Evaluation of topographic effects on multiscale leaf area index estimation using remotely sensed observations from multiple sensors

Local topography significantly affects remotely sensed reflectance data and subsequently impacts leaf area index (LAI) retrieval over mountainous areas. Therefore, mountain vegetation LAI mapping from satellite observations at multiple scale levels is often obstructed by topographic distortion. To analyze the effects of topography on multiresolution LAI retrievals, consistent LAI estimations were first generated across six spatial scales (i.e., 960 m, 480 m, 240 m, 120 m, 60 m and 30 m) from MODIS and Landsat OLI reflectance data using the ensemble multiscale filter (EnMsF) approach over rugged surfaces. Subsequently, the topographic influence on LAI was evaluated based on spatial patterns and retrieval accuracies at multiple scale levels by comparing the EnMsF-based multiscale LAI results obtained before and after terrain correction of a Landsat image. The results demonstrated that the multiresolution LAI values retrieved from topographically corrected surface reflectance data outperformed those without topographic correction, regardless of various slopes or aspects at the 30 m scale or the differences in spatial resolutions. The accuracies were determined for the retrieved LAI values before (coefficient of determination (R-2) = 0.32 and root mean square error (RMSE) = 1.03) and after (R-2 = 0.60 and RMSE = 0.82) topographic correction when compared to the field measurements over slopes facing toward the sun at the 30 m resolution. The R-2 and RMSE values were 0.57 and 0.54, respectively, for the LAI estimations with terrain corrections in the shady aspects. Moreover, the topographic effect on the LAI estimations depended on our spatial scales. The finer the spatial resolution is, the more significant the topographic effects on the multiscale LAI retrieval are, and vice versa. Thus, the results of this study open an encouraging path to deepen the understanding of terrain effects on multiscale LAI estimations, and further improve the LAI retrieval algorithm across different spatial scales over mountainous areas.