Scientific post-processing of climate data

🌍 Improving Climate Model Accuracy Using Machine Learning: A Multi-Model Ensemble Approach 📢 Just wrapped up an exciting project where I used Bayesian Optimization + XGBoost to compute a Multi-Model Ensemble (MME) of Global Climate Models (GCMs). 🧠 The Goal: Climate models vary widely. Instead of relying on a single GCM, I combined outputs from multiple models—CESM2-WACCM, INM-CM4-8, and EC-Earth3—to better match observed record. 🔧 The Process: ✅ Data Preprocessing ✔ Cleaned + normalized GCM & observed data ✔ Filled missing values and ensured time-consistent splits ✅ Bayesian Optimization Used scikit-optimize to find optimal hyperparameters for an XGBoost model, accelerating convergence with smart probabilistic search. ✅ Grid Search Refinement Fine-tuned the best Bayesian result using a local Grid Search for extra precision. ✅ Evaluation 📊 Metrics: R², RMSE, and NSE 📈 Visuals: Time series comparison + residual analysis 🔍 Why It Matters: MMEs are crucial for reducing uncertainty in climate predictions. By integrating machine learning with GCM outputs, we can boost reliability for real-world decision-making—from water resource management to climate adaptation strategies. 🚀 Youtube video link🌱 https://lnkd.in/d5k3wFrx #ClimateChange #MachineLearning #XGBoost #BayesianOptimization #GCM #EnvironmentalScience #AI4Climate #Hydrology #DataScience #ClimateModeling #Python #TimeSeries #MME

All climate models have systematic bias. How do scientists deal with this KEY issue? Is it possible to correct the distribution while preserving the variability and the long-term trends? As you know, Global Climate Models are the primary source of information for reproducing and exploring climate scenarios. But there is a HUGE difference between Climate models running many years ahead, and the real-time Weather forecasting models. The difference is because: Climate models don't assimilate any satellite observations, any airplane or radar or ship measurement. As a matter of fact, there is nothing to assimilate in 2024-2050! This is why all Climate models "live" in their own alternative "reality". Climate models reproduce hurricanes and droughts, but this data should be corrected. And this is where you see the difference between true climate experts and the amateurs. It is necessary to bias-correct the raw climate model outputs in order to recalibrate and improve climate projections. This is the first and, actually, it is the most IMPORTANT part in data processing! Climate data should be aligned with the observations! And there are several tricks! The long-term tendency, for example! As of today, "quantile mapping" is the most advanced methodology for bias correction of climate models. According to this methodology, the bias is unique for each geolocation, for each weather parameter (temperature, wind, precipitation), for each calendar month, and for each part of the distribution curve (quantile). At Weather Trade Net we apply this methodology to assess the skill score and the capability across different climate models and across different climate change scenarios. University of East Anglia METEO FRANCE MetOffice World Energy & Meteorology Council (WEMC) Folmer Krikken Victor Estella Perez Paul-Antoine Michelangeli