3给水排水--外文翻译-外文文献-英文文献.doc

Relations between triazine flux, catchment topography and distancebetween maize fields and the drainage networkF. Colina,*, C. Puecha, G. de Marsilyb,1aUMR “Systemes et Structures Spattiaux”, Cemagref-ENGREF 500, rue J.F. Breton 34093, Montpellier Cedex 05, FrancebUMR “Structure et Fonctionement des Systemes Hydriques Continentaux”, Universite P. et M. Curie 4, Pl. Jussieu 75252,Paris Cedex 05, FranceReceived 5 October 1999; revised 27 April 2000; accepted 19 June 2000AbstractThis paper puts forward a methodology permitting the identification of farming plots contributing to the pollution of surface water in order to define the zones most at risk from pesticide pollution. We worked at the scale of the small agricultural catchment (0.27.5 km2) as it represents the appropriate level of organisation for agricultural land. The hypothesis tested was: the farther a field undergoing a pesticide treatment is from a channel network, the lower its impact on pollution at the catchment outlet.The study area, the Sousson catchment (120 km2, Gers, France), has a “herring bone” structure: 50 independent tributaries supply the main drain. Pesticide sales show that atrazine is the most frequently used compound although it is only used for treating maize plots and that its application rate is constant. In two winter inter-storm measurement exercises, triazine flux values were collected at about 30 independent sub-basin outlets.The contributory areas are defined, with the aid of a GIS, as different strips around the channel network. The correlation between plots under maize in contributory zones and triazine flux at related sub-basin outlets is studied by using non-parametric and linear correlation coefficients. Finally, the most pertinent contributory zone is associated with the best correlation level.A catchment typology, based on a slope criterion, allows us to conclude that in steep slope catchments, the contributory area is best defined as a 50 m wide strip around the channel network. In flat zones, the agricultural drainage network is particularly well developed: artificial drains extend the channel network extracted from the 1/25.000 scale topographic map, and the total surface area of the catchment must be taken to account. q 2000 Elsevier Science B.V. All rights reserved.Keywords: Pesticide catchment; GIS artificial network1. IntroductionThe use of pesticides in western agriculture dates back to the middle of the 19th century (Fournier,1988). Since then, because of their intensive use,yields have increased and the demand for agricultural products has been satisfied. However, the pollution created by their use threatens both drinking water resources and the integrity of ecosystems. Therefore, there is a great demand for the reduction of pollution.The remedies lie in changes in the way that agricultural land is managed. The problem of agricultural Journal non-point source pollution by pesticides must be taken from the field, the level of action, to the catchment,the level of control of the water resource.Between these two spatial scales, different levels of organisation can be found. Fields, groups of fields,basins and main catchment, can be viewed together as nested systems (Burel et al., 1992). For each scale level, the main processes governing water movement and soluble pollutant transport are different, as are the variables characterising the system (Lebel, 1990):flow in macropores at local scale, preferential flowpaths at the hillslope scale, flows in connection withthe repartition of different soils at the catchment scale,geology influence at the regional scale (Blosch and Sivapalan, 1995).At the field level, an experimental approach can be used and the relative weight of each variable can be experimentally tested (Scheunert, 1996; Bengtson et al., 1990). The major factors that concern agricultural practices have been identified and many agricultural management indicators have been developed (Bockstaller et al., 1997).Nevertheless, this approach cannot be applied at the catchment scale for several reasons: the need to measure the pollution and the environmental factors simultaneously, multiple measurement difficulties, the complexity of analysis. The variability of observations has temporal and spatial components. Rain induces pesticide leaching and therefore causestemporary high pesticide concentrations in the water; the closer the pesticide spreading date in thefield is to the measurement, the greater the concentration levels (Seux et al., 1984; Reme, 1992; Laroche and Gallichand, 1995). The extensive use of Geographical Information System (GIS) has made it possible to analyse the impact on the pollution of the spatial characteristics of agricultural zones (Battaglin and Goolsby, 1996). But so far, the results of these experimentshave only led to an approximate estimate of the risks (Tim and Jolly, 1994).In order to progress in the search