Uncertainties of LAI estimation from satellite imaging due to atmospheric correction

Abstract

Leaf area index (LAI) is a plant development indicator that as an input parameter strongly influences several relevant hydrological processes represented in Soil–Vegetation–Atmosphere-Transfer (SVAT) models. Generally, temporal measurement or monitoring of LAI is challenging or even impossible in remote areas. High-temporal resolution remote sensing imaging can be used to estimate LAI from vegetation indices calculated from band ratios. This paper shows the sensitivity of LAI estimation from satellite imaging to atmospheric correction (with ATCOR) and evaluates the effects of LAI uncertainty on water balance modelling. LAI as a SVAT model input parameter was estimated based on the empirical relationship between field measurements, and the vegetation indices NDVI (Normalized-Difference Vegetation Index), SAVI (Soil-Adjusted Vegetation Index) and SARVI (Soil–Atmosphere Resistant Vegetation Index) for six RapidEye images obtained between 2011 and 2012. In summary, we found that the ATCOR parameter ‘visibility’ has the strongest influence on LAI estimation. Likewise, atmospherically corrected successive images gathered from around the same time period had low LAI differences (mean absolute difference of 0.09 ± 0.08) on overlapping image areas. This uncertainty is negligible in SVAT modelling in most cases, thereby allowing mosaicked successive atmospherically corrected images to be used. We showed that LAI uncertainties arising from atmospheric correction (ATCOR 3) can translate into small (LAI ± 0.1 ≈ evapotranspiration ± 0.9%, interception ± 2.5%, evaporation ± 3.3%, transpiration ± 0.7%) to moderate (LAI ± 0.3 ≈ evapotranspiration ± 4.1%, interception ± 7.5%, evaporation ± 9.9%, transpiration ± 2.4%) SVAT model uncertainty.