arrow right cross menu search arrow-left arrow-right menu open menu close plus minus

In-Season Ag Planting Dates at Scale: An Illinois Case Study

By Shane Bussmann

Planting is a critical time for farmers: If you plant too early, a late freeze or frost could ruin your entire crop; plant too late and the risk of reduced yields rises in the face of a shortened growing season. Yield is the most important outcome for many farmers, those who provide services to farmers, and much of the commodities market as well. It stands to reason that understanding planting dates, not just for a particular farm, but across regions and at scale offers a critical clue to the status of coming prices, demand, business opportunities and the entire farming ecosystem. 

One of CIBO’s key differentiators is the ability to estimate planting date for corn and soybeans at the individual parcel level across the corn belt. In this blog post, we discuss why estimating planting dates is so important and we explore our methodology for determining them.

Planting Dates in IL

Figure 1: Planting date at the parcel level across all of Illinois (a handful of counties appearing fully white are missing due to the absence of parcel data). This map highlights regional trends in planting dates, such as the shift in corn planting date from April and early May on the northwest side of the Illinois River to late May and June on the southeast side of the river.

Why Is Planting Date Important?

Planting date is an important component of the in-season yield trendline for a field. Knowing the in-season regional yield trendlines can give a farmer and an ag-related service business the competitive edge in the agricultural marketplace. For example, supply is likely to be lower if regional-yield trend lines indicate lower than usual yield. If demand remains constant, then prices for goods are expected to be higher. Armed with this knowledge, a farmer will have more insight when making in-season management decisions, calculating land valuation, and other applications. Armed with this knowledge, ag-service businesses such as input suppliers, bank and credit providers, transportation providers, and elevators (the list goes on) can better anticipate demand for their services and the capacity they should plan on.

Despite the importance of planting dates, publicly available data is limited. The USDA publishes planting progress percentages for each state on a weekly basis. This data can be used to infer the range of typical planting dates across the state, but it fails to provide the granularity needed for an accurate regional assessment of planting date. Farmers with strong social networks can sometimes fill in the gaps through word of mouth, but this remains a challenging task and it neither scales nor makes the data available to all who are interested.

An alternative solution is now available from CIBO. Our data scientists, crop modelers and ag specialists work together to use computer vision and data from remote sensing satellites to determine the most likely planting dates for crops across the US. In the remainder of this article, we detail how we at CIBO use leading-edge technology to estimate planting date measurements at the parcel level.

Using Remotely Sensed Data to Infer Planting Date

Our work begins with phenology curves, which are critical to determining planting dates. A phenology curve is a set of vegetation index (VI) measurements that chart a crop’s life stages over the course of a growing season. There are many types of VI measurements, but they are all designed to approximate the growth of vegetation. A commonly used VI is the enhanced vegetation index (EVI), which relies on blue, red, and infrared wavelengths of light and helps most vegetation stand out from non-vegetation like roads or buildings. This is because vegetation appears dimmer in red light than near-infrared light compared to non-vegetation. By looking at the EVI phenology curve, we are able to determine when crops emerge from the ground and even when they mature. Combining this with ground truth planting date data over a range of distinct geographical locations, we can infer the most likely planting date in any given field. Of course, the trick is doing all this for every parcel in the U.S where corn or soybeans are grown. 

The VI measurements that comprise the phenology curve can be obtained from a variety of sources, such as unmanned drones, airplanes, or space satellites. Of these data sources, only space satellites are capable of covering the entire planet. The LANDSAT and Sentinel satellite missions provide imaging at blue, green, red, and infrared wavelengths, with 30m and 10m resolution pixels, respectively. These satellites’ orbits allow them to pass over any given spot on the Earth up to twice a week. In addition, the satellites’ data is publicly available at minimal cost, making them viable options for creating phenology curves at the parcel level across the entire country. 

Figure 2 EVI images taken by the Sentinel satellites of a parcel of land in Illinois

Figure 2: EVI images taken by the Sentinel satellites of a parcel of land in Illinois for two dates within the same year. Light green indicates less vegetation, darker green indicates more vegetation. Roads and buildings appear as white or very light green and do not change over time. Forest and grassland become green earlier in the year than land used for agriculture.

The planting date on a given parcel of land has a strong impact on the phenology curve associated with that parcel. EVI begins to rise when a crop first emerges from the ground. The later the planting date, the later the characteristic rise in EVI occurs. Another key factor is the type of crop planted. If soybeans and corn are planted on adjacent parcels that are materially similar, then the parcel with corn will show an earlier rise in EVI—only to see soybeans reach a higher peak value later in the growing season. 

At CIBO, we use a proprietary algorithm based on the principles of machine learning and Bayesian inference to solve for both the crop type and planting date. Bayesian inference is a statistical modeling approach that allows us to extract the maximum amount of information from the available data. It also provides confidence intervals for the crop type and planting date. For example, in the model fit to the EVI curve shown here, the crop type is corn with 99% probability and the most likely planting date is May 3, 2020, give or take a few days. 

Figure 3 EVI curve made from Sentinel images in 2020 of the parcel in Illinois

Figure 3: EVI curve made from Sentinel images in 2020 of the parcel in Illinois shown earlier in this post. The EVI values begin to increase in late May as the crop emerges from the ground. The EVI values reach a maximum of around 0.65, typically signifying the presence of corn rather than soybeans. Model curves from our proprietary algorithm for corn (orange) and soybeans (blue) are shown, indicating that corn provides a much better fit to the EVI data, which is consistent with expectations. The estimated planting date for this parcel is May 3, 2020.

We apply this algorithm to infer corn and soybean planting dates for all parcels in the corn belt and are in the process of expanding to additional crops and additional regions across the United States. Through our science-based approach to understanding the land, we are able to better deliver decision-ready insight to anyone who has an interest in or provides services to the agriculture space. Planting date data is now available in our August, 2020 county-by-county yield forecast report. People and businesses that rely on a detailed understanding of when crops are likely to be ready for harvest, sale, transportation, storage and processing should be very interested in seeing this data. We invite you to register for CIBO to have this information at your fingertips. Start now, it’s free.

In this blog post, we have shared the results of this effort for the state of Illinois. We believe this approach to be a potentially powerful tool in the hands of the right user, providing crucial information that will help farmers and related businesses across the country take their work to the next level.

About Shane Bussmann 

Shane Bussmann is a Senior Data Scientist at CIBO, a science-driven software startup. Prior to CIBO, he worked as the Lead Data Scientist for Understory, Inc. He holds a B.A. in Physics and Astrophysics from UC Berkeley, along with a Ph.D. in Astronomy from the University of Arizona.



Scroll to Top