Tracking the Heat and Drought

Most commentators attribute this year's bad crop progress with drought--a lack of rainfall.  The problem with traditional drought measures is that they don't predict crop outcomes especially well.  Our measure of extreme heat--degree days above 29C--predicts crop outcomes a lot better.

Extreme heat is correlated with traditional drought measures, but only very roughly.  But it probably has a stronger association with water stress in plants, since there tends to more evaporation and evapotransporation when it's very hot (vapor pressure deficit increases).  Also, heat can have its own direct damaging effects on plants.

So, here's the current situation for extreme heat and precipitation the US relative to history since 1960.  We obtain a single nationwide index by weighting counties by trend production.  County level measures are derived from our own daily fine-scale (4km grid) weather data, which combines PRISM monthly data with daily weather station data.

These plots were generated by my colleague Wolfram Schlenker.  We're also going to start cross-posting at our joint site G-FEED, a group that includes Wolfram as well as David Lobell, Solomon Hsiang and Jarrod Welch.

The gray lines show earlier years, the hottest prior to this year being 1988 and 1983, which had really bad crop outcomes.  The driest year was also 1988; but 1983 is one of the lines in the middle somewhere.

From the extreme heat measure, it looks like we're on track for the worst yield outcome (relative to trend) in over 50 years. Precipitation doesn't look quite as bad, but close.

One key feature that I think makes this year especially bad:  the slope of the extreme heat line during the month of July appears to be the steepest on record while the precip line for July appears to be the shallowest on record.  This could turn out a good bit worse than 1988.


  1. One question I keep coming back to in discussions of the extremely poor growing conditions this summer is what does a "very bad" harvest mean? Looking back at yield per acre graphs it looks like both '83 and '88 corresponded to drops in yield of ~40 bushels per acre or a 1/3 reduction in total yield per acre (when comparing to yields of other years in the same era). Of course since then the expected yields in a normal year have increased significantly.

    From your experience with these datasets, do you have a sense of whether it makes more sense to look at the percent reduction or absolute reduction in yield when drawing comparisons to previous bad years?

  2. Good question James. As a first cut, I would look at percent reduction from trend. And I think your assessment is about right. I'd guess a about 1/3 reduction in yield from what was expected early season, maybe a bit worse. But then the very early planting may make this year a little different that what our models tell us, plus some new drought tolerant varieties were introduced this year. So, I'd put a fat uncertainty bands around that prediction.


Post a Comment

Popular posts from this blog

Nonlinear Temperature Effects Indicate Severe Damages to U.S. Crop Yields Under Climate Change

Commodity Prices and the Fed