Model. Each 1-km pixel is a simple water balance of the top 50 cm of soil, tracked as the fraction of plant-available water between the wilting point and field capacity. Daily PRISM rainfall, less 5 mm of canopy and litter interception, fills the profile; a seasonal drying rate then draws it down. That rate was derived from the drydowns recorded by the Kansas Mesonet soil moisture network on rain-free days: about 0.5% of the available water per day in January and about 4% per day in July.

Soils. The wilting point and field capacity of each pixel are estimated from SSURGO clay, sand and bulk density with a pedotransfer function. Compared with the limits the stations themselves reach, this gives an error of roughly 20 mm; about half of the remaining station-to-station differences are local and cannot be resolved by any soil map.

Anchoring to the network. Every day the map is compared with the stations. The mismatch of the last three days is interpolated into a smooth surface, strongest near stations and fading to nothing far from them, and 30% of it is applied to the map. This relaxes the map toward the measurements over a few days, removing persistent bias without copying day-to-day sensor noise. Sensors in frozen soil are ignored. Against the network the map has a median correlation of 0.89 and an unbiased error of 0.12.

Reading the map. 0 is wilting point, 1 is field capacity. Station markers show the sensor reading on the same colour scale. The statewide chart compares the spatial mean of the map with the network mean, shown only when at least 30 stations report. Water stored is the sum over the state of the water held in the top 50 cm, in cubic kilometres.