Validation of global moderate resolution leaf area index (LAI) products over croplands in northeastern China

Over the last decade, a series of global moderate resolution leaf area index (LAI) products have become available and been widely applied in many disciplines. At the same time, there is an increasing demand for the uncertainties associated with these products, which has to be determined through rigorous validation studies. This study validated seven global LAI products - EPS, GEOV2, GLASS, GLOBMAP, MODIS, PROBA-V, and VIIRS - over typical agricultural croplands in northeastern China. Seasonal continuous LAI measurements were obtained from field campaigns in paddy rice fields in 2012 and 2013, and in maize, soybean, and sorghum fields in 2016. High resolution reference LAI maps were first derived from HJ-1, Landsat 7, and Sentinel-2A images with the look-up table (LUT) inversion method and were evaluated with the field measured LAI (R-2 = 0.85 and RMSE = 0.66). Subsequently, the moderate resolution LAI products were validated with the upscaled high resolution reference LAI. All LAI products show typical seasonal variation patterns of agricultural crops, but distinct differences exist among the products. The product quality indicators show large deviations during the peak growing season, whereas the relative uncertainties are higher during the green-up and senescent phases. Both EPS and GLASS show some saturation effects at LAI similar to 4.0 and underestimate the reference LAI (> 0.5), whereas GLOBMAP shows the largest overestimation (bias = 0.96). GEOV2 and PROBA-V significantly overestimate the LAI for all crops. In contrast, MODIS and VIIRS underestimate and show high variations (RMSE >1.50, RRMSE >47%) compared with the reference LAI. In general, the global moderate resolution LAI products show moderate agreement with the reference LAI (RMSE: 0.80-2.0 and RRMSE: 25-60%). The product uncertainties are higher over paddy rice fields than those over the other crop fields. The uncertainties are mainly attributed to the lack of regional tuning of the global algorithms for agricultural crops at different growth stages. Further algorithm improvement and validation studies are necessary to improve the global LAI products for regional applications.