Field Monitoring Insights From Statistical Evaluation With High-Resolution Remote Sensing on a California Annual Rangeland

An optimized sampling strategy is critical for effective environmental monitoring, bridging between local data collection over limited sites and inference over a broader study area, especially for those environmental attributes subject to large spatial and temporal variations.This study leveraged a unique 10-ha rangeland dataset from the Central California Coast to determine the optimal sample size required to quantify peak standing crop (PSC) at given levels of accuracy and statistical significance. Imagery from drone and satellite were fused and calibrated with field clipping data to map PSC at a 0.3 × 0.3 m resolution. We assessed the spatial and temporal variability in PSC to characterize its true population and landscape variability for both a wet and dry year. The mean PSC was 2.8-fold higher in the wet year (2 793 kg · ha −1 ) than in the dry year (983 kg · ha −1 ), whereas the coefficient of variability among surrounding quadrats was 50% higher in the dry year and showed a logarithmic decrease with increasing PSC levels indicating the need for a greater number of samples in years having lower production. We found that 10 to 12 randomly selected samples provided accuracies within ±20% of the catchment true mean at a confidence level of 95% in the drier year, whereas only five samples were required in the wetter year. Furthermore, the use of the 0.3 × 0.3 m versus 1 × 1 m quadrat achieved similar sample-size requirements, whereas providing a ∼10-fold savings in field work time/effort. These findings provide data-driven guidance on the optimized field sampling for accurate PSC monitoring and assessment to guide policy and management decisions. The statistical evaluation framework leveraging very high-resolution remote sensing can be applied to other systems for the sampling strategy optimization to capture the true spatial or temporal variability.