Precision Agronomy and Management Zones
High resolution, quality management zones are a power-house of ag data. Properly designed and executed, management zones can be utilized to maximize ROI of numerous crop inputs:
- Seed
- Fertilizer
- Soil Amendments
- Pesticides
- Irrigation
When zones reflect the various microenvironments within a field, they can even be leveraged to guide early-season scouting before NDVI and aerial mapping become a viable option.
However, there is no industry standard for management zone creation, and many data sets, often with questionable value, are employed for the purpose of zone creation. This can cause farmers to inadvertently make mistakes that they wouldn't even make with conventional methods; such as increasing seeding rates in an area that *should* have dramatically lower rates, as is the case in soybeans where the disease white mold is a concern.
SSURGO soil survey map units. The scale at which soil variability is actually mapped is grossly insufficient to be used in a site-specific fashion. Of course, that hasn't stopped many from trying to leverage this "free" data in management zone creation.
Above: zone delineation with SMART4D at left, soil survey map units at right.
For decades, yield data has been largely seen as the holy grail of ag data. In practice, however, it leaves a lot to be desired in terms of actionability and agronomic relevance. Sure, yield maps are a great tool for financial analysis (such as profit mapping), but far too often they don't answer the question of 'why did this area yield this way'. Instead, yield maps simply show what came off the field and where, assuming they are properly calibrated to begin with.
NDVI maps, if properly timed, can serve as a proxy for yield data in many crops. However, they are not well suited for quantifying specific yields (bu/a, t/a, cwt/a etc) and instead at best allow agronomists to infer relative yield differences between regions within the same field. Like yield maps, NDVI does not tell the 'why' of the story and without that context, farmers are not well equipped to address the issues at hand. A common example in yield and NDVI maps is the misclassification of excessively wet and excessively dry regions; both look similar in these datasets, absent another layer (or ground-truthing) to verify.
In the above example, the Spring NDVI classified wet soils the same as dry soils. The colorized map on the left is the zone map, which accurately characterizes soil moisture, as seen in the resulting crop development above
Elevation data, in its raw format is simply distance above sea level. In some situations, this can line up with soil moisture, but this can vary widely from region to region or even field to field. Therefore, using raw elevation for creating management zones can be agronomically questionable. Fortunately, this same data can be used to calculate scores of different models that may be more relevant. Slope, aspect, closed depressions, watershed modelling and other outputs are common examples.
Grid soil sampling, which has also been in practice for decades and still quite popular in many regions, does have unique value over other means of delineating and quantifying field variability. The main problem with grids, however, is resolution. Standard grids are taken at 2.5 acres, or 330 ft between sample points. Most fields, depending on the region, can experience significant fluctuations in natural and man-made variability in this span. In local demonstrations (central MN), we have had to go down to 30 ft between sample points before we can accurately represent changes in the soil. The investment of labor in sampling at 30 ft intervals is cost-prohibitive for most broad-acre farmers. However, with high-intensity grid sampling, we can do a better job than high resolution soil mapping technologies (EC, EM, optical sensors) at delineating certain micronutrients, pH and phosphorus variability in the soil. Unfortunately, the labor investment and lab costs are prohibitively expensive for most crops at this resolution.
Electrical conductivity mapping (and other terrestrial-based sensors) is a significant improvement upon the resolution of standard grid sampling practices, with swaths usually set 40-60 ft apart and data points collected approximately every second. All things being equal, EC mapping can be a great way to infer soil texture as clays will often be more conducive than sands, for example. However, all things are not equal in the field and EC can also be driven by salinity, soil moisture and some extent temperature. The best use-cases for this technology are regions with soil salinity issues, deep-rooted crops (where we are actively managing inputs below the top ~1 ft, perennial crops, and fields that do not have high quality soils imagery.
Optical sensors exist in both terrestrial and aerial platforms. Where they are used in aerial platforms and as a source for radiometric analysis, optical sensors represent the highest resolution soil variability mapping technology the industry has to offer. Depending on the sensor, altitude and other variables, 3 ft resolution (or less) can often be achieved for a fraction of the cost of terrestrial-based sensors. This is a powerful dataset for efficiently and accurately delineating and classifying soil variability. However, this is not a one-size-fits-all tool, since the conditions at sensing can impact the quality of the data obtained. Fortunately, with experienced and knowledgeable precision agronomists, farmers can reap great value from this inexpensive, high resolution soil variability data. For example, it is possible to accurately infer relative changes in soil moisture, organic matter, and other changes in the soil before even stepping foot into the field.
Finally, identifying the right dataset for your grower, crop, region and specific yield-limiting factors is only half of the challenge. How you combine these layers, account for man-made variability, and execute the zone management program will ultimately determine the success of your precision ag platform.
Owner, Hinkson Land Tech
3yWe found in some cases that EC data in “micro topography” and EC data in “macro topography” are inversely correlated which goes to the point brought up that any one data layer is not necessarily a good management zone.
Nice article, Michael. Do you also combine ground truth technology?
Precision Ag CEO @ GeoPard Agriculture | Precision Agriculture | Software Engineering | API | Automation | Sustainability
3yGreat article Michael, as always! You know you are an advanced precision ag specialist if you use Geological software! 😁 Which imagery do you use to map bare soil? We see a big shift into understanding the value of several data layers to create Zones for a field (we at GeoPard call it multi-layer analytics). EC/Topography (about 10 models)/Multi-year Vegetation/Yield/Soil sampling/EC/EM/Bare soil - the number of possible combinations only from listed layers is 40320 (8 factorial). And I'm not even mentioning setting up of different weights per layer, which create almost uncounted amounts of options.
Experienced Innovative Product Development Leader
3yThis is a great overview, Micheal. Thanks for posting!
Head of Agronomy and Digital Sales Enablement - Simplot Grower Solutions
3yVery informative article, thank you