A new satellite remote sensing tool for improving agricultural land use observation

With the global population continuing to grow at a high pace, it is essential to optimize the use of water resources and to increase agricultural production in view of the prospect of having to feed 8 billion humans in 2030. Scientists have for many years been using remote-sensing satellite observations to improve water balance and farming yield assessment on large geographical scales (at the level of irrigated agriculture areas, catchment basins and so on).

Until quite recently, scientists had two different observation methods available for doing that: wide field-of-view sensors (TERRA-MODIS or SPOT-VEGETATION), which allowed daily observation of the entire globe but with a resolution on the kilometric scale, generally extending far over that of one parcel of crops, or decametre-scale-resolution sensors (SPOT, Landsat, ASTER), which can yield only one or two observations per month.

Since 2004, the Taiwanese satellite FORMOSTAT-2 has been in operation, combining the functional features of these two observation techniques, albeit without providing an exhaustive cover of the continents. It gives the possibility for daily observation of small areas of around 500 km² at a spatial resolution of about 8 metres.

Research conducted by an IRD team at the ‘Centre d’Etudes Spatiales de la Biosphère’ at Toulouse, using images taken by FORMOSAT-2, gave the opportunity to study two agricultural areas where farmers make extensive use of irrigation: the Tensift Plain around Marrakech in the centre of Morocco, and the Yaqui Valley in the State of Sonora in North-West Mexico. In these agricultural areas, irrigated cultivation of cereals, fruit trees and vegetables is practised over several thousand square kilometres.

This activity draws on limited water resources, mainly coming from precipitation received by the nearby mountain ranges: the Moroccan High-Atlas in the case of the Tensift Plain, the western Sierra Madre for the Yaqui Valley. Both regions have an arid climate: average rainfall is 200 mm per year. But the water demand is seven times as high (the potential evapotranspiration of the plant cover is about 1500 mm/year). It is therefore essential to portion off the water as equitably as possible, according to the needs of the different types of crops that make up the agricultural landscape.

For these two areas, the study demonstrated all the potential capabilities of the new imagery technique with its high spatio-temporal resolution. With the support of CNES, the satellite FORMOSAT-2 was used to obtain time series of images, from November to May of the following year, that is throughout one entire farming season, at the rate of one shot every 5 days. Processing of the resulting satellite data first helped compile land use maps (including crop rotation and succession patterns) with an excellent degree of discrimination between the different crop types.

The availability of a large set of observations also allowed detailed monitoring of the plant cover with time, achieved by determination of changes in reflectance, the proportion of the light reflected by the land surface. These measurements and associated vegetation indices gave the bases for determining the variables that describe the state of the soil-plant system such as the degree of land plant cover or the green leaf area index: a reliable index for vegetation, a figure below 0.15 corresponding to bare soil whereas one of over 0.70 characterizes a crop that is on the point of ripening (see the series of satellite images).

This detailed and practically continuous description of the plant cover made it possible to improve yield assessments and modelling of water transfer between soil, vegetation and atmosphere. The investigations run in Morocco on parcels of wheat in particular showed that evapotranspiration from the plant cover, the principal factor in water loss, could be evaluated with a margin of error of between 10 and 20%. Moreover, the plant biomass (dry above-ground phytomass) and grain yields could be estimated to an accuracy of about 25% at parcel scale.

In conclusion, this observation technique opens up the possibility to track the various successive operations (sowing, ploughing, irrigation, harvest and so on) effected as an agricultural season progresses. Making sets of such data available for local agencies responsible for agro-environmental management should facilitate the reorientation of irrigation rapidly towards the crop parcels where it is most needed. In addition, it also makes it possible to adapt and improve all the information and advice issued to farmers and thus enable them to make significant improvements to their cultivation practices.

Grégory Fléchet – DIC

1. This research work was conducted as part of the SudMed and MedMex programmes (IRD-CESBIO, Toulouse), in conjunction with the Faculté des Sciences Semlalia of Marrakech, the Office Régional de Mise en Valeur Agricole of Haouz (Morocco) and the University and Technological Institute of Sonora (Mexico). It was financed by the European Commission, Coopération Universitaire Franco-Marocaine and French national space research programmes (INSU and CNES). The FORMOSAT-2 image series were made available by the NSPO and processed in the framework of an agreement between the CNES, the NSPO and SPOT-IMAGE.