Development of Hourly Resolution Air Temperature Across Titicaca Lake on Auxiliary ERA5 Variables and Machine...

Area of Science:

• Meteorology
• Data Science
• Environmental Science

Background:

• Meteorological data often suffers from gaps, inconsistencies, and outliers.
• High-resolution, reliable data is crucial for environmental monitoring and climate studies.
• Automatic weather stations provide valuable but sometimes imperfect datasets.

Purpose of the Study:

• To develop and evaluate an integrated procedure for enhancing meteorological data quality.
• To combine advanced quality control (QC) and machine learning (ML) for reliable data.
• To address data gaps and inconsistencies in air temperature records from the Lake Titicaca region.

Main Methods:

• Applied advanced QC statistics (Interquartile Range, Biweight, Local Outlier Factor) to clean raw data.
• Implemented spatial and temporal gap-filling techniques using nearby station data and ERA5-Land reanalysis.
• Utilized ML models including Random Forest (RF), Support Vector Machine (SVM), Stacking (STACK), and AdaBoost (ADA).

Main Results:

• The Random Forest (RF) model demonstrated superior performance, achieving R² values up to 0.9 and RMSE below 1.5 °C.
• Spatial gap-filling was effective for highly correlated station data, while temporal methods suited areas with low correlation.
• The integrated procedure significantly enhanced the reliability and completeness of meteorological data.

Conclusions:

• The combined QC and ML approach effectively produces reliable, continuous, high-resolution meteorological data.
• The method is adaptable to different geographical conditions and can be extended to other meteorological variables.
• This procedure offers a robust solution for improving environmental data quality in data-scarce regions.