Posts by Collection

portfolio

publications

Machine Learning Solutions for Reservoir modelling

Published in American Geosciences Union Fall meeting 2021, 2021

The forecast of near real-time water availability in reservoirs helps in effective water resources management, flood control, hydropower generation, irrigation, drought mitigation and climate analytics. This study uses learning techniques to forecast reservoir water availability based on historical water availability, hydro-meteorological data, and water demand. Three deep learning architectures have been used fully connected deep neural network, long short-term memory network, and WaveNets. They take inputs as storage, inflow, outflow, and precipitation leading up to the reservoir and use these inputs to predict the storage level 30, 60 and 90 days ahead. This study uses SHAP (SHapley Additive exPlanations) - a game theoretic approach to explain the output of the machine learning models. This predictive modelling exercise is complemented with hydrological modelling using Soil and Water Assessment Tool (SWAT) and the best model combinations are found. The visualization and dissemination of the results of this exercise and other important hydro-meteorological datasets is done through an interactive dashboard to help stakeholders in making informed decisions. The final product updates on daily input data and builds predictions and visualizations.

Recommended citation: W. Samuel, S. Garg, et al. Machine Learning Solutions for Reservoir Modelling. https://ui.adsabs.harvard.edu/abs/2021AGUFM.H55Q0931S/abstract

WeatherBench Probability: A benchmark dataset for probabilistic medium-range weather forecasting along with deep learning baseline models

Published in Arxiv, 2022

WeatherBench is a benchmark dataset for medium-range weather forecasting of geopotential, temperature and precipitation, consisting of preprocessed data, predefined evaluation metrics and a number of baseline models. WeatherBench Probability extends this to probabilistic forecasting by adding a set of established probabilistic verification metrics (continuous ranked probability score, spread-skill ratio and rank histograms) and a state-of-the-art operational baseline using the ECWMF IFS ensemble forecast. In addition, we test three different probabilistic machine learning methods – Monte Carlo dropout, parametric prediction and categorical prediction, in which the probability distribution is discretized. We find that plain Monte Carlo dropout severely underestimates uncertainty. The parametric and categorical models both produce fairly reliable forecasts of similar quality. The parametric models have fewer degrees of freedom while the categorical model is more flexible when it comes to predicting non-Gaussian distributions. None of the models are able to match the skill of the operational IFS model. We hope that this benchmark will enable other researchers to evaluate their probabilistic approaches.

Recommended citation: S. Garg, S. Rasp, N. Thuerey. WeatherBench Probability: Medium-range weather forecasts with probabilistic machine learning methods. https://arxiv.org/abs/2205.00865