Problems Modeling Soil Moisture: Calculating Rainfall

All attempts to model soil moisture or drought condition use observed NEXRAD precipitation as inputs, but the raw data must first be corrected.

East of the Continental Divide, radar imagery is compared to ground rain gauges and a correction factor is calculated. In mountainous areas West of the Continental Divide, a different method is used to derive the "observed precipitation".  Ground rain gauge data is compared to average precipitation data (PRISM) and departures from that average are interpolated between gauge locations.  The end result is 4km resolution rainfall totals.  

Once observed precipitation is calculated, accumulated precipitation can be viewed for any time period using NOAA's website.

Clearly, 4km by 4km grid cells hide a great deal of local variability.  For example, the Rainlog network of rain gauges in Tucson records highly variable rainfall at locations less than a kilometer apart during the monsoons (W Miracle Mile is about 2 km):

Methodological problems in the way PRISM fills gaps using modeled historical data may bias against extreme or unusual rainfall patterns.  Also, numerous sources of bias in both the radar and the rain gauges have to be accounted for manually.  For example, radar can be biased by hail, angle, and artifacts created by birds and insects.  Rain gauges can also malfunction in an endless variety of ways, including sensor error, human error, and when ice and snow block the gauge.  These uncertainties in observed precipitation can jeopardize efforts to model soil moisture such as the PDI.  Also, they call into question research that has revealed an increase in extreme precipitation events.

Existing large-scale methods of modelling soil moisture are unconstrained by field measurements, so the advent of satellites offering weekly global measurements of soil moisture are an important step forward.  These satellites (such as SMOS) can image vast swaths of the Earth’s surface to infer average soil moisture at the surface, but this imagery has an accuracy of +/- 4% soil moisture over pixels that are 35-50 km on a side.  A new satellite launched this year (SMAP) has better resolution, approximately 9 km, but still nowhere near field-scale resolution.  Local hill slope, vegetation land cover, and soil texture differences mean that county-level averages aren't accurate enough to apply on individual acres.

There are some companies that claim to be able to remotely monitor acre-by-acre soil moisture for farmers, but that is not possible without field measurements.