Canada1Water - Machine-Learning Based Bias Correction of ERA5-Land and FLDAS Soil Temperatures Using SWE and NDVI

On May 26-29, 2024 - Tyler Herrington attended the Canadian Geophysical Union’s (CGU) annual conference to present his recent work titled 'Machine-Learning Based Bias Correction of ERA5-Land and FLDAS Soil Temperatures Using SWE and NDVI'. The presentation addressed critical issues identified in previous studies, highlighting substantial biases in reanalysis and LDAS-based soil temperature estimates. After evaluating the performance of three distinct bias correction methods: mean bias subtraction (MBS), multiple linear regression (MLR), and random forest (RF). The central research questions guiding his investigation were whether machine-learning approaches could effectively correct biases in ERA5-Land and FLDAS soil temperatures and how different correction techniques compared in reducing these biases.

A week after, Tyler had attended the Eastern Snow Conference where he presented another talk on this work. Review the poster below to explore the innovative research that was done on using machine learning to improve soil temperature accuracy in ERA5-Land and FLDAS datasets.

Click here to see a high-resolution copy of the poster.

Machine-Learning Based Bias Correction of ERA5-Land and FLDAS Soil Temperatures Using SWE and NDVI - Presentation Screenshot.