Latest articles

The rain dilemma for U.S. corn, soybeans

Localized storms are making widely used weather models less meaningful to the market


Rain has been falling across the U.S. Corn and Soybean Belt this month but crop ratings have been low or declining, a sign that some farmers may be getting too much moisture while others have parched fields.

Because summer storms have been so localized, the two leading weather models used by traders may be little help in gauging whether this summer’s hit-or-miss weather pattern may cause U.S. corn and soybean yields to end up lower than expected.

July and August are the key months for rainfall in the United States that dictate the yields for corn and soybeans. But as of July 23, the U.S. Department of Agriculture rated 57 per cent of soybeans and 62 per cent of corn in good or excellent condition, both five-year lows for the week.

The two weather models that commodity markets trade on, the U.S. GFS and the European EC, have greatly improved the resolution to just a few miles in the past couple of years in an attempt to offer more precision on the location of potential events.

Observed weather conditions may be charted or mapped allowing analysts to see how much rain actually fell over a given period of time. But the spatial resolution on these maps can be highly variable, and precipitation events often get smoothed over too much.

The U.S. GFS model – or Global Forecast System – is a gridded model that computes temperature and precipitation forecasts four times per day, extending out 16 days. Up to day 10, the spatial resolution or grid size on the output is eight miles (13 km).

Beyond day 10, GFS resolution is much coarser at 21 miles (34 km). For reference, the average span of a county in Iowa, the country’s top corn producer, is about 24 miles.

The European EC model has an even higher resolution at nine km (5.6 miles), but each run only covers 10 days and the model is run twice per day – neither of them landing when the Chicago grain markets are open for trading. Nevertheless, the EC model is often touted as being more accurate than the GFS.

Localized rain

The typical summer storm activity that accounts for a large portion of the seasonal rainfall has been suppressed this year across the western Corn Belt in particular, based on the recent positioning of upper-atmospheric features.

A stubborn high-pressure ridge has been parked over the Plains, bringing heat and drought conditions to the western Corn Belt, with the worst of it in the Dakotas. Meanwhile, all the main storm activity has been riding the jet stream along the east side of the ridge straight into Wisconsin, northern Illinois, and the eastern belt.

The same weather pattern has been responsible for burdensome rainfall amounts recently from Wisconsin to Ohio. This pattern has drawn a harsh line right through the core corn- and soybean-growing areas, thinly separating the wet regions from the dry ones.

Along this boundary, rainfall amounts have been highly variable, which is why farmers’ crop feedback has been anything from terrible to excellent.

Take Cedar Rapids and Iowa City in eastern Iowa. According to data from the National Weather Service, Cedar Rapids received three inches (76 mm) of rain between July 1 and July 26, about 80 per cent of normal for that period.

About 23 miles (37 km) to the south, Iowa City observed 6.5 inches (165 mm) of rain over the same period, roughly 50 per cent more than average.

Another example lies farther west in the state, where monthly totals through July 26 for the capital city of Des Moines was half of that in Ames, which is 30 miles (48 km) due north.

Maps vary

Many analysis maps that market participants use to see how much rain actually fell are often smoothed over twice as much or more as the weather model output. This can make it difficult to see the highly localized discrepancies such as the Iowa example above.

Corn-heavy northern Illinois, where heavy precipitation events have been common this month, provides a great example of how different analysis maps could lead to different conclusions.

A smoother, gridded observation map would say that the northernmost counties in the state have received four to five inches (102 to 127 mm) between July 1 and July 26. However, a much finer-scale map using weather station data suggests the same region may have actually observed anywhere from two to 10 inches (51 to 254 mm).

But there are limitations to the station method. Sometimes there is only one station in a county, which may be the same size as the grid on a more smoothed map anyway, rendering the station method no better.

The station method also comes up short where station coverage is lacking, which is often a problem in sparsely populated areas. And since temperatures tend to have much less spatial variation than precipitation, the smoothing effect on that data is often less consequential.

About the author

Columnist/Reuters

Karen Braun is a Reuters market analyst based in Chicago. The views in this column are her own.

explore

Stories from our other publications

Comments