Indian scientists have developed a mathematical equation that they say can be used to forecast near-accurate crop yield. They claim their generic equation can predict the state-wise crop yield with better than 90 per cent accuracy and help calculate crop yields in other countries as well.
The researchers - Ramesh Singh, Anup K Prasad and Vinod Tare of the Indian Institute of Technology (IIT) Kanpur along with US researcher Menas Kafatos of the Center for Earth Observing and Space Research in Fairfax, Virginia - have used long-term satellite and meteorological data for their equation.
"We used 20 years of satellite-derived surface and meteorological data to develop this prediction model for the first time," Ramesh Singh, who works at the civil engineering department at IIT, said.
Singh and his co-workers have described their model in a paper that is to appear in the International Journal of Remote Sensing.
"Our equation is giving very good results," Singh said. "The Indian ministry of agriculture has been looking precisely for such a model," he said.
For India, the "yield equation" has been developed for rice and wheat.
The rice and wheat yields predicted by their model were close to the observed values for three consecutive years - 1997, 1998 and 1999 - tested, said Singh.
Because of large variations in the climatic conditions and agricultural practices in different parts of India, separate equations have been developed for each state for estimating rice and wheat yield, Singh said.
He said the methodology developed by his team for rice and wheat "can be used to derive such empirical equations for other crops as well".
The equation can be modified and extended to other countries where crop production is "primarily dependent" on weather and climatic conditions. "In fact, we have tested this model for corn and soybean in the Iowa state of the United States," Singh said.
The conventional approach of forecasting crop yield using ground-based data collection is tedious, time consuming and often a difficult job. Remote sensing satellites have made this job easier, the IIT scientist said.
According to Singh, their approach assumes that crop yield mainly depends on four key factors that vary year to year from region to region - soil moisture, surface temperature, rainfall and another parameter called "normalised difference vegetation index or NDVI".
These four parameters are routinely collected by American earth observation satellites for the entire globe and are available in archives. What Singh's team has done is to analyse 20-year satellite data and use the average seasonal value for each of the four parameters to develop their crop yield equation specific for every unit (50x50 km area) in crop growing regions in India.
To make a prediction for a particular year, the satellite data for that is plugged into the equation and the estimates for each unit are added up to get the total yield for the country as a whole.
The yield equation takes into account only the variations in surface and meteorological data but not factors like pests, plant diseases, or farmers switching to new hybrid crop varieties.
Under these other circumstances the forecast could go wrong, Singh admitted. "This is a limitation to any forecasting method."
The IIT scientists recommend that the government build a database of vegetation index, soil moisture, surface temperature and rainfall in 50x50 km grid for all crop growing regions in the country and execute the crop prediction model separately for each "window".
"Grid size from suggested 50x50 km may be further reduced to a higher resolution for a more realistic estimate of crop yield," they suggested.