The Influence of Logger Bias on Reported Temperature Trends: Implications for Temperature Monitoring Networks

Abstract:

There has been increasing interest in river temperature monitoring and research in recent years. This has been driven by factors including a greater awareness of the importance of river temperature for freshwater ecology, the potential for detrimental extremes under climate change and the availability of increasingly affordable dataloggers. A number of studies have attempted to collate and analyse pre-existing long-term (decadal) datasets to assess for evidence of temporal trends. These studies require considerable care given the magnitude of temporal trends (often < 1 degree per decade), the low signal to noise ratio in the data and the potential for bias across different makes, models and individual dataloggers. Despite these issues, data quality control often receives only a superficial consideration with subjective assessments of data quality or a reliance on manufacturer reported accuracy with consequences for the reliability and interpretation of findings. This study assessed the potential influence of logger bias on reported temperature trends in the Girnock Burn, Scotland over > 25 years. The bias of temperature measurements made by different dataloggers (two makes and five models) was determined through cross-calibration against a reference datalogger. Long-term trends in stream temperature metrics (daily mean, max, min) were characterised using Generalised Additive Mixed Models (GAMM). Models were fitted to (1) the raw data and (2) data corrected for logger bias. Significant non-linear temporal trends were observed in the raw data. These trends were accentuated when corrected for logger bias. Given the potential to accentuate or remove long-term trends, it is suggested that robust internal and external calibration and quality control procedures should be established for new temperature networks. Such approaches are capable of removing logger bias and improving accuracy by an order of magnitude over manufacturer stated values.