How do data collection and processing methods impact the accuracy of long-term trend estimation in lake surface-water temperatures?

Identifying significant changes across lake ecosystems is important for understanding impacts of global environmental change. Synthesizing data on lake warming trends is challenging because individual lake datasets differ in the: (1) length of the time series available for analysis and (2) frequency of data collection (e.g., daily vs. monthly observations). This study aimed to address how dataset length, frequency of data collection, and strength of temperature trends could impact both the accuracy of summer surface-water temperature trends and their statistical significance. Using Monte Carlo simulations, we found that accuracy in trend estimates and the ability to recover statistically significant trends were both directly related to trend strength, dataset length, and sampling frequency. To consistently retrieve statistically significant trend estimates that deviated < 25% from the true values, 30-yr datasets with high warming rates (≥ 0.75°C decade⁻¹) were required. These findings have important implications for efforts to analyze lake temperature trends, as the characteristics of many existing datasets fall within a range where our simulations predict low accuracy in trend estimates as well as a low probability of achieving statistical significance. Longer datasets are needed to accurately estimate warming trends and evaluate drivers of lake surface-temperature changes, highlighting the need to support existing long-term monitoring projects occurring across the globe, and to encourage updates to remotely sensed lake temperature datasets.