Skip to content

Forecast Data

The GEOGLOWS model produces ensemble streamflow forecasts using data from the ECMWF (European Centre for Medium-Range Weather Forecasts) ensemble system. Each forecast includes 52 ensemble members: 51 low-resolution members with a 15-day forecast horizon, and 1 high-resolution member with a 10-day horizon. The low-resolution members have a time resolution of 3-hour intervals, while the high-resolution member provides forecasts at 1-hour intervals. These forecasts offer probabilistic insights into future streamflow conditions, allowing users to assess potential variability.

image

Forecast plots are designed to help users interpret the range of possible outcomes and uncertainties. The most commonly used forecast plot presents the median, 20th percentile, and 80th percentile, which represent 60% of the probability distribution within the ensemble members. These statistics give insight into the potential variability of future streamflows. This allows users to gauge the likelihood of different flow scenarios.

image

In some cases, the plot may also include mean, minimum and maximum flow values, as well as the high-resolution member, offering a comprehensive understanding of the forecast. This statistical breakdown allows users to assess the likelihood of various flow scenarios, making it a valuable tool for making informed decisions related to water resource management, flood forecasting, disaster preparedness, and risk mitigation.

image

Forecast Data Presentation

The Colab notebook provides an interactive guide on accessing and visualizing forecast data from the GEOGLOWS Model. It demonstrates how to retrieve streamflow forecasts, plot the data using Python libraries, and interpret key statistics for effective water resource management and planning:

Forecast Simulation Colab

Forecast Bias Correction

The GEOGLOWS model applies bias correction to its forecast data by assuming the forecast shares the same biases as the retrospective simulation. This process involves mapping forecasted streamflow values to a non-exceedance probability using the historical simulation's flow duration curve and then replacing the forecasted values with corresponding values from the observed flow duration curve.

image Forecast Bias Correct Methodology

This method helps improve forecast accuracy, particularly during earlier forecast lead times, aligning the data more closely with historical observations. However, improvements are limited by the assumption that the biases in forecast data are identical to those in the retrospective simulation​.

image Kling-Gupta efficiency (KGE) evaluating the performance of GEOGLOWS forecast simulation.

image Kling-Gupta efficiency (KGE) evaluating the performance of GEOGLOWS bias corrected forecast simulation.

Bias Correction Presentation

This Colab notebook offers a step-by-step guide for performing bias correction on GEOGLOWS forecast values. It shows how to adjust forecasted streamflow values using historical observations, improving the accuracy of predictions and aligning the data with real-world measurements for better hydrological analysis:

Bias Correction Colab