for ways to reduce pesticide pollution, it would be worthwhile to improve our assessment of how spatial structure and organisation affects the levels of pollutants measured.This paper presents the results of a study that concerns a particular aspect of the influence of spatial organisation on pesticide transfer: the effects of the distance between the cropland and the channel network. The longer the distance between a cultivated field and a river, the greater the retention and degradation processes (Leonard, 1990; Belamie et al.,1997). One might therefore imagine that the greater the distance, the lower the pollution level. However,few studies have given a numerical value to the critical distance at which a field does not influence river pollution significantly. Usually, when dealing with risk zone definition, experts establish an arbitrary distance (Bouchardy, 1992). Our main goal is to determine through spatial analysis the critical distance from a hydrographic network. The zones most at risk from pesticides, including the plots, which contribute most of the pollution, can then be determined.The study area, the Sousson catchment (Gers,France) has certain physical characteristics, which allows sampling of most of the independent subbasins, defined here as agricultural production zones. Its particular morphology made the comparative study of the production zones possible. The method involves a statistical comparison between pollution measurements and spatial characteristics of the catchments. In order to establish the boundaries ofthe contributing areas, the pollution flux measured at the production zone outlet is compared to the landcover, estimated within strips of variable width around the channel network. Results are shown and discussed from a mainly practical viewpoint.2. The study area and collected data2.1. Study area descriptionThe study area is the Sousson catchment, in southwestern France (Gers). The Sousson River is a tributary of the river Gers. The catchment area is 120 km2. The 32 km long hydrographic network has a herringbonepattern: 53 sub-basins with fairly homogeneous surfaces areas ranging from 0.2 to 7.5 km2 serve the central drain (Fig. 1).The wide, gently sloping and heavily cultivated left bank, differs from the right bank, which is narrow, steep and mainly made up of forest and pastureland.The Sousson catchment area is exclusively agricultural.There is no industry or settlement of more than 200 inhabitants. The two main crops cultivated aremaize and winter wheat (17 and 15% of the catchment surface area, respectively). The maize fields are usually situated, on the left bank, in the upstream middle of the catchment area, and along the main river. There are two types of soil: a calcareous soil, which is quite permeable, and a non-calcareous soil called locally boulbenes with an top limoneous layer and a lower silty layer. In order to avoid the stagnation of water in the upper layer caused by the silty impermeable layer, the fields on boulbene soil are artificially drained. Maize is cultivated for preference on thistype of soil.No significant aquifer has been found in the catchment, as the substratum is rather impervious (clays).2.2. Collected data2.2.1. Spatial dataA GIS was developed for the area, which contains the following information layers: the hydrographic network and the catchment boundaries digitized from 1/25.000 scale topographic map; a gridded Digital Elevation Model (DEM) of the zone providing landsurface slopes generated from DEM with a resolution of 75 m; the boundaries of cultivated fields digitized from aerial photos at scale of 1/15.000; landcover for both 1995 and 1996 was defined in detail in the study area. For 1997, landcover was identified by remote sensing. Knowledge of agricultural antecedents enhanced the classification of a SPOT (Satellite Pour l0Observation de la Terre) image. As a result, the maize areas for the entire Sousson catchment were determined for 1995, 1996 and 1997 (Fig. 2).GIS functions are capable of determining the landcover of each catchment by intersecting the two information layers “landcover” and “catchment boundaries”, or defining a zone of constant width around the hydrographic network, which is called the buffer zone. In order to evaluate the pesticide application rate, figures for local pesticide sales were collected. Atrazine, alachlor and glyphosate are the most commonly used compounds, atrazine far outstrips the others triazines as the most frequently used product (ten times less simazine is sold). In this region, atrazine is only used in maize cultivation. The application rate (mass of atrazine sold/maize surface area) does not vary from one municipality to another.To simplify the investigations, we chose to study the atrazine spread on maize plots in May. We assume that all the maize plots are treated with atrazine and that the application rate is uniform.2.2.2. Water pollution