Systematic errors in borehole temperature monitoring records in permafrost – Part 1: Stability of calibrated temperature sensor offsets

Abstract. This study investigated some of the systematic errors that contribute to the uncertainty budget associated with short- and long-term temperature measurements in permafrost. We quantified errors associated with four types of sensors' behaviors that are not addressed by the most commonly carried out single-point 0 °C bias correction: i) long-term drift of the 0 °C-offset using sensor recalibration data after six and nine years, ii) non-linear sensor offsets dependence on temperature (multiple-step calibration), iii) drift of the 0 °C-offsets following thermal stress, and iv) hysteresis of temperature values measured by the sensor. 
We measured sensor drift rates up to 0.0157 °C per year, with 95.3% of evaluated sensors drifting towards warmer temperatures. Extrapolated over 30 years, the observed positive drift of 0.471 °C would be comparable to the rate of permafrost warming reported from some permafrost regions, assuming that at the deployment temperatures, the drift rate remained nearly linear. We also confirmed non-linear sensor offsets across the -20 °C to 20 °C range, with larger errors at negative temperatures. These results support the argument for multi-point calibration, particularly at and below 0 °C, to improve the representation of phase-change processes, especially in fine-grained and saline soils. Additionally, 43 thermal cycles between -20 °C and 20 °C in the lab produced measurable offset drift, while the sensor readings exhibited hysteresis, indicating dependence on prior thermal state. Based on the physical mechanisms underlying these errors, we argue that similar drift-inducing and drift-enhancing processes are likely present in other sensor technologies commonly used in permafrost monitoring. We therefore recommend extending current calibration best practices to include periodic recalibration of individual sensors, comprehensive reporting of sensor metadata and calibration histories, and calibration guidelines tailored separately to long-term monitoring and short-term measurement campaigns. Such measures would improve the quality, traceability, and comparability of permafrost temperature records that we are acquiring as a permafrost community.