Evaluation of an Operational Leaf Area Index Retrieval Approach Using Vegetation and Modis Data

An operational method has been proposed to estimate the leaf area index (LAI) from satellite imagery in the framework of EUMETSAT Satellite Application Facility on Land Surface Analysis (LSA SAF). This study evaluates the performance of the LSA SAF LAI retrieval algorithm when prototyped to VEGETATION/CYCLOPES and MODIS reflectances over Europe for the 2000-2003 period. The results indicate that LSA SAF algorithm retrieves consistent LAI estimates from multiple remotely sensed imagery even when input reflectances present systematic differences. High spatial and temporal consistencies between LSA SAF prototyped LAI and CYCLOPES and MODIS products are found. Differences in LAI between CYCLOPES products and LSA SAF estimates are lower than 0.4 LAI units in terms of RMSE. Larger discrepancies are found when comparing LSA SAF prototyped estimates against MODIS products due, in part, to differences in products assumptions (RMSE ranging from 0.2 up to 0.8 with higher (lower) LSA SAF LAI values compared to MODIS for herbaceous (woody) biomes). Direct validation indicates that LSA SAF prototype estimates achieve similar performances (0.8 and 0.6, respectively) as CYCLOPES and MODIS LAI products. This study constitutes a step forward for the validation and consolidation of the LSA SAF LAI algorithm. INTRODUCTION Leaf area index (LAI), typically defined as the one-sided area of green foliage projected onto a unit area of ground, is a key biophysical parameter of vegetation canopies controlling the exchanges of fluxes of energy, mass (e.g., water and CO2) and momentum between the land surface and the atmosphere. LAI is required for a wide range of environmental applications related to vegetation monitoring, weather prediction and climate change. Satellite remote sensing constitutes the single effective means of deriving continuous and global LAI estimates. Estimation of LAI from remotely sensed optical imagery can be achieved by several operational algorithms which span from simple empirical ones based on calibrated relationships with vegetation indices, up to complex physical algorithms based on the inversion of canopy reflectance models. In Satellite Application Facility on Land Surface Analysis (LSA SAF), a computationally efficient method based on spectral mixing has been developed to estimate the LAI from the EUMETSAT satellites SEVIRI/Meteosat and AVHRR/MetOp. Since 2005, LAI from SEVIRI/Meteosat is routinely estimated by the LSA SAF system (1). But estimating LAI from merged data of MetOp and MSG satellites is foreseen until 2012. Prototyping the LSA SAF algorithm to other sensor data appears to be useful for algorithm validation and improvement. In this study, a prototype of the LSA SAF retrieval algorithm (2) is applied to VEGETATION/SPOT and MODIS/TERRA reflectance data. The prototype estimates are compared to similar LAI products derived from the same sensors. In particular, version 3 of CYCLOPES LAI products (3) and collection 5 of MODIS LAI products (4) are considered. The assessment is achieved over Europe for the period of overlap between CYCLOPES and MODIS products (2000-2003). The main aim of this study is to assess the performance of the LSA SAF prototype estimates compared to the consolidated and validated MODIS and CYCLOPES LAI products, and EARSeL eProceedings 8, 2/2009 181 ground-based maps. Emphasis is also given to the evaluation of the impact of the algorithm and input data on LAI retrieval discrepancies. METHODS LAI algorithm In the LSA SAF system, LAI is estimated from the fractional vegetation cover (FVC) based on the combined use of a novel mixture modelling method (2) and the FVC-LAI relationship proposed by (5):