In the previous post I looked at the ranked highest temperatures recorded at GHCND stations since 1970. I was very surprised to find an apparent downward trend in the American Mid West. This does not seem to be consistent with the commonly held view that maximum daily temperatures would tend to rise locally as the global mean temperatures increase.
This was such an odd finding that I needed to try a more robust analysis to check it. Taking a single high temperature is clearly not a good metric So I took the yearly mean and all the days in summer, defined by the readings from June, July and August. I then fitted a regression on the yearly mean and the mean summer maximum temperature against the year since 1970. All global data sets and mast local stations show a trend of increasing temperatures that range between 0.1 and 0.7 degrees per decade (with some outliers)
The expectation would be to see higher summer temperatures in most regions.
The first map shows the mean annual TMAX temperatures. The slope is cut into ranges and represents the trend in the change in temperature per decade. Only stations with a significant (P<0.05) trend are included. The second map shows only the summer months of June, July and August. The maps are zoomable and best viewed fullscreen (right click on the full screen icon and open in new tab). There is a lot of data here, so it requires some exploration. Clicking on any point shows the details for the station concerned (slope, p-value, estimated land use, station name, station ID)
This was by far the most surprising result of any analysis of climate data that I have ever conducted. I remain very concerned that I could have made a mistake. All the data are publicly available from ghcnd so the result can be replicated easily. I have not included the full R code here as much of it concerns downloading the entire data set. The key result requires independent validation to ensure that any errors I may have committed are not repeated. If it does replicate then the implications are quite profound.
Assuming the results are valid, summer temperatures in the “wheat belt” of the United States have shown a declining trend. It is rather difficult to explain this, but one possible explanation involves the greening trend shown in areas of cropland. If crops are transpiring more water then the result would be to raise humidity slightly. This would be imperceptible when crops are interspersed with other land uses. However the mid west of the USA constitutes the largest contiguous area of cropland in the world.