Sampling Design for Global Scale Mapping and Monitoring of Agriculture

Gathering timely information of the global agriculture production of major and commercially important crops has become essential with globalization of the agriculture commodities. Remote sensing based crop production forecasting and monitoring is emerging as one of the most viable solutions for such large area monitoring task. A suitable sampling strategy is the basic requirement towards this. In the present study, different sampling sizes using agricultural area as the sampling frame has been used to analyse the optimum sampling size for continent level assessment. Land use/cover map of the world using 300 m resolution MERIS data was used to generate the agriculture area mask. Grid size of (i) 5° × 5° (ii) 1° × 1° (iii) 30′ × 30′ (iv) 15′ × 15′ (v) 7.5′ × 7.5′ and (vi) 5′ × 5′ were used. Percent crop area was estimated for the grids of all sizes. The grid size of 15′ × 15′ was found to be optimum for global monitoring, as not much change Ws observed in the distribution of the grids after reducing the sample size. Stratification was done using simple random and stratified random sampling method. Stratification using the ‘cumulative square-root of frequency method that resulted in five strata performed best in terms of the variance of the population.