Maximizing Agricultural Signal from Gap-Limited SMAP Observations: Does Gap-Free 1-km Soil Moisture Strengthen County-Level Soil Moisture–Yield Coupling Across CONUS?

Soil moisture (SM) is a fundamental driver of crop water stress, yet agricultural monitoring remains constrained by the gap-ridden nature of satellite SM observations. NASA’s Soil Moisture Active Passive (SMAP) satellite has provided global SM retrievals since April 2015, but the 1-km THySM (Thermal Hydraulic disaggregation of Soil Moisture) product—the highest-resolution observation-based SMAP product for Contiguous United States (CONUS)—exhibits 41–99% daily data gaps due to revisit cycles and frozen soil masking. Using DOF_SM (DINEOF+ derived Observation-Faithful Soil Moisture), a gap-free daily 1-km SM dataset reconstructed from THySM with the Data Interpolating Empirical Orthogonal Functions (DINEOF+) algorithm, we present the first CONUS-wide evaluation of whether temporal completeness strengthens the association between SM and county-level yield anomalies of corn, soybeans, and winter wheat. Four SM products—SMAP Level 3 (9-km gapped), SMAP Level 4 (9-km gap-free), THySM (1-km gapped), and DOF_SM (1-km gap-free)—are compared against USDA National Agricultural Statistics Service (NASS) county yield anomalies using county fixed-effects regression with a prespecified phenological window and plant-available water (PAW) feature on a common sample of county-years. Three findings emerge. First, gap-filling weakens the association for corn and soybeans (corn silking Rwithin2 0.014 vs. 0.020 for THySM): DOF_SM restricted to SMAP-observed days reproduces THySM, and window means from reconstructed gap days alone carry a weaker association, so the reduction arises from the reconstructed values. Second, spatial resolution carries most of the SM signal for winter wheat grain fill: coarsening DOF_SM from 1 to 9 km over nearly the same counties reduces Rwithin2 from 0.197 to 0.061. Third, SM–yield coupling is 2.2–5.1× (corn) and 1.5–2.6× (soybeans) stronger in five water-limited rain-fed states than across all rain-fed counties. Finer resolution strengthens county-level SM–yield association only when within-county SM variability reflects landscape-scale moisture gradients rather than field-management noise.