ECMWF are aware of a systematic wet bias in IFS forecasts of precipitation over Southeast Asia, with an average bias of 1.61 mm/day in the boreal summer half-year, and precipitation overestimations on forecast day 3 of up to 31%. 

We are interested to hear from forecast users about their experiences and perspectives related to this bias in precipitation over Southeast Asia. Any feedback and comments may be used to inform ongoing research and diagnostic work. Some of the types of feedback we're interested include:

  • Further case studies / dates where this issue has been observed
  • Have you noticed the same issue in forecasts from other models?
  • What type of synoptic situations are contributing to this bias?
  • What processes could be causing these issues?

However, we also welcome any other insights, experiences or feedback you may have related to this issue. 


Left: April to September precipitation biases (mm/day) on forecast day 3 at 0000UTC at 5219 stations globally; blue colours indicate an overestimation of precipitation. 
Centre and Right: Example precipitation forecasts for two events, from 0000UTC 25 June to 0000UTC 26 June 2019 (centre) and 0000UTC 21 May to 0000UTC 22 May 2020 (right). Circles represent the observed precipitation totals. 


This issue has been under investigation and a detailed overview of the work done so far can be found in a recent paper by David Lavers, Shaun Harrigan and Christel Prudhomme, published in the Journal of Hydrometeorology (open access) in April 2021: Precipitation Biases in the ECMWF Integrated Forecasting System

These biases also have implications for other aspects of the hydrological cycle - soil moisture, evaporation, temperature, snowpack, and land-atmosphere feedbacks, alongside impacts on river flow forecasts generated using the IFS precipitation forecasts (the Global Flood Awareness System, GloFAS). Comments on these aspects are also welcome. 


If you would like to provide any feedback related to this issue, or any information, perspectives or experiences with using these forecasts, please comment below. We may not be able to respond to every comment, but will be monitoring and joining the discussions.