On optimal sampling design for rainfall field

Abstract

This paper reviews the sampling error formalism by North and Nakamoto (1989), North et al. (1991), and Yoo (2000) to provide an idea of optimal sampling design for rainfall field. Rain gauge networks, a line of rain gauges, microwave attenuation measurements, and satellite measurements are evaluated. Combined sampling cases using any two different sampling methods are also evaluated. All cases are applied to the GATE Phase I data, the same as in North and Nakamoto (1989), for the comparison. Summarizing the results is as follows. (1) Without any consideration of heterogeneities like orographic effect or diurnal cycle, the most economical rain gauge network becomes to use evenly spaced 16 gauges or satellite sampling every 24 hours for the sampling of 500km×500km area of GATE. For both cases the sampling error was estimated to be about 10% of standard deviation of areal average values. (2) The sampling error could be decreased significantly for the combined sampling of two different sensors, especially when they satisfy the condition for design orthogonality. For example, the sampling error is decreased up to 50% by combining the evenly spaced 16 rain gauges with satellite sampling every 24 hours.