Abstract
Improvement in agricultural productivity has become a necessity due to the limitation in expansion of cultivated acreage and ever increasing food demand. To forecast the growth of agricultural productivity, crop production data aggregated by administrative zone such as states and nations along are not sufficient because it is necessary to take into account environmental and land conditions specific to each location. With this background, this paper discusses a method to estimate the distribution of agricultural productivity as case study of India. The distribution of NPP for each grid cell (8km resolution) was estimated using the NDVI data derived from NOAA AVHRR and PAR. NPP in crop land was extracted using the land use data and has been summed up by each districts. Correlation of the agricultural productivity and NPP were investigated by comparing statistical crop production data on each district. As the result, it was found estimated from remote sensing data is effective to complement data of crop production. Finally the grid-level agricultural productivity was estimated using NPP data.
Introduction
Food demand is foreseen to continue increasing along with the increase in the world population. However, an expansion of cultivated acreage can not be expected because of the limitations arising from environment conservation, especially, forest conservation. Therefore, an improvement in farm productivity has become a necessary for coming days.
The growth of the agricultural productivity depends on the land conditions, for example availability of irrigation water resources, soil deterioration, in addition to an improvement of agricultural technology. Mechanization is also influenced by the topography. Whereas, increase in productivity by applying fertilizer are limited due to their side effect of environmental loading. Like this, to foreseen the growth of the agricultural productivity, various environmental factors must be considered, and grasping on detailed space distribution of the crop production becomes indispensable. Moreover, monitoring of the farming area in cultivating stage is effective, because the amount of harvest is insufficient. So remote sensing technology can be very effective for spatio-temporal supplementation.
The purpose of this study is to develop a basic method to estimate distribution of the agricultural productivity using remote sensing data, gives an approach to a fundamental research. India has been selected as study area, because the ratio of the crop land is relatively higher and availability of statistical data. To do this, NPP distribution (8km resolution grid cell) was estimated using NDVI of NOAA AVHRR and PAR data derived from TOMS data set. Next, NPP on crop land (named "crop land NPP") was extracted using the land use data, and totaled in each district. The distribution of the agricultural productivity was estimated, by corresponding with crop land NPP and the statistical data of crop production.
Estimation of NPP distribution using NDVI
NPP distribution was estimated with a method proposed by Hirakoba (1998) based on Production Efficiency Approach. NPP is estimated from solar radiation, absorbed by a plant, and the efficiency index (e) which is the ratio of radiation energy changed into the organic substances.
The radiation energy in wavelength which a plant can use for the photosynthesis is called PAR (Photosynthetically Active Radiation) and data set of PAR was made using the data on TOMS (Dye, 1993). The ratio of PAR absorbed by the canopy of the plant is named fapar (the fraction of PAR), and has been estimated using NDVI of NOAA AVHRR. Solar radiation absorbed by a plant is named APAR (Absorbed PAR) and have been calculated by multiplying PAR and fapar . Therefore, APAR can be calculated from the data set of PAR and NDVI.
In this study, using 8km resolution monthly NDVI data derived from NOAA AVHRR data (figure 1), APAR was estimated. Moreover, monthly NPP was estimated using APAR multiplied by efficiency index (e) which was assumed 1.5 g/MJ, and annual accumulation value was calculated. The NPP distribution estimated in this study is shown in figure 2. Crop land NPP was extracted from NPP distribution using the land use data (figure 3).
Figure 1: Monthly NDVI (example of March, June, September, December)
Figure 2: NPP distribution (annual, 1990)
Figure 3: Crop land NPP distribution
NDVI data for the year 1990 was used to match with the statistical data. PAR data available with us was only for the year 1979-1989, therefore monthly mean value of PAR data in this period was calculated and used.
