How Universal is the Relationship between Remotely Sensed Vegetation Indices and Crop Leaf Area Index? A Global Assessment

This study aims to assess the relationship between Leaf Area Index (LAI) and remotely sensed Vegetation Indices (VIs) for major crops, based on a globally explicit dataset of in situ LAI measurements over a significant set of locations. We used a total of 1394 LAI measurements from 29 sites spanning 4 continents and covering 15 crop types with corresponding Landsat satellite images. Best-fit functions for the LAI-VI relationships were generated and assessed in terms of crop type, vegetation index, level of radiometric/atmospheric processing, method of LAI measurement, as well as the time difference between LAI measurements and satellite overpass. These global LAI-VI relationships were evaluated on the variation of coefficient values using a site-based approach and validated at three different spatial scales by comparing LAI estimates from global LAI-VI relationships with independent in-situ measurements and other reference datasets. The main findings suggest that, on a global scale, vegetation indices derived from remote sensing can provide reasonable LAI estimates for crops on an average basis and explain more than half of the total variance using three or less regression coefficients without any prior knowledge about the land surface. The global LAI-VI relationships are mostly non-linear and appear to be the strongest when it is crop-specific and is based on radiometrically corrected satellite data. Site-based evaluation and validation processes both indicate that the global LAI-VI relationships are stable over varied locations, but might induce systematic errors/differences if applied to individual local-scale study locations. To this end, we find some degree of global universality that would allow for the generation of spatially and temporally explicit LAI maps from remotely sensed observations, despite the heterogeneity of land surface in which they are applied. As LAI provides the key link between remotely sensed observations and various agro-ecosystem processes, the findings also contribute to the operationalization of large-area crop modeling and, by extension, aid in various agroecosystem research and application questions.