Citrus orchard mapping in Juybar, Iran: Analysis of NDVI time series and feature fusion of multi-source satellite imageries

Nowadays crop mapping as an interdisciplinary hot topic attracted both agriculture and remote sensing researchers' interests. This study proposed automatic method to map citrus orchards in Juybar, Iran, where planting trees is booming there. In this regard, 148 Sentinel-1, Sentinel-2, ALOS Digital Surface Model (DSM) tiles are processed Google Earth Engine provide a hybrid feature set including initial satellite images, Gray Level Co-occurrence Matrix (GLCM) textural features, spectral features such vegetation, built-up, bare-soil indices, the Vegetation Dynamic Index (VDI). A semi-automatic sample selection paradigm also developed based on time-series analysis of 12 monthly Normalized Difference Indices (NDVIs), Otsu thresholding, multi-level thresholding (MLT), using two indices called Evergreenness (EGI) Water-covered or No-vegetation (WCNV) index, finally human post-revision. The output Support Vector Machine (SVM) classification 60,000 samples post-classification operation showed that classified has average overall accuracy (OA) kappa coefficient (KC) equal 99.7% 0.992, respectively. results show multispectral bands lonely extracted with high (OA: 99.55%, KC: 0.986), rest only made slight improvement that. For year 2019, area about 4351 ha estimated orchards, which accordance department's reports (~4700 ha). indicate from 2016 despite “citrus non-citrus” land-use conversion 754 ha, was totally expanded by 17%. Comparing images indicates precise extraction 10 m spatial resolution. To use for extensive cases areas other types evergreen trees, it recommended high-resolution normalized DSMs (nDSMs) features. • aimed Juybar (Iran) automatically. Ever greenness key up potential masks. pointwise have better performance than manual polygon method. Multi-modal increased classification. croplands were detected