Ouaadi N., Jarlan Lionel, Khabba S., Le Page Michel, Chakir Adnane, Er-Raki S., Frison P. L. (2023). Are the C-band backscattering coefficient and interferometric coherence suitable substitutes of NDVI for the monitoring of the FAO-56 crop coefficient ?. Agricultural Water Management, 282, 108276 [16 p.]. ISSN 0378-3774.
Titre du document
Are the C-band backscattering coefficient and interferometric coherence suitable substitutes of NDVI for the monitoring of the FAO-56 crop coefficient ?
Ouaadi N., Jarlan Lionel, Khabba S., Le Page Michel, Chakir Adnane, Er-Raki S., Frison P. L.
Source
Agricultural Water Management, 2023,
282, 108276 [16 p.] ISSN 0378-3774
Rationalizing the use of agricultural water is a key issue in semi-arid areas that face more and more water shortages while food security is already threatened by the increasing population. The FAO-56 approach has been developed to estimate the crop water requirement. It relies on an accurate estimation of the "basal crop coef-ficient" Kcb-act that has been shown to be closely related to NDVI. Nevertheless, optical data can be inoperant in case of persistent cloud cover. Within this context, the objective of this study is to assess the potentiality of the all-weather C-band Sentinel-1 radar observations available with 6-day revisit time at the field scale. To this end, the empirical relationships between Kcb-act, on one hand, and the interferometric coherence at VV (pVV) and VH (pVH) polarizations and the polarization ratio, on the other hand, were assessed on two wheat fields during two crop seasons and compared to the classical Kcb-act-NDVI method. It is demonstrated that while good statistical metrics are obviously obtained between Kcb-act and NDVI derived from Sentinel-2 (R = 0.77/0.88 and RMSE = 0.14/0.15 for Field 1/Field 2), close results are highlighted with radar data. The best metrics are found with pVV: R = 0.76 and 0.77 and RMSE = 0.18 and 0.28 for Field 1 and Field 2, respectively. Using the calibrated re-lationships on one season of Field 1, reasonable estimates of ETc-act was found on Field 1 (R = 0.70, RMSE = 0.75 mm/day and bias =-0.18 mm/days using Kcb-act-pVV). By contrast, a significant overestimations is high-lighted both with pVV (bias = 0.73 mm/day) and NDVI (bias = 1.46 mm/day) over Field 2. Interestingly, the Kcb-act-pVV relationship is more consistent in the estimation of ETc-act when changing from one plot to another. These outcomes open new perspectives for the estimation of ETc-act from radar data as a potential substitute of NDVI in case of persistent cloud cover.