dataTwo series of measurements were made during the winter period: 23 sub-basins were sampled on December 3rd and 4th 1997, and 26 sub-basins were sampled March 17th to 19th 1998. Hence, the atrazine treatments were carried out 7 or 10 months before and the maize harvest was 1 and 4 months before the measurements were taken.To obtain stable hydrological conditions, the chosen measurement dates coincided with decreasing flow as shown in Fig. 3. The same operator collected the quality samples and gauged the river flow in order to limit measurement errors.The triazine concentration was measured with an ELISA water test (Transia Plate PE 0737). This measurement technique is less accurate than the classical chromatography technique, but it permits a faster analysis of a large number of samples (Rauzy and Danjou, 1992; Lentza-Rios, 1996). As atrazine is the most widely commercialised triazine product in this region, we will consider that observed triazine concentrations are representative of atrazine concentrations.December 1997 values, and March 1998 values were grouped together in order to assemble a large enough sample for statistical analysis (Fig. 4). The instantaneous triazine flux was obtained by multiplying the triazine concentration with the dischargevalue. As shown in Table 1, water flow in December 1997 was double that in March 1998, but the corresponding triazine flux are comparable.2.2.3. Quality assuranceTo control the quality of ELISA water-test measurements, each concentration was analysed142 F. Colin et al. / Journal of Hydrology 236 (2000) 139152 Fig. 2. Hydrographic network (topographic 1/25.000 map) and subcatchments, parcel limits and land-cover (example of maize plots). twice. A maximum difference of 20% is tolerated between two duplicate samples, the median error is 10%, and mean values are used. It is possible that ELISA measurement induces a consistent error by comparing with gas chromatography measurements (Tasli et al., 1996), but this bias is compensated by comparative reasoning on all the samples.A few points were measured two or three times during the exercise in order to evaluate the daily variations during the sampling period. Table 2 shows that the flux variation between different days of a sampling period ranges from 2 to 49%. It is therefore possible to compare the different samples from the period in question. All the measurements from each period are then grouped together.The uncertainty on the triazine flux is the sum of the uncertainty of discharge and concentration measurements. The uncertainty on the discharge measurements ranges from 15 to 20%. Therefore, the triazine flux value is given with a maximum uncertainty of 40%.3. MethodTo define the zones most at risk we tested how the distance to the river of the areas where pesticides are applied influence pollution levels. Thus, we have to determine the relative position of the hydrographic network and the contaminating plots. In our case, the data on pollution is provided by triazine flux measurements taken at basin outlets and the potentially contaminating fields are maize plots.3.1. Efficiency curve and spatial partitionThe basic hypothesis is that the impact of the field as a contributor to pollution decreases the further it is from the channel network. Thus, there is a critical distance at which the field makes little contribution to outlet pollution. In other words, we assume that plot contribution to pollution level can be modelled through a decreasing efficiency curve. This hypothesis will be tested with a very simple curve: a step function. This curve is defined using only one parameter, the threshold limit distance, d, beyond, which a plot stops contributing to river pollution.In practice, this hypothesis implies a three-step approach: determination of the location of the maize fields; definition of a buffer of width d, equal to the threshold distance and, which surrounds the channel network; determination of the contaminating fields inside these limits.The fields define the contributing maize areas depending on the buffer width (Fig. 5). At this stage, GIS functionality is required, particularly for the buffer function.3.2. Correlation between contributing area and pollution at the catchment outletWe studied the correlation level between triazine flux measured at the catchment outlet and the different contamination contributing areas defined by strips of variable width. Three parameters are used to determine the correlation level (further information is provided on this point in Appendix A): The Kendall rank correlation coefficient (Siegel, 1956) t gives a measure of the degree of association or correlation between two sets of ranks. It expresses the difference between the probability that the two data sets are ranked according to the same order and the probability that they are ranked according to a different order. If t . 1.21.; a positive (negative) relation exist