446 Remote Sensing and Hydrology 2000 (Proceedings of a symposium held at Sanla Fe, New Mexico, USA, April 2000). IAHS Publ. no. 267, 2001. Satellite remote sensing and GIS used to quantify water input for rice cultivation (Rhône delta, France) PHILIPPE CHAUVELON, ALAIN SANDOZ, VINCENT HEURTE AUX & ANTOINE BERCEAUX Fondation Sansouire, Station Biologique de la Tour du Valat, Le Sambuc, F-13200 Arles, France e-mail: chauvelonfsitour-du-valat.com Abstract In the Rhône River delta (France), flooded rice cropping requires constant irrigation from April to September. Irrigation volumes used are highly variable, and depend greatly on specific cropping practices. Land use characteristics were digitized, updated with classified satellite images, and placed in a GIS format. The actual irrigation volumes were calculated from estimated pump flows and the distribution of rice area within each irrigation basin. Empirical relationships were obtained between specific irrigation volume and rice area for each monitored pumping station. Thus it was possible to estimate water input for previous years from geo-referenced rice-yield data only. This step is necessary to model the hydrology of the system. Without geo-referenced knowledge of irrigation water input, it is not possible to estimate runoff to the lagoons. Key words GIS; irrigation volumes; land use; Rhone River delta, France; rice INTRODUCTION Remote sensing and GIS are important tools for characterizing land use dynamics and spatial structure for water resource management, and especially for input to hydrological models (Baumgartner & Apfl, 1996; Rango & Shalaby, 1996). The Great Camargue is located in the central Rhone delta, (Fig. 1(a)). It is a highly anthropic and complex hydro-system, with agricultural basins, marsh areas, and shallow brackish lagoons. Rice is cultivated on levelled fields of 1 to 3 ha, delimited by bunds. The fields are flooded during the growing period (mid April or early May, to September) and drained by a meshed network of ditches and canals. Naturally draining basins (Fig. 1(a)) drain into the lagoons, and total 112.5 km 2, while polder areas total 310.7 km 2. Apart from the spatial variability of irrigation input, problems for hydrological modelling lie in the fact that some irrigation basins are drained by two different drainage networks, and that some drainage transfers may occur between polderized and unpolderized basins (Heurteaux, 1994; Chauvelon, 1998). This hydro-system experiences a Mediterranean climate, with an average annual precipitation of 620 mm (250-1050 mm) and average évapotranspiration of 1300 mm.
Quantification of water input for rice cultivation (Rhône delta, France) 441 (a) (b) Fig. 1 (a) Map of the Great Camargue showing the different drainage basins and position of pumping stations, (b) Structure of the GIS database, discretized in irrigation basins. MATERIALS AND METHODS GIS and remote sensing Aerial photographs (1:20 000) were digitized, ortho-rectified, and geo-referenced. A total of 11 000 agricultural parcels were identified (Fig. 1(b)). Land use was further updated with classified satellite images (Table 1). Rice fields are best identified with images acquired during July or August.
448 Philippe Chauvelon et al. Table 1 Satellite images and classification methods used. Date of image acquisition Satellite Channels Classification 23 July 1975 Landsat MSS 1,3,4 Unsupervised 28 April 1981 Landsat MSS 1,3,4 Unsupervised 14 July 1984 Landsat-5 TM 1,3,4 Unsupervised 17 July 1985 Landsat-5 TM 1,3,4 Unsupervised 05 June 1987 Landsat-5 TM 1, 3,4 Unsupervised 13 August 1989 Landsat-5 TM 1,3,4 Unsupervised 19 August 1991 Landsat-5 TM 1, 3,4 Unsupervised 23 July 1993 Landsat-5 TM 1,3,4 Supervised 24 June 1994 Landsat-5 TM 1, 3,4 Supervised 29 July 1995 Landsat-5 TM 1,3,4 Supervised 11 July 1996 Spot XS 1,2,3 Supervised 08 May 1997; 12 June 1997; ERS-2 C-band Multidate and 21 August 1997 supervised 18 May 1998 Landsat-5 TM 1,3,4 Supervised Hydrometrics There are 74 farms irrigated by private irrigation stations, totalling 100 pumps. Irrigation collectives total 53 pumps in 13 pumping stations. All the pumping stations have been described previously (Gindre, 1995). Since the objective of this study is to model the hydrology of the lagoon system with its natural drainage basins, the data is concentrated on the FUM drainage basin (Fig. 1(a)), which has been gauged since 1993 (Chauvelon, 1998). Monitoring of the eight private stations and two collective stations for this basin was conducted from 1993 to 1998, and the complete drainage network was digitized. The Department of Agricultural Services provided the monthly pumped volumes since 1994, which were derived from pump characteristics, gauging tests, duration of pumping and/or energy consumption. RESULTS AND DISCUSSION During the four sampled years, it was observed that the rice cultivated area increased from 5850 ha in 1975 to more than 12 500 ha in 1995 (Fig. 2). Crop rotation with wheat occurs every two to three years in the region north of the delta, where soil salinity is not too important. The specific irrigation volume (SFV) (irrigation volume divided by irrigated area) was calculated for each monitored irrigation basin (five to six years from 1993 to 1998). From this dataset, linear or second order polynomial regressions were calculated between SFV and rice area for each basin. Because of hydraulic constraints and non-optimized water delivery schedules, SIV is negatively correlated to rice area within an irrigation basin (Fig. 3). The range of variation and the minima of SFV were also greater for the collective than for the private irrigation (2100 to 5000 mm, and 1500 to 3500 mm, respectively). From the relationships between SIV and rice area, SIV was estimated for the years before 1993, and the total irrigation input calculated for all irrigation basins (private and collective) of the FUM drainage basin and all collective irrigation basins from polder drainage basins. A method
Quantification of water input for rice cultivation (Rhône delta, France) 449 Fig. 2 Maps of rice field distribution within the Great Camargue in four study years. 1 300 1 400 1 500 1 600 1 700 1 800 Rice area (ha) Fig. 3 Examples of empirical relationships between specific irrigation volume (SIV) and rice area for a private (a) and a collective (b) irrigation basin (the CAZ and PML basins respectively).
450 Philippe Chauvelon et al. to transform irrigation input into drainage output has yet to be completed. After removing évapotranspiration from SIV, we obtain net SIV. The next step will consist of establishing a transfer function from net SIV to drainage output. Acknowledgement This work was funded by the French National Research Programme on Wetlands and the Ricefields, and the Conservation of Nature Programme of Fondation Sansouire. REFERENCES Baumgartner, M. F. & Apfl, G. M. (1996) Remote sensing and geographic information systems. Hydrol. Sci. J. 43(6), 593-607. Chauvelon, P. (1998) A wetland managed for agriculture as an interface between the Rhône river and the Vaccarès lagoon (Camargue, France): transfers of water and nutrients. Hydrobiologia 373/374, 181-191. Gindre, D. (1995) L'irrigation et le drainage en Camargue. Description et fonctionnement des infrastructures hydrauliques. Réflexion pour une gestion globale de l'eau. Rapport polycop. ISARA, Parc Naturel Régional de Camargue, France. Heurteaux, P. (1994) Essai de quantification des termes du bilan hydrique des étangs du système Vaccarès (Camargue, France). Ann. Limnol. 30, 131-144. Rango, A. & Shalaby, A. 1.(1996) Operational applications of remote sensing in hydrology: success, prospects and problems. Hydrol. Sci. J. 43(6), 947-968.