Posted at 2:25 PM on April 23, 2008
by Craig Edwards
More than two decades ago, when I was the chief of forecast operations at the Indianapolis Weather Service Office, we were wrestling with the challenges presented by predicting the crest along major rivers. For the most part, we dealt with early spring snow melt combined with observed rainfall, as well as the impact of ice jams.
Rises on the river were based on observed stages, antecedent conditions,(i.e. soil moisture), temperature forecasts, the time of year and reported rainfall.. The river forecast models would ingest observed 24 hour rainfall, averaged over the basins, forecast the flow along the tributaries and then forecast the rise and fall at specific locations for the next five days.
Unfortunately, in the 1980s we had yet to boldly accept the challenge to attempt to incorporate what is now known as quantitative precipitation forecast, or QPF.
Currently skilled hydrologists and meteorologists in NOAA's River Centers and local offices are adding this additional forecast information. Daily computer model runs are adjusting the crest forecast based on anticipated rainfall, averaged across the basin for the next seven days.
These critical forecasts of quantitative precipitation help to keep the stair-stepping appearance to a minimum, since forecasters are now adding expected additional rainfall, not just observed rainfall. In the 80s, we often knew the river could continue to rise higher if more rain occurred. So a forecast crest of 15.5 feet might step up to 16.0 feet if rain fell in the basin while the river had yet to crest.
You can follow along on the river crest forecast on the National Weather Service River Forecast link. It may be of interest with the periods of wet weather expected in the next three to four days.
This was one of those "paradigm shifts" mentioned in The Seven Habits of Highly Effective People that I watched (and participated in) while working as a hydrologic forecaster at the West Gulf RFC about a decade or so ago. As you said, our forecasts were done without any future precipitation, but the forecasters "knew" stages would go higher with additional rainfall. The kicker was how much higher.
Toward the end of my career there (late '90s), we started routinely incorporating (HPC) QPF into our forecasts, and that seemed (in a qualitative sense) to make them better. But we didn't yet have the tools to quantify that.
Another problem with the QPFs down there (and across much of the country in the warm season) is that heavy rainfall is usually convective, and thus is hard to forecast spatially. That 5-inch bulls-eye shifted 50 miles any direction would be upstream of, say, San Antonio, or Austin, or downstream of both; any of those would yield very different river forecasts.
While the amounts in the QPF were (and still are) mean areal amounts, I would double those values for point totals. For example, I expected to see a 6-inch report inside a 3-inch isopleth.
And just as the accuracy of weather forecasts decreases the farther out you go, the QPF also decreases in accuracy. A seven-day QPF is pretty much useless past day 3, IMHO, and many times, past day 2.
But, back to your original point, the inclusion of QPF in river forecasts is a good thing, despite the problems with it being a forecast.