2018 Corn Yield Forecast Verification

AER’s inaugural corn yield forecast was produced in February of 2018.  The methodology for producing this forecast, as-well-as an analysis of would-be performance in prior years (a hindcast) is provided in a previous blog post that may be found here.  What follows in this blog post is a detailed accounting of the performance of AER’s February 2018 corn yield forecast, as-well-as an assessment of the perceived risks to the 2018 growing season, as highlighted in the February forecast, and how these risks manifested during the growing year.

We will be releasing our 2019 Crop Forecast in early March. We will be providing a similar breakdown of forecasted yield, and perceived risks to the 2019 corn growing season, so please visit us again for a first look at the upcoming season.

We start with comparing our forecasted yields and the final verified yield in 2018 from the USDA.  The USDA estimated the final yield at 176.4 bushels per acre (bpa), approximately 3.0 percentage points above trend. This final yield compared very favorably to our August 1^st forecast of 177 bpa and compared well to our February projection of 173 bpa. Figure 1 shows our two projections from last year in comparison to the USDA’s initial forecast a week later and their forecast in September.

Figure 1. Yield curve for the 2018 season. The probability density function curve (red line) is based on the final de-trended yields of 15 previous seasons that have had corn condition reports that were within the range of the three weeks following silking in 2018 (66-83% combined good to excellent). Going from left to right, the solid green line represents the yield at trend for 2018 (~171 bpa), the solid blue line represents our analog based projection back in February (~173 bpa), the first solid black line represents the USDA’s final 2018 verified yield of 176.4 bpa, the dashed blue line represents our August forecast of 177 bpa, the dashed black line represents the USDA’s projection of 178.4 bpa in August, and the final solid black line represents the USDA’s projection of 181.3 bpa in September.

We also want to evaluate how we performed last year with our “Primary Risks for the Upcoming Season” as shown in Table 1. The remainder of this blog is an evaluation of each of these in order from what we were most confident in to least confident in. Note that each risk has a confidence of high, medium, or low. Each risk also denotes whether we were discussing the entire Corn Belt or just a certain section. These risks were based on the summer forecast and analogs, which were prepared by Dr. Judah Cohen, AER’s Director of Seasonal Forecasting.

Table 1. Our Primary Risks for the Upcoming Season for the 2018 season that we released in our Crop Forecast in February 2018.

  Risk: Abnormally cool spring; Everywhere

Confidence: High

We were very confident of an abnormally cool spring across the Corn Belt. Figure 2a shows that our confident projection of an abnormally cool spring for the whole Corn Belt was accurate as temperatures were mostly below normal. This was particularly true in April, when temperatures were some of the coldest on record for much of the region (Figure 2b). It should be noted that May 2018 was one warmest May’s on record, but this warmth was not enough to offset the persistently cold temperatures prior to May in terms of the seasonal average. The cold start to the spring did lead to delays in planting regionwide; thus we feel that our forecast was not only accurate but useful.

Figures 2a-b. Departure from average temperatures (based on 1981-2010 normals) for the spring season (Fig. 2a), the month of April (Fig. 2b). Figures compliments of the High Plains Regional Center.  

Risk: Wet spring; Eastern Corn Belt

Confidence: High

Our relatively confident projection of a wet spring in the Eastern Corn Belt mostly was mostly accurate for Ohio and southern Michigan but generally not true for Indiana (Fig 4). However, I would point out that Ohio and southern Michigan were one of two regions of the Corn Belt to have above average precipitation last spring, as most places were drier than average.

Figure 3. Percent of normal precipitation for spring 2018. Figure compliments of the High Plains Regional Center.

Risk: Increased number of severe storms; Everywhere

Confidence: Medium

 We were reasonably confident of an increased number of severe storms across the Corn Belt. Using total storm report data from the Storm Prediction Center (Table 2), we found that 2018 was (in general) more active for severe weather in the Corn Belt States than the median (2013-2017).

Table 2. Total number of storm reports in each Corn Belt state in 2018 compared to the median number of reports of the previous five seasons. Data used to populate the table are from the Storm Prediction Center.

Risk: Warm nights in July; Everywhere

Confidence: Medium

We were reasonably confident of a warm spell with warmer nighttime temperatures for much of the region during July for much of the region. According to the USDA’s weekly Weather and Crop Bulletin (Figures 4a-b), the first two weeks in July were somewhat above average for much of the region. Additionally, I wrote a blog in mid-July that discussed the abnormally quick accumulation of growing degree days (GDD’s) that we had seen from planting to the early reproductive stage.

Figures 4a-b. Departure from average temperature for the United States during the weeks of 1-7 July and 8-14 July 2018. Maps compliments of the USDA’s Weekly Weather and Crop Bulletin.

Risk: September frost; Northern Corn Belt and western Nebraska

Confidence: Medium

We were also reasonably confident that locations in the northern section of the Corn Belt and western Nebraska had a risk of frost in the last half of September. Figure 5 confirms that there was a light freeze in the last third of September in much of Minnesota, North Dakota, northeastern South Dakota, and western Nebraska. This also was a prelude to a cool, wet fall (in many locations) that led to more harvest delays than in recent years, especially west of the Mississippi River.

Figure 5. First date of a 32F freeze this fall/winter. Map compliments of the Midwest Regional Climate Center.

Risk: Prolonged cool spell in late summer; Everywhere

Confidence: Medium

Based on our internal seasonal forecast and associated analogs, we determined that there was a modest chance of a cool last half of July and August across the Corn Belt. Figure 6 shows that there was abnormally cool weather in much of the north central United States the last part of July. Figure 7 shows that cooler than average weather persisted into August for much of the western Corn Belt, while slightly to moderately warmer than average temperatures prevailed elsewhere in the region.  

Figure 6. Departure from average temperatures during the period from 22-28 July. Figure compliments of the USDA’s Weekly Weather and Crop Bulletin.

Figure 7. Departure from average temperatures for August 2018. Figure compliments of the High Plains Regional Climate Center.

Risk: Dry spell in July; south of Interstate 80

Confidence: Medium

We determined that there was a chance of a drier than average July for the southern part of the Corn Belt. Precipitation was indeed below average across much of the Corn Belt that resides south of I-80 (Fig. 8). However, precipitation was also below average across much of the region in July with above average precipitation being mostly confined to Nebraska (except the far east) and the northwestern portion of the Corn Belt.

Figure 8. Percent of normal precipitation in July 2018 across the Corn Belt. Normals based on 1981-2010 data. Figure courtesy of the High Plains Regional Climate Center.

Risk: Severe drought in the Southern Plains build north into the Corn Belt; Everywhere

Confidence: Low

There was a severe drought across much of the Southern Plains at this time last year. Based on our seasonal forecast, we determined that it was unlikely that drought would expand into the entire Corn Belt. We were mostly correct about this. As shown in Figure 9, most of the Corn Belt was drought free in early August, as was the case for most of the summer. Flash drought did develop and intensify in northeast Kansas and northern Missouri in the early summer, later spreading into southern Iowa and parts of western Illinois. However it never spread beyond this area and by late summer, significant precipitation reduced or eliminated drought in these areas.

Figure 9. Drought intensity classification on 7 August 2018 as determined by the United States Drought Monitor. Figure courtesy of the National Drought

Please direct any feedback or questions my way ( Eric.Hunt@aer.com ).