How Planet’s Productivity Forecasting Solution is helping farmers and economists move forward with this year’s harvest

Contributed to data analysis and visualizations: Guyon Duifhuizen, Max Borrmann and Rob Simmon

This harvest season, global eyes are watching US crop reports with anticipation after fields around the world have been challenged by severe weather, droughts, heat waves and political turmoil. The United States is the world’s largest producer of corn and the second largest grower of soybeans, and news regarding production yields can affect global prices for the entire season. Some recent reports have indicated that corn yields in the United States could be the lowest since the drought-ridden 2012, and that corn quality could be affected as a result of hot, dry conditions. However, erratic weather patterns can make predictions difficult, and reports can vary widely. Given these challenges, Planet’s Rield Forecasting solution offers a technologically advanced way to help farmers and economists better prepare for the future.

What we see in the future in the 2022 harvest season

With a variety of unusual weather events, 2022 has already proven to be a challenging year for growing crops in the US Midwest. However, unlike some national agricultural productivity reports, Planet’s Rield Forecasting solution sees some late-season improvements for corn and soybeans. With projected yields of 177.6 Bu/acre for corn (Fig. 1a) and 51.1 Bu/acre for soybeans (Fig. 1b), near the final yield of 2021, these crops could provide some much-needed relief for an already strained global market. .

Compare predictions

The U.S. Department of Agriculture (USDA) released its latest forecast for September for this season: 172.5 bushels per acre (bu/acre) for corn and 50.5 bushels/acre for soybeans, significant declines compared to the August forecast.

However, as we already see with planet predictions, the market perception can differ from the reports.

In early August, the situation was completely different. The corn forecast matched well between USDA (175.4 po/acre), market perception (175.9 po/acre) and Planet (176.0 po/acre). For soybeans, USDA still forecast near-record yields (51.9 pu/acre), while market perception (51.1 pu/acre) and planet forecasts (50.4 pu/acre) were calmer.

Figures 1c and 1d visualize how USDA and planet predictions compare historically with actual final yields. The discrepancy in these numbers is an example of the associated uncertainty that underlies these expectations. This uncertainty can translate into market volatility and an overreaction to short-term weather events. Therefore, to put these planet predictions into perspective, it is important to understand their basics.

Figure 1a, Time series image of the 2022 US National Maize Forecast throughout the growing season. The root square error (rmse) is a measure of a model’s uncertainty, by calculating the average of the absolute differences between the yield values ​​predicted by the model and the final yield values ​​for the 2003-2021 historical runs.

Figure 1b, time series image of the 2022 US National Soybean forecast throughout the growing season.

Figure 1c, US National September corn yield forecasts from both Planet and USDA, compared to final yields.

Figure 1d, US National September soybean yield forecasts from both Planet and USDA, compared to final yields.

Analyze how revenue forecasting works

To understand what might be behind these trends, it is useful to first understand the underlying data archive that underpins our statistical prediction model. The US national corn and soybean yield forecast in season is based on some of our unique set of remotely sensed planetary variables including:

  • Soil water content Measure the volume of water in the first 5 cm of soil
  • Earth’s surface temperature Measuring the radiant skin temperature of the Earth’s surface at night
  • visual depth of plants Microwave-based botanical evidence relating to the water content of plants, or simply, an indicator of vegetation density

Next, let’s delve into the factors that influence our predictions. We’ll take a deeper look at this year’s planetary variables in Illinois, one of the top corn and soybean producing states. Spring began with unusually wet conditions, which can be seen from the soil water content data (Fig. 2a). These conditions are very unfavorable for farmers to start planting activities, causing a delay in the start of the growing season – about 15 days on average when looking at emergence data from the USDA over the same state (Fig. 2d). These emergence data describe the percentage of newly planted crops emerging from the soil.

Although delays do not necessarily result in lower crop yields, they can increase the risk of drought and heat problems and harvest later in the season. It also adds to the uncertainty of all currently available forecasts, with the growing season being up to half a month later than the early emergence years.

Here’s what distinguishes Planet’s Yield Forecasting solution from others: Due to its twenty-year historical archive, it can provide natural insights into this uncertainty over time (see Figure 1) and into space (Figure 3). This ultimately helps agronomists and economists see long-term trends to support balanced decision-making versus overreaction to short-term weather events.

Figure 2a, chronological image of soil water content in Illinois, showing an unusual wet period in early spring 2022.

Figure 2d, Illinois USDA emergence dates, with an apparent delay of 2022 at the start of the season.

After the late emergence, the crops experienced a widespread heat wave in June that can be seen in the land surface temperature data (Fig. 2b), along with reduced precipitation. This is an unwelcome combination that resembles the patterns of 2012, a year that saw record low returns. While fortunately some relief was seen afterwards, with more favorable climatic conditions in many areas, it is difficult to determine the exact impact of this on crop yield. However, what we see in soil water content is that moist early-season conditions are likely to provide a valuable buffer during the summer heat wave.

Figure 2b, Illinois night surface surface temperature time series image, showing a sustained period of high anomalies in 2022 from the second half of May through early August.

This interpretation is supported by a third variable, vegetation visual depth (VOD). VOD is a coarse-scale microwave-based vegetation index that correlates closely with vegetation water content, that is – the amount of water retained in a plant’s land surface structure, such as the stem and leaves. This regional signal (Fig. 2c) shows that the previously recorded delay in plant emergence may be largely offset by a strong increase in plant growth in July.

Figure 2c, optical depth time series image of plants in Illinois, with stronger-than-normal increase in biomass in July 2022.

Although this information is indeed valuable, providing unique insights into crop conditions throughout the region of interest, this does not yet give us the quantitative production value of the Midwest that everyone from farmers to investors seek. So, how can we understand all this in a quantitative way?

Why is this beneficial for farmers and investors

Each year’s crop is the result of an incredibly complex combination of factors. Genetically improved plants grow across different local climates on nearly 500 million acres of arable land. By capturing this multi-dimensional view of crop health in near real-time, without cloud interference, Planet can provide a forecast of the crop every day of the growing season. This is done using an algorithm that makes optimal use of Planet Archive records, giving our statistical models an unparalleled set of data to correlate with historical revenue. The final results are based on the aggregation of more than a thousand smaller models, which are out-of-sample multivariate regressions, trained daily for all individual counties in the state. This means that these models are able to find the best combinations of the most influential planetary variables and their exact time in the growing season. As more information becomes available throughout the season, the model will be adjusted accordingly and begin to include PV data from newly available key moments. Taken together, it sums up to over 2 million typical improvements per season per crop leading to these daily national outlook values!

Figure 3, Spatial map of maize forecast uncertainty on September 8, 2022, based on historical simulations of same-day crop yields in the growing season (the beginning of which depends on the day when the 50% crop is on this case). The value of the uncertainty is expressed as root square error (rmse) as a percentage of the trend return.

As we initially shared, with a variety of unusual weather patterns, 2022 has already proven to be a challenging year for crops in the heart of America. But Planet’s Yield Forecasting solution sees some late-season improvements that raise our forecasts to 177.6 bu/acre for corn and 51.1 bu/acre for soybeans compared to the USDA reports. These values ​​are similar to those of 2021, and could provide some much-needed relief to an already strained global market.

Leave a Comment