Unclear Computer Models Made 2013 Colorado Flood Prediction Difficult

Unclear computer weather models made it difficult to predict the location and magnitude of heavy rains during Colorado’s deadly September storms, the National Weather Service said.

In a report assessing its performance during the flood, the agency said forecasters were still able to give an average of 69 minutes’ warning of flash floods, beating the national goal of 58 minutes.

The report was released this week.

Forecasting models gave varying predictions on the time, location and magnitude of rainfall, the agency said, and some significantly underestimated the amount of rain.

The forecast system that proved most useful during the storm predicted 8 inches of rain, but components of the system were not readily available to forecasters. Even that model was inconsistent on the expected time and location of the rain.

The seven-day storm dropped from 8 inches to more than 17 inches of rain along the mountains and foothills of the Front Range. State officials said nine people were killed in the flood and a 10th died in the recovery effort.

The floods caused $2 billion in damage.

The report said several flash flood warnings on the morning of Sept. 12, the storm’s second day, provided no lead time. The report did not specifically address how forecasters achieved the 69-minute average while contending with unclear modeling data.

The report said the Weather Service’s parent, the National Oceanographic and Atmospheric Administration, doesn’t put as much time and money into improving precipitation forecasts as it does for hurricane forecasts. It recommended a long-term program aimed at substantially upgrading its ability to predict extreme precipitation.

River forecasting centers had multiple problems, starting with the lack of accurate rainfall predictions, the report said. Many gauges that measure stream flow washed away or transmitted inaccurate data, and glitches in a new automated reporting system kept forecasters from getting measurements from some sites.

The review also cited problems with slow Internet connections, limited coordination between local forecasting offices and the national Weather Prediction Center and differences in the level of information that forecasters passed on to local officials.