Impact of Delayed Planting on Corn Yield Forecasts


The most popular question from customers and prospective customers of CropProphet this year is “how do you account for the late planting delays in your yield forecasts?” This post addresses that question by looking at the performance of CropProphet in prior late planting years.

May 2019 = Wet in the Midwest

At this point in the US crop season, just about everyone knows about the very wet May that occurred and is watching carefully for information regarding its impact. On a production-weighted basis, corn growing areas in the United States received 163% of normal precipitation while soybean growing areas received 155% of normal precipitation.

May 2019 US County Precipitation Totals and Deviation from Normal

The excessive rainfall greatly increased soil moisture, caused widespread flooding, and prevented producers from being able to plant their crop when they normally do.

2019 Corn Planting Delays

The chart below shows US corn planted progress from 1990 to 2019 with 2019 highlighted as the heavy black line. As can be seen, there are not many years with similar planting delays in the past 30 years.

29 Years of USDA National Corn Planting Progress Reports – 2019 Highlighted in Black

We selected five years from this data that represent other late planting years. The years were 1993, 1995, 1996, 2011, and 2013, and these are highlighted in red in the graphic below.

29 Years of USDA National Corn Planting Progress Reports – Delayed Years Highlighted in Red

CropProphet and Planting Delays

The CropProphet modeling methodology for corn and soybean yield and production does not directly include the planting progress of each season. However, because the model is weather-based, relationships such as the following are included in the model:

bad weather -> causes planting delays -> shortens growing season -> reduces yield

Numerous other statistical relationships between spring weather and subsequent crop outcomes are also represented. CropProphet therefore captures the impact of planting delays on yield because we are quantifying the impact of early season weather via robust statistical relationships.

With 40 years of historical weather and US crop data available to CropProphet, it is possible to accurately quantify the impact of weather on crops, and as part of the modeling approach we recreate yield forecasts for prior years. This data is available in our Modeler product option. An artistic rendering of 33 years of US corn yield forecasts is shown below.

Example of 33 years of CropProphet corn yield forecasts shown in a single time series

Corn Yield Forecasts and Delayed Planting Years

Using the data from the 33 years of daily crop yield forecasts from the Modeler data set, we extracted the corn yield forecasts for the five prior years with planting delays to examine the forecast performance, measured as the mean average error (MAE) in these years compared to the average MAE performance of CropProphet. The results for the yield and production forecasts are shown below.

In this case, we show CropProphet forecasts from September 7th of each year and compare the result to the subsequent USDA mid-September yield and production forecasts.

CropProphet Yield Forecasts and Error During Delayed Planting Years for Forecasts made on September 7th of Each Year

CropProphet Production Forecasts and Error During Delayed Planting Years for Forecasts made on September 7th of Each Year


The conclusion from this analysis is that the CropProphet:

  1. yield forecasts have been better in previous delayed planting years (1.6 bpa mean average error) than in normal years (2.3 bpa)
  2. yield forecasts during delayed planting years are substantially better than the USDA forecasts in those years
  3. production forecasts are very similar in skill (198 mm bushels delayed vs 188 bushels all years) in delayed planting years.
  4. production forecasts are substantially better than the USDA forecasts during delayed planting years

Most importantly, CropProphet forecast skill was not degraded in prior years that experienced planting delays, suggesting that the relationship of weather to corn yield is appropriately captured by the model.