It is intuitively obvious that soil conditions, particularly temperature and moisture content, play a critical role in the efficiency of agriculture, affecting all decisions from planting through harvest. As such, many agricultural and weather service providers are starting to offer measurement solutions and/or model-based estimates and forecasts of soil temperature and moisture.
Recently, meteorologist Tim Marquis posted an excellent blog outlining some of the challenges in forecasting soil conditions, especially with regard to extracting soil forecasts from the land surface models (LSMs) commonly available from public numerical weather prediction (NWP) models that are distributed by government weather services or run internally by numerous private weather companies. I would like to expand on his list to not only include forecast soil conditions, but also current, historical, and climatological values, whether measured or modeled, particularly in the case of soil moisture. In fact, it is nearly impossible to effectively use a numerical value of soil moisture without the context of climatological values from the same source of information at the same location and resolution. This is because for a given patch of soil, two different measurement techniques may yield substantially different results. Likewise, even with the same weather input and soil taxonomy information, two different LSMs or even the same LSM run at different resolutions may provide drastically different results, even though trends may be similar (see Figure 1). Thus, the only real solution is to understand the current measurement or forecast in the context of the model’s response over a long history, which can then be correlated with field performance, irrigation activities, and moisture-sensitive diseases for the same periods. This issue has been a challenge for weather modelers for many years when trying to initialize a regional, high-resolution forecast model’s LSM from a coarser model that has different land use, resolution, and potentially even different physics in how the LSM functions.
Figure 1. Soil moisture for the same location for two different land surface models running at the same resolution using the same weather forcing.
So, given this and all of the other issues raised in the referenced blog, is hope lost for having meaningful soil temperature and moisture forecasts using land surface models? Absolutely not! In fact, at DTN, we agree that the best answer to solve these problems is through collaboration with experts in the various fields. So, our scientific staff consists of agronomists, land surface modelers, meteorologists, and computer scientists, all working together to solve problems like these.
First, rather than rely on soil forecasts coming from NWP models, we have implemented two state-of-the-science LSMs within our own infrastructure, providing full control over how we configure and apply the output. Doing this allows us to drive the models at whatever resolution we need, limited only by the precision of the soil and vegetation coverage information. Furthermore, we are able to run these models from our global 35+ years of historical weather data using the exact same configuration as our real-time forecasts, and we re-accomplish this every time we make an improvement to the system. This allows us to present the soil moisture conditions relative to the model’s own climatology, giving an easy indication of how recent, current, and forecast moisture compares to a relevant climatology. For example, Figure 2 shows the output of our system for my location. We had a wetter than normal July, and you can see where after receiving over 1” of rain on the 21st and 22nd of July, my soil moisture reached 100% of the climatological maximum (likely equivalent to the soil saturation point in this case as well).
Figure 2. Soil moisture history and forecast for my house from July 8th, 2015 through October 17th, 2015. Orange vertical line is the dividing line between past and forecast information.
But, what about all of the other concerns that were raised? Those are legitimate issues, and we have addressed those, too. Because we run our own LSMs, we have been able to optimize the configuration specifically for agriculture in many ways:
– We have increased the number of vertical layers and decreased their thickness to better handle different rooting zones, as well as to better model nutrient leaching and provide improved field accessibility forecasts. Most “standard” LSMs have a 10 cm top layer, whereas we currently have a 2 cm layer.
– Rather than rely on standard land conditions and associated physics, our ClearAg agronomists have developed replacement coefficients based on crop type and growth stage to account for root depth and density to better simulate uptake and evapotranspiration of moisture from the soil. This is in contrast to the standard LSM configuration used in the various publicly available NWP models where crop height, leaf area, and rooting depth are the same for all cropland.
– We have developed a field-specific version of our LSM system, in which we can override soil and vegetation information based on user-provided data, and in turn this allows proper selection of the crop coefficients as described above. Even if the grower is unable to provide more detailed information on their field, we can still estimate crop growth stage from our internal crop growth and health models provided we at least have a crop type and estimated planting date.
– The field-specific modeling capability accounts for irrigation activity, whether reported by the farmer or in the future by receiving data directly from the irrigation equipment. Thus, the soil information is updated by both natural rainfall from our advanced eMPower weather content as well as the grower’s activity. In the case of fields having soil moisture sensors, we can leverage those in our machine-learning capabilities to further calibrate the forecasts, although measurements of soil moisture bring additional complexities beyond those discussed here.
So, not only does hope exist, but science-based capabilities are a reality today and are ready to be delivered via our ClearAg application, as stand-alone components, or via APIs. ClearAg scientific staff have creatively joined forces to combine detailed agronomy with advanced soil modeling and leading weather content to help growers, crop consultants, retailers, and agribusinesses be more efficient and productive.