FLI : a new spectral index to characterize flooding irrigation

Flooding irrigation is a method still widely used in certain farming systems and in foothill areas. Although this traditional technique remains water-consuming, it offers significant external benefits such as groundwater recharge and biodiversity preservation. For example, on the Crau area (600 km²) in the south of France, flooding irrigation of grasslands contributes to 70% of the total recharge of the aquifer, which is strategic for a large number of human activities (irrigation of orchards, drinking water, industry). Remote sensing makes it possible to characterise the area of irrigated grasslands with a high degree of accuracy (Abubakar et al., 2022). However, the number of water cycles, which is determined by meteorological conditions and possible restrictions, remains poorly characterized as well as the irrigation dose, which depends on the length of the plot along the water flow axis. There is therefore a challenge in detecting irrigation events and the direction of flow.

In order to obtain a territorial view of flooding irrigation on grasslands, the objective of this study is to use high spatial resolution (~10m) remote sensing to characterize irrigation patterns (irrigation period, frequency and dose) at the plot scale.  A previous study (Bazzi et al., 2020) based on radar imagery shows that it is possible to detect flooding irrigation, but there are still many errors, mainly when vegetation is dense. In the present study, analysis of Sentinel 1 time series in both radar configurations did not show a clear signal of irrigation. The dense vegetation of the grasslands probably masks the water layer during irrigation or the wet soil after drying. On the other hand, plots undergoing irrigation appear clearly when Sentinel 2 measurements are placed in a diagram relating the reflectance in the SWIR -band 11 (RSWIR) and the NDVI. Plots undergoing irrigation have RSWIR that deviates from the RSSWIR=f(NDVI) relationship. The distance from the RSWIR/NDVI point to this relationship can therefore be used to identify flooded pixels. With adequate thresholding of this distance, it can be shown that the plots identified as being irrigated are indeed irrigated in more than 90% of cases. Intra-plot mapping of irrigated areas makes it possible to identify the direction of irrigation and some times the direction of flow, which makes it possible to specify the water amount applied and, consequently, the amount drained, contributing to groundwater recharge. Temporal analysis of the territory allows the identification of the start and end of irrigation period. The proposed method thus makes it possible to sample a large number of irrigation events and thus enable more realistic modelling of irrigation schedules.

Abubakar, M., Chanzy, A., Pouget, G., Flamain, F., Courault, D., 2022. Detection of Irrigated Permanent Grasslands with Sentinel-2 Based on Temporal Patterns of the Leaf Area Index (LAI). Remote Sensing 14, 3056. https://doi.org/10.3390/rs14133056

Bazzi, H., Baghdadi, N., Fayad, I., Charron, F., Zribi, M., Belhouchette, H., 2020. Irrigation Events Detection over Intensively Irrigated Grassland Plots Using Sentinel-1 Data. Remote Sensing 12, 4058. https://doi.org/10.3390/rs12244058