HidroCL: Flow forecasting model based on machine learning

Currently, Chilean society has incorporated the usefulness of weather forecasts into their daily lives. From knowing if today will rain to whether we can have plans outside this weekend. However, in terms of precipitation, knowing if it will rain or how much it will rain does not necessarily translate into a reduction in risk in extreme situations such as river floodings, turbidity events, or droughts. This requires going a step further in forecasting to quantify how much water our rivers will have to decide and make preventive actions.

For this purpose, we have created HidroCL, a 5-day ahead streamflow prediction platform available for hundreds of gauges in the country. We have combined multiple meteorological variables from international forecasting systems such as precipitation, temperature, and wind. Also, we considered historical hydrological variables such as mean precipitation, evapotranspiration, aquifer depth, and hydraulic conductivity. We added this information with satellite products that allowed us to incorporate new variables such as vegetation and reservoir water storage, which are essential to quantify the availability of water resources in a basin.

This large amount of data was processed and incorporated into multiple models, from conceptual models, decision trees, and the most recent machine learning techniques to obtain the best prediction at the gauges.

This platform will place Chile at the level of developed countries in terms of predictive capacity. However, this represents only the first step in our quest for more accurate predictions, on more sites, and free accessibility to the community.

Estero Llico

Río Frío

Río Llancahue

Valle del Elqui

Estero Pocuro

Río Futa

Río Ñuble

Ventisquero Colgante

Estero Yerba Loca

Río Golgol

Río Paine

Río Baker

Río Huequecura

Río Simpson

Río Clarillo

Río Lebu

Río Volcán

previous_button next_button