Estimating Optimal Sampling Locations of Surface Soil Moisture at Different Scales Using Various Methods and comparing them for a Lower Himalayan Watershed Region

To compute optimal sampling locations for a lower Himalayan watershed various methods were used. The methods were compared at various scales: watershed scale, landform scale and seasonal scale. The methods compared for the evolution of optimal sampling strategy are statistical sampling, geostatistical sampling, stratified sampling, bootstrap methods and random combination method. To examine the methods, field experiments were conducted in a sampling domain of 425 km² to study the patterns. At the watershed scale total number of 24 locations were evaluated which were distributed into 12 agricultural, 6 forest and 6 grassland landforms. To study the seasonal patterns comparison for the Rabi and Kharif seasons was done. The results indicated that the random combination method provides a simplified and efficient sampling strategy compared to the other methods. Further the random combination method has an inherent advantage of requiring very minimal input information whereas, the statistical and stratified sampling strategy requires data that has to be independent and normally distributed. The geostatistical methods requires a semi variogram model to get the necessary results. To obtain the results at the same level of error the random combination method gives lesser number of sampling locations required. Additionally the computational efficiency of the random combination method can be increased generating smaller groups of samples for the optimality estimation.