A Simple Solution for Refining Lake Water Temperature Profiles Data Arrayed from High-Frequency Monitoring Sensors
The revolutionized aquatic monitoring sensors are essential in capturing environmental patterns that traditional discrete samplings might not be able to. They allow scientists to further synthesize and better conclude processes in aquatic ecosystems. These sensors produce high-frequency data that provide information on a fine temporal scale, even near real-time. The massive quantities of the streamed data, however, create challenges for scientists to grasp the concrete information. Filtering data quality, on the other hand, is another problem scientists might have encountered as sensor accuracy and precision may drift along the line. Hence, quality assurance and quality control might be quite labouring owing to the size of datasets to handle. This paper proposed a semi-mechanistic algorithm to improved false water temperature data. Using “theoretical” thermal stratification as a reference, this algorithm fixed sensors error readings. A 5-month dataset of water temperature profiles of Lake Maninjau, West Sumatra, captured every 10 minutes from a set of sensors in thermistor chain was applied. We found that most data fit to the theoretical temperature profile, R<sup>2</sup> = 0.962, RMSE = 0.081<sup>o</sup>C. A number of errors, however, were observed in the upper layer of the lake (<20 m), the most dynamic layer in terms of its thermal variation. Sensor drifts in this active upper mixed layer can be related to the generated errors. Through this simple solution, not only improving the quality of the observed water temperature data, but was also able to identify the most probable source of errors
