Crop Gross Primary Productivity Estimation using Satellite Data
Gitelson, Anatoly1; Peng, Yi1; Sakamoto, Toshihiro2; Rundquist, Donald1; Verma, Shashi1; Suyker, Andrew1
1University of Nebraska, UNITED STATES; 2National Institute for Agro-Environmental Sciences, JAPAN

In this study, a paradigm was considered to assess gross primary productivity (GPP) in crops via the estimation of total crop chlorophyll (Chl) content. Based on this paradigm, a simple model was developed to estimate crop GPP using a product of Chl-related vegetation index (VI), retrieved from MODIS 250 m and Landsat data, and potential photosynthetically active radiation (PAR). Potential PAR is incident photosynthetically active radiation under a condition of minimal atmospheric aerosol loading. This model is based entirely on satellite data, and it was tested for maize and soybean GPP estimation, which are contrasting crop types different in leaf structures and canopy architectures, under different crop managements and climatic conditions. Using Landsat data, this model was able to accurately estimate GPP in maize-soybean croplands in Mead, Nebraska during growing seasons 2001 through 2008. The indices using green and NIR Landsat bands were found to be the most accurate in GPP estimation with coefficients of variation (CV) below 13% for maize and 15% for soybean. The algorithms established in the Nebraska AmeriFlux sites were validated for the same crops in AmeriFlux sites in Minnesota, Iowa and Illinois. Using MODIS 250 m data, with much higher temporal resolution than Landsat data, the model was capable of estimating GPP accurately in both irrigated and rainfed croplands. Among the MODIS-250 m retrieved indices tested, EVI and WDRVI were the most accurate for GPP estimation with CV below 20% in maize and 25% in soybean. It showed that the developed model was quite sensitive to detect GPP variation in crops where total Chl content is closely tied to seasonal dynamic of GPP. However, the coefficients of model are affected by leaf structure and canopy architecture and, thus, requires re-parameterization when apply for different species. To make it non-species specific, the use of red edge spectral bands is suggested. We present the significance of the red-edge bands of the MSI sensor on Sentinel-2 for monitoring crop physiological status. These narrow MSI spectral bands (15 nm width) are centered at 705 nm and 740 nm, and they have good potential for retrieving GPP, green LAI, chlorophyll and N contents of C3 and C4 crops with no re-parameterization of algorithms. In combination with the high spatial resolution (20 m) and short revisit time, it offers improved applications in fields like precision farming. Similar band positions have been applied for the MERIS sensor on board of ESA’s Envisat satellite (which stopped operation early 2012) and will be available on the upcoming OLCI sensor on the Sentinel-3 satellite constellation. However, both MERIS and OLCI combine a worse spatial resolution (300 m) with a better spectral resolution (about 10 nm), making it applicable at other spatial scales than Sentinel-2 data.