Abstract
The increasing development pressures from urbanization, aquaculture and tourism on worldwide coastal environments and the ecosystem services that mangroves provide make it essential to monitor and manage these environments more effectively. Measuring and monitoring mangrove structure through variables like Leaf Area Index (LAI) is an essential part of this action. This study investigated the effects of different mangrove environmental settings, satellite image spatial resolutions, spectral vegetation indices (SVIs) and mapping approaches for LAI estimation. We compared and contrasted the ability of WorldView-2 (WV-2), ALOS AVNIR-2 (AVNIR-2) and Landsat TM (TM) image data (2 m, 10 m and 30 m pixel sizes, respectively), to estimate LAI through regression analysis at sites in Moreton Bay (Australia) and Karimunjawa Island (Indonesia). We also investigated the effect of different pixel averaging windows (3 x 3, 5 x 5, and 7 x 7 pixels) and multi-resolution segmentation scale parameters (10, 20, 30, 40 and 50) applied to the WV-2 image for LAI estimation. The results showed that LAI estimation using remote sensing data varies across sites and sensors. Estimation of LAI in this study was influenced by the local spatial variation of mangrove phenological stages and canopy cover. The regression analyses showed significant coefficient of determination (R-2) values ranging from 0.50 to 0.83 across different sensors (TM, AVNIR-2, WV-2), segmentation scales (10, 20,30, 40, 50) and SVIs (SR, NDVI, SAVI, EVI). The sensor and SVIs assessment identified the ALOS AVNIR-2 and NDVI as the optimal estimators of LAI, with R-2 = 0.83, RMSE = 0.54 for Moreton Bay, and R-2 = 0.82, RMSE = 1.31 for Karimunjawa Island. The optimum image pixel size for estimating LAI was related to the average canopy size (about 10 m in diameter) and the field sampling size (10 m). Image segmentation significantly increased the LAI estimation accuracy by approximately 14% for both sites. The findings of this study provide an understanding of the relationship between image spatial resolution, field sampling size and spatial variation of mangrove vegetation for estimating LAI. These findings can be potentially used as a guide for selecting the optimum imagery for LAI estimation. (C) 2016 Elsevier Inc. All rights reserved.