To estimate crop yields, we use an LUE-based crop growth model similar to those used in a number of studies (Boschetti et al. 2011; Hashimoto et al. 2013; Kumar and Monteith 1981; Peng et al. 2014; Running et al. 2004). The model computes daily above-ground biomass accumulation similar to gross primary production (GPP) in MOD17, the growing season sum of which forms the basis for assessing crop yields. The model accounts for sub-optimal conditions of temperature, humidity, and soil moisture throughout the season, along with levels of cloudiness to estimate AGB (above-ground biomass (g/m2)).
where ABG is above-ground biomass (g/m2); PAR is incident photosynthetically active radiation (MJ/m2); FPAR is the fraction of absorbed PAR, R is the fraction of biomass allocated to above-ground; is the actual LUE (g of dry biomass per MJ), determined as follows.
where maxis the maximum LUE for a given crop; Kv is a scaling factor varying from 0– 1.0 depending on the sensitivity of photosynthesis to humidity deficits; Kt is a scaling factor varying from 0 to 1.0 depending on the sensitivity of photosynthesis to daily temperature variations; Kw is a scaling factor varying from 0 to 1.0 depending on the
sensitivity of photosynthesis to soil moisture conditions. For further details on the downregulation scalars, please see the section on weather data and its use in the study. Crop yield is then derived according to:
where HI is the harvest index that determines the fraction of above-ground biomass that results in harvestable products and M is the moisture content in harvested grain (fraction).
For conducting the CCEs, we used an intelligent sampling scheme that selected 40 GPs in each of the 25 districts. We used satellite data at the peak growing season to identify three categories of crop performance: low, medium and high. We chose 15 GPs each in low and medium and 10 GPs in high performance category. In each of the GPs we conducted 4 CCEs, totalling 160 CCEs in each district.