Becoming carbon-neutral and offsetting an organization’s carbon footprint is a top priority for organizations this year. Yet, it can be difficult to find credible carbon market programs and understand how these offsets are created. To help organizations figure out how to build carbon-neutral programs into their agriculture supply chain, we went back to the basics behind carbon credits – crop modeling in our recorded webinar “Transforming the Future of Ag with Crop Modeling.”
What is Crop Modeling?
A crop model is the set of mathematical quotations describing how crop genetics, crop management practices, and the environment interact together to determine crop growth. Think of a crop model as a framework that integrates knowledge across several disciplines. It is a mathematical representation of the world. Crop models bring a consistent framework to show how a crop grows and responds to the environment and management practices. At CIBO, we use crop models to predict the crop growth stage, when a crop matures or flowers. The model can calculate how the crop will respond to the temperature which will drive the development rate of the plant.
At CIBO, we have a deep library of public, private and proprietary data we use for crop modeling. The data brings the model to life and provides “what-if” scenarios of how the crop grows. The data we use from growers is for yield, biomass and even some of the important management information like planting dates, amount of fertilizer applied, row spacing, etc. Our model requires a fair amount of soils data (deep soil data) from a particular field. Some of this data is not available from growers so we rely on the SSURGO (Soil Survey Geographic Database)—a web soil survey database.
Watch our recording to hear our detailed discussion of all things crop modeling. We explained how crop models are used for regenerative agriculture, what makes CIBO’s crop models unique, how we use them and took questions from our live audience. See some of the questions and their answers below!
Questions from the Audience
Which soil database do you use to simulate your crop with the SALUS crop model, and how do you handle the variability of soil physical and chemical properties among soil types and across large areas?
We use the SSURGO soil database and run SALUS on different soil polygons per parcel, then aggregate model outputs across soil polygons. We know using SSURGO does not account for soil topography and subfield variability, so we have been working on improving the soil properties at the initialization of the model and a downscaling method to account for subfield variability.
Which model do carbon credit purchasers (collectively) tend to prefer: “pay for practices (model-predicted SOC changes)”, or “pay for performance (actual measured SOC changes)”?
Model predictions and physical measurements are two ways of estimating actual SOC changes, and both have uncertainties associated with them. Carbon credit purchasers strongly prefer carbon credits that are certified by a third party, such as CAR, Verra, or Gold Standard. These agencies have strict requirements for measuring and accounting for the uncertainties associated with estimated SOC changes.
If you have a ten year crop yield history, can you realistically predict the current year’s yield? Updated daily based on the current year’s fertilizer program, soil types and actual weather patterns?…To clarify…If you have only a ten year crop yield history and access to NOAH weather history and USDA Soil types… AND you want to predict the current year and see daily updates based on predicted and actual weather results.
When using a crop model to predict field-level crop yield using weather forecasts, the yield accuracy will heavily depend on the uncertainty in the weather forecasts. Early-season weather forecasts may vary widely, leading to a large uncertainty in resulting yields. As the season progresses and more of the weather is measured rather than predicted, the accuracy of yield prediction will increase. The ten year crop yield history would have been used to calibrate and validate the model first.
How is the interest of growers about crop models?
Growers are generally fascinated by what crop models can do and understand the model limitations in terms of accounting for subfield variability, which can generally be addressed by integrating remote sensing or user-supplied data into model simulations. There are many challenges to establishing grower confidence in crop model results. When building that trust, caution should be taken not to overstate model performance or capability while emphasizing model uncertainty.
Can you incorporate any disease effect? For example, this year Downy Mildew is affecting some varieties in the south… Can you also cover those factors, to some extent, in sensitivity analysis?
There are a number of ways to model disease effects on crop growth and yield. In the US, these effects are usually mitigated by treatment (fungicide application in the case of Downy Mildew). For this reason, we do not currently include modeled disease effects in our estimates for US row crops. As we move into other cropping systems and geographies, this may change.
However, other crop models account for disease effects in a simple way by modeling the actual disease impact on the crop rather than modeling the biological dynamics of disease evolution. This approach would typically parameterize the disease susceptibility of the host cultivar, then the disease onset and severity are calculated based on environmental factors like temperature, relative humidity and leaf wetness. The disease severity is combined with the crop susceptibility to modify crop leaf area (which is the main tissue pests are likely to attack), which propagates the disease effect to crop growth. More sophisticated approaches include modeling pathogen growth across different population classes, and progressive crop disease stages from healthy to fully infectious. The most important thing is to include the elements of the disease triangle and apply the disease impact to the right plant tissues.
How do you calibrate the model if you do not have long-term data like SOC for the past five years?
For SOC-related applications, we need some form of SOC data for performance assessment. In general, there are many published long-term SOC studies in the literature with more than one-time measurements. We use such studies to assess the performance of Salus to predict SOC change.
If a region has several dominant cultivars, how do you pick which one needs to go into the model?
First, we identify the cultivars that account for a high proportion of land use in the region (e.g. five cultivars may account for 80% of land use). Some of those cultivars may have such similar inherent performance that they can be treated as equivalent. The unique cultivars are then characterized by their crop parameters. Finally, we run the simulation model for each unique cultivar and aggregate the results.
If you are interested in learning more about Crop Modeling, watch our recorded webinar, where we cover everything in more detail!
