Gro Analysis Points to Bigger US Cotton Crop

As harvesting of US cotton gets underway, a series of USDA downgrades of the crop’s condition has increased uncertainty about what was expected to be a bumper crop this year.

A Gro analysis that compares this year’s growing conditions with previous years’ indicates that there’s reason for optimism. Although final yields may come in below USDA NASS expectations in some major cotton producing states, gains in Texas, the biggest producer by far, may more than offset declines elsewhere, our analysis found.

The USDA’s most recent production estimate for the 2019 upland cotton crop is 21.1 million bales. Upland cotton represents the vast majority of US production. By contrast, Gro’s analysis indicates total upland cotton production this year that is 2.0% higher at 21.6 million bales. This compares with 2018 upland production of 17.6 million bales, one of the highest levels in the previous decade. This year, the USDA reduced the share of the cotton crop rated good or excellent from a historically strong level of 61% in late July to a much weaker 39% by late September. The department rates the crop on a weekly basis.

This map shows readings of NDVI anomalies in July for the major cotton-producing states. Anomalies compare current NDVI, a satellite-derived measure of vegetative health, with the average of the previous 10 years. A Gro analysis found that NDVI anomalies for July, when cotton plants are in their squaring stage, historically are strongly associated with final crop yields. 

The US is the world’s largest exporter of cotton by far, so its production is important for international trade flows and supplies. Price volatility in cotton futures rises in November as markets assess the size and quality of the US crop. Similar bouts of market volatility occur in the spring when Brazil and Australia, the world’s second- and fourth-largest cotton exporters, also harvest.

In the Gro analysis, we looked at state-level data for each day of the growing season for measures including NDVI, soil moisture, and rainfall, traditionally strong predictors of cotton yield. We then compared these daily data points to their trailing 10-year averages. And we did that for each year going back to 2011. All of this data is available for free in the Gro web app. For higher geographic granularity, county-level data can be accessed through Gro’s API.

The next step in our analysis was to look at the most critical water-availability period for a cotton plant, which is around the squaring phase of development when water and nutrient availability determine the number of cotton bolls on each plant. Measures of vegetative health taken during this phase can strengthen a model’s predictive power. Accuracy could be increased further if each state’s crop development schedule is addressed individually. For the sake of simplicity, however, our analysis assumed that the seven top producing states were all significantly into their squaring stages in July.

We then calculated for each state the relationship between water availability and ultimate yield. Of the three measures we examined, we found that NDVI was the most reliable for predicting final yields, especially in Arkansas, Georgia, Missouri, and Texas. Soil moisture and rainfall, perhaps the most commonly watched metrics, had far less predictive power than did NDVI.

This chart shows the percentage of cotton crop the USDA rated good or excellent in 2019 (line with markers) compared with previous years (other lines). This year, the USDA reduced the share of ratings that were good or excellent from a historically strong level of 61% in late July to a much weaker 39% by late September. 

The levels of NDVI seen this past July are associated with yield in Arkansas that is 3% lower than the USDA NASS estimate reported on Sept. 12th, 15% lower in Missouri, and 8% lower in Georgia, according to our analysis. For Texas, however, the analysis estimated yield would be 11% higher than the USDA forecast. Because Texas’ expected crop of 8 million bales is roughly 50% larger than the expected crops of the other three states combined, our projections are for a net 2.0% increase from USDA NASS’ national projection.

Our analysis shows that NDVI readings in July, when cotton plants are in their squaring stage, have historically been predictive of final crop yields. By contrast, USDA NASS crop condition ratings in July have for the most part shown little correlation to historical final crop yields. Only Arkansas, the No. 4 producing state, has had a strong association historically between July conditions and final yields.