References
Assemble nationale (2010) Arrt du 21 janvier 2010 modifiant larrt du 11 janvier 2007 relatif au programme de prlvements et danalyses du contrle sanitaire pour les eaux fournies par un rseau de distribution, pris en application des articles R. 1321-10,R. 1321-15 et R. 1321-16 du code de la sant publiqueBRGM (2009) La Banque du sous-sol : un capital pour lenvironnement et la connaissance gologiqueEmsellem Y, Pfister C, Millerliri A, et al (1994) La gestion active des aquifres laide des modles. BRGM, pp 8995 Lacherez-Bastin S (2005) Contribution ltude de la migration des nitrates dans le sol et la zone non sature de la nappe de la craie dans le Nord de la France : modlisation intgre des nitrates dans le bassin versant de lEscrebieux. Universit Lille1 - Sciences et Technologies, FranceMgnien C (1970) Atlas des nappes aquifres de la rgion parisienne.
BRGM, FranceParlement europen (1998) DIRECTIVE 98/83/CE DU CONSEIL du 3 novembre 1998 relative la qualit des eaux destines la consommation humaineRoux JC (2006) Aquifres et eaux souterraines en FranceVernoux JF, Wuilleumer A, Drflinger N (2007) Dlimitation des bassins dalimentation des captages et cartographie de leur vulnrabilit vis-vis des pollutions diffuses
Springer-Verlag Berlin Heidelberg 2016