Investigating the Potential of Long Time Series Remote Sensing NDVI datasets for Forest Gross Primary Productivity Estimation over Continental U.S.
Xiaolei Yu, Xulin Guo, Zhaocong Wu

Abstract
Several global remote sensing normalized difference vegetation index (NDVI) datasets have been established for vegetation monitoring and climate change study, including the MODIS NDVI product, the GIMMS and so on. Many researches focused on estimating forest gross primary productivity (GPP) by MODIS data, which is only available from 2000. However, the GIMMS dataset has a long record period from 1981 to 2006, and will be continued to most recent year (GIMMS3g). That is very suitable for long time period GPP monitoring. Nonetheless, there is a lack of comparison between those two different NDVI datasets for forest GPP estimation. As an attempt to deal with that problem, MOD13A1 from MODIS and GIMMS from AVHRR were used in this study to test the potential of NDVI for forest GPP estimation over continental U.S. from the greenness and radiation (GR) model. Meanwhile, three different predictors: NDVI, NDVI*NDVI and NDVI*NDVI*PAR were chosen to accompany with linear, quadratic and exponent functions for GPP estimation at 15 Ameriflux sites. Results showed that both MODIS and AVHRR NDVI datasets have the ability to predict GPP. However, the MODIS data outperforms AVHRR data. Meanwhile, the predictor of NDVI*NDVI*PAR has better explanatory power than others. As for three statistical models, no one is obvious better than others.

Full Text: PDF