Composite estimation to combine spatially overlapping environmental monitoring surveys

Long-term environmental monitoring surveys are designed to achieve a desired precision (measured by variance) of resource conditions based on natural variability information. Over time, increases in resource variability and in data use to address issues focused on small areas with limited sample sizes require bolstering of attainable precision. It is often prohibitive to do this by increasing sampling effort. In cases with spatially overlapping monitoring surveys, composite estimation offers a statistical way to obtain a precision-weighted combination of survey estimates to provide improved population estimates (more accurate) with improved precisions (lower variances). We present a composite estimator for overlapping surveys, a summary of compositing procedures, and a case study to illustrate the procedures and benefits of composite estimation. The study uses the two terrestrial monitoring surveys administered by the Bureau of Land Management (BLM) that entirely overlap. Using 2015–18 data and 13 land-health indicators, we obtained and compared survey and composite indicator estimates of percent area meeting land-health standards for sagebrush communities in Wyoming’s Greater Sage-Grouse (Centrocercus urophasianus) Core and NonCore conservation areas on BLM-managed lands. We statistically assessed differences in indicator estimates between the conservation areas using composite estimates and estimates of the two surveys individually. We found composite variance to be about six to 24 units lower than 37% of the survey variances and composite estimates to differ by about six to 10 percentage points from six survey estimates. The composite improvements resulted in finding 11 indicators to statistically differ (p <0.05) between the conservation areas compared to only six and seven indicators for the individual surveys. Overall, we found composite estimation to be an efficient and useful option for improving environmental monitoring information where two surveys entirely overlap and suggest how this estimation method could be beneficial where environmental surveys partially overlap and in small area applications.