Improving Global Simulation of Irrigation within the Land Information System

Irrigated agriculture increases evapotranspiration and near surface humidity, decreases air and 31 ground temperatures, augments downwind precipitation, enhances recharge, and depletes surface 32 water and/or groundwater resources. However, model developers have been slow to incorporate 33 irrigation. Using the latest satellite-based land cover and irrigation datasets and innovative 34 modeling approaches, we developed a scheme for simulating irrigation in the Land Information 35 System (LIS). We reconciled global and national irrigation, land cover, and crop class databases 36 in an effort to (1) distinguish irrigated and rainfed agriculture, (2) approximate fractional crop 37 cover in each model grid pixel, and (3) apply the most likely irrigation method based on the crop 38 class and country or state level information on the relative usage of sprinkler, furrow, paddy, and 39 drip irrigation. Each irrigation method is represented by a separate algorithm. We used 40 agricultural water withdrawal data, as available, for calibration and evaluation. The results 41 compare favorably with available observational and reported data in terms of patterns. The 42 simulated mean annual global irrigation rate, 2765 km3/yr, is on par with an observational 43 estimate (2500 km3/yr) and other irrigation simulating models (mean 2700 km3/yr). Over the 44 contiguous U.S., based on county level reported irrigation water usage data, the new scheme had 45 a root mean square error of 0.22 km3 while correctly simulating the occurrence of irrigation in 46 3053 of 3220 irrigated counties, compared with 0.24 km3/yr and 2334 for LIS’s original scheme.