R‐Learner Data Integration for Estimating Population Trends From Opportunistic and Structured Survey Data

Accurately quantifying wildlife population change is essential for assessing ecosystem health and guiding conservation action. Model‐based integration of structured and opportunistic survey data can expand the spatial scope and improve the precision of trend estimates but requires methods that reconcile differences in observation processes, sampling coverage, and data volume across surveys. We present the integrated R‐learner trend model, a modular framework for estimating spatially explicit trends from multiple surveys. In Stage 1, the double machine learning (DML) trend model is applied separately to each survey, isolating ecological signals from intra‐ and interannual confounding. In Stage 2, survey‐specific trends are combined using the R‐learner framework, with weights based on sampling coverage probabilities estimated via a spatially explicit mixture‐distribution sub‐model. This two‐stage design enables bias mitigation, reconciles differences in observation processes and coverage, handles high‐dimensional feature sets, and balances influence across datasets of differing size and quality. Using simulations, we show that incorporating mixture‐distribution weighting improves accuracy and statistical power relative to independent survey‐specific models. We then apply our model to data from the North American Breeding Bird Survey and the eBird participatory science project, demonstrating how complementary coverage from structured and opportunistic surveys can be leveraged to produce high‐resolution, spatially explicit bird population trends.