Identifying Outliers of the MODIS Leaf Area Index Data by Including Temporal Patterns in Post-Processing

Abstract

The moderate resolution imaging spectroradiometer (MODIS) calculates the leaf area index (LAI) for each pixel without incorporating the temporal correlation information, leading to a higher sensitivity for the LAI that produces uncertainties in observed reflectance. As a result, an increased noise level is observed in the timeseries, making the data discontinuous and inconsistent in space and time. Therefore, it is important to identify and handle the outliers during the post-processing of MODIS data. This study proposed a method to identify the MODIS LAI outliers based on the analyses of temporal patterns, including the interannual and seasonal changes in the LAI. The analysis was carried out utilizing the data from 278 global MODIS LAI sites and the results were verified against the measurement obtained from 52 ground stations. The results from the analyses detected 50 and 92 outliers based on 1.5σ and 1.0σ standard deviations, respectively, of the difference between the MODIS LAI and ground measurements; correspondingly, 46 and 65 outliers, respectively, were identified by incorporating temporal patterns during the post-processing of the data. The validation results exhibited improved values of the coefficient of determination (R2) after eliminating the MODIS LAI outliers identified through the interannual and seasonal patterns. Specifically, the R2 between the ground measurement LAI and MODIS LAI increased from 0.51 to 0.54, 0.88, and 0.90 after eliminating MODIS LAI outliers when considering the interannual patterns, seasonal patterns, and both the interannual and seasonal patterns, respectively. The results from the study provided valuable information and theoretical support to improve MODIS LAI post-processing.