Proof Point: Surplus following drought in the Amazon Rainforest

6 May 2025

In this installment of the Proof Point series, we detail current weather-related events affecting the Amazon rainforest and the historical accuracy of the Water Security Indicator Model (WSIM) forecasts.

Description of the Event: Emergence of surplus following prolonged drought in the Amazon Rainforest

Date of Event: February - March 2025

What happened, where and when? Accurate prediction of transition from deficit to surplus in the Amazon Rainforest

Flooding following prolonged drought in the Amazon

Since June 2023, the Amazon rainforest has been experiencing widespread exceptional drought, which has had a devastating impact on regional communities and ecosystems. Last November, we assessed the region’s increasing drought risk and its struggle to recover from four consecutive “one-in-a-century” dry spells, which caused vital river systems in the region to fall to unprecedented low levels.

In late February 2025, after this prolonged dry spell, the region saw heavy rainfall, which triggered landslides and caused destructive flooding across multiple regions of the Amazon, specifically affecting indigenous communities along the Ene River. Agricultural and communal infrastructure were destroyed, which aggravated regional food security and displaced both residents and local wildlife, including the endangered population of Amazon jaguars. This transition from water deficit to surplus is clearly depicted in the composite water anomaly maps computed using observed data from ERA5 shown below. Surplus conditions continued through March 2025.

WSIM Forecasts

This transition and the resulting flood events were predicted by WSIM forecasts as far as four months in advance. As each monthly forecast was released, lead times shortened, and the accuracy of the forecasted anomalies improved.

In early November 2024, WSIM produced a forecast of one-month surface water anomalies for December 2024 through March 2025 using observed data from ERA5 through October 2024 and an ensemble of 28 NOAA CFSv2 forecasts issued the last week of October 2024 as shown in the figure below. This forecast, produced four months in advance of the February transition, depicted an emergence of surplus in a drought-stricken region of the Amazon rainforest, initially appearing mostly in northern Brazil. However, the timing of the forecast transition was slightly off, appearing in March 2025 instead of late February 2025. Furthermore, the forecast only called for severe surpluses instead of exceptional surpluses. These types of errors in timing and magnitude are common with long lead-time forecasts. Nonetheless, the forecasts are still useful for contingency planning.

Two months later, in early January 2025, WSIM produced a forecast of one-month surface water anomalies for January through March 2025 (1-3 month lead times), correctly depicting the timing, magnitude, and location of the February transition. These anomalies spread into significant portions of the Bolivarian Nations, most notably throughout Colombia. As lead times shortened, WSIM forecasts correctly depicted a transition to exceptional surpluses consistent with the floods events that subsequently occurred February and March 2025.

Our final figure displays the WSIM forecast for March issued in early March using observed data from ERA5 through February and 28 CFSv2 forecasts issued the last week of February. The one-month lead forecast closely matches the pattern in the observed data shown above.

Conclusion

WSIM correctly predicted the February 2025 transition from widespread exceptional drought conditions to exceptional surpluses in the Amazon Basin as long as four months in advance. While the forecasts issued in early November 2025 were imprecise, they were still useful for contingency planning purposes. Subsequent forecasts issued two months in advance had much higher skill.