Importance of soil organic carbon in near-surface soil water content estimation: A simple model comparison in dry-end Canadian Prairie soils

Abstract

The relationship of soil organic carbon (SOC) to soil water content (SWC) is primarily included in hydrological models through porosity where SOC is known to influence soil structure. The Soil Moisture Active Passive Validation Experiment 2012 (SMAPVEX12) field sampling campaign in southern Manitoba provided both SWC and SOC data over three soil textural classes and a range of soil wetness conditions. Linear regressions were developed to predict time-averaged SWC from SOC in the surface 5 cm of 50 cropland fields during SMAPVEX12 when averaged across three wetness ranges. The prediction of SWC from SOC was tested on fields in the same vicinity as SMAPVEX12 that were sampled for SWC in 2008 when soils were very dry. The measured SWC on the five sampling dates in 2008 was compared to SWC estimated for those dates from (1) the single-parameter regression equation from SOC during SMAPVEX12 dry days (1pSOC); (2) a ‘decay’ formula (5% reduction of a declining balance) (2pDM); and (3) a model based on field mean SOC with daily computation (3pSOCM). Both daily iteration models (2 and 3) required initial field capacity and rainfall data from 1 April. Models were compared with root-mean-square error (RMSE) and delta Akaike information criterion (∆AIC). The two-parameter model (2pDM) had the lowest RMSE and most favorable ∆AIC scores for mean SWC of the five sampling dates over all fields. The single-variable SOC equation was equally viable in RMSE and ∆AIC to the other models on sand soils. On clay soils, 3pSOCM was advantageous on some sampling dates, demonstrating a benefit of including SOC with field capacity and precipitation. The predictive ability of SOC to explain time-averaged SWC variability in dry soils would be valuable for SWC model initialization and as a covariate for geostatistical downscaling from satellite SWC scales.