Retrieval of Leaf area index and stress conditions for Sundarban mangroves using Sentinel-2 data

Abstract

ABSTRACTThe potential of Sentinel-2 (S2) data in mapping Leaf area index (LAI) of mangroves having heterogeneous species composition, variable canopy density, and complex backgrounds was studied. Out of the three available near-infrared bands in S2, band-8 of 10 m spatial resolution was found to be the most suitable one for deriving the Normalized Difference Vegetation Index (NDVI) for mangroves. The LAI-NDVI relation did not accord apparently with the earlier reports and the underlying complex background effect was validated with Airborne Visible Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) hyperspectral data. It simulated spectral and spatial conditions of S2 by linear mixing of canopy and background that confirmed the effect of background contributions to the canopy reflectance decorrelating the NDVI from LAI. The compensation for diverse backgrounds was accomplished with optimum-scaled NDVI (scNDVIm) obtained from the mean of scaled NDVIs derived with different backgrounds in the mangroves. LAI was well correlated with composite NDVI (NDVIcom), derived empirically from the most appropriate NDVI (NDVIS2) and scNDVIm where ground observation controlled the threshold arbitration in extracting the range of scNDVIm. It was shown that an improved LAI estimate with a coefficient of determination (R2) of 0.69 and root-mean-square error (RMSE) of 0.02 could be obtained with NDVIcom. This method has the advantage of compensating the contaminations due to background reflectance. While the relation between LAI and NDVIcom was found to be consistent, the application of the same methodology in similar mangroves should be site-specific with ample ground observation. The fusion of NDVI and scNDVI obtained from S2 yields better LAI retrieval for mixed mangroves, such as that of Sundarban.