| Popis: |
The availability of accurate information on the water consumed for crop irrigation is of vital importance to support compatible and sustainable environmental policies in arid and semi-arid regions. This has promoted several studies about the use of remote sensing data to monitor irrigated croplands, which are mostly based on statistical classification and/or regression techniques. The current paper proposes a new semi-empirical approach that relies on a water balance logic and does not require local tuning. The method stems from recent investigations which demonstrated the possibility of combining standard meteorological data and Sentinel-2 (S-2) Multi Spectral Instrument (MSI) NDVI images to estimate the actual evapotranspiration (ETa) of irrigated Mediterranean croplands. This ETa estimation method is adapted to drive a simplified site water balance which, for each 10-m S-2 MSI pixel, predicts the irrigation water (IW), i.e. the water which is consumed in addition to that naturally supplied by rainfall. The new method, fed with ground and satellite data from two years (2018–2019), is tested in a Mediterranean area around the town of Grosseto (Central Italy), that is covered by a particularly complex mosaic of rainfed and irrigated crops. The results obtained are first assessed qualitatively for some fields grown with known winter, spring and summer crops. Next, the IW estimates are evaluated quantitatively versus ground measurements taken over two irrigated fields, the first grown with processing tomato in 2018 and the second with early corn in 2019. Finally, the IW estimates are statistically analyzed against various datasets informative on local agricultural practices in the two years. All these analyses indicate that the proposed method is capable of predicting both the intensity and timing of the IW supply in the study area. The method, in fact, correctly identifies rainfed and irrigated crops and, in the latter case, accurately predicts the IW actually supplied. The results of the quantitative tests performed on tomato and corn show that over 50 % and 70 % of the measured IW variance is explained on daily and weekly bases, respectively, with corresponding mean bias errors below 0.3 mm/day and 2.0 mm/week. Similar indications are produced by the qualitative tests; reasonable IW estimates are obtained for all winter, springs and summer crops grown in the study area during 2018 and 2019. |
Full Text from ScienceDirect