Industry Sampling Method
Error in sampling methods is moving Nutrient Results across the Maintenance Band
The Playing Field
An even footing for testing.
- Conventional Tillage Field in Northeast Indiana (Unnamed for privacy of farmer)
- Corn / Bean Rotation
- Sampled in Wheat to Beans in August, 2022
- 21 Samples on 2.5 ac grid
- Field Conditions: medium dry soil inconsistent ground hardness. Decent texture for pulling
- These field conditions were about the best you can get for an auger
- And quite good for the Wintex as well.
- This field had pretty good fertility levels overall
Robot – ROGO Robot sampling.
Hand-Probe – Trained operator using a hand probe at a normal speed, trying for accuracy
Auger – Standard auger operated by a 3rd party professional sampling company (Not ROGO)
Wintex – Wintex 1000 ran by a trained operator
High Quality Hand Check
Attempted Very Slow, Near-Perfect Hand probes as a check in the study
- Nothing should theoretically be better than a perfect core, every single time
- Discarded bad partial cores when they occurred (about 10% of the time)
- Depth was not perfect (sometimes slightly too deep) due to inconsistent ground hardness, but very good
- Unsustainably slow pace to achieve this accuracy (maybe ~100 acs per day)
The robot outperformed every other method by a large enough margin to move across the maintenance band and move the needle on fertilizer applied (pH the most, then K, then P), validated by the Check trial. Of non-robot methods, hand-probing and augering were similar and better than the Wintex.
How many samples had significant error?
Significant error means likely to move across the maintenance band. This is defined as an error greater than ⅔ of the maint. band, since that could easily “trick you” into thinking you are in the maint. band, but are not.
The other methods have meaningful amounts of samples with significant error.
On P, it ranges 20-40%. on K, it ranges 10-40%. on pH, it ranges 50-80%.
Absolute Error Percentage
- Error % is the difference from truth divided by the truth
- This ranged from 10-32% on P, 4 to 15% on K, and 1 to 5% on pH
- The robot performed the best by a factor of 2x+ vs any other method
- The auger and handprobe performed similarly. Wintex was last place, especially on P
Data Deep Dive
- P had the most difference between the Robot and the Hand-check, presumably due to depth on the hand-check not being perfect
- Wintex was wildly off on P, several times
- The difference between the robot and the other methods was least pronounced on P
P-Distribution of Sample Point Error by Method
- The bigger the box and the whiskers, the larger the error
- The vertical position of the boxes indicates their bias (high vs low)
- The boxes represent 50% of the data range, the whiskers represent 95% of the data range. Points outside of that are the full minimum and maximum
Check(Blue) and Robot (Green) track very closely
K-Distribution of Sample Point Error by Method
- The Robot was very tight
- Among the others, the auger led, followed by the hand probe, followed by the Wintex
- Robot(green) and Check (blue) almost right on top of each other, Closest of all 3 traits
- The others generally track with them, but also several segments of samples where they are significantly off
pH-Distribution of Sample Point Error by Method
Very clear visual of how accurate the robot was, compared to the others
Actual A&L Great Lakes raw lab data available upon request.
a New way to sample soil
Build a Modern Fertility Program with SmartCore™
Robotic soil sampling increases sampling precision and accuracy, and allows for farmers and ag retailers to build accurate fertility management plans.
Get better insight to ensure the fields are being sampled, and to your specifications.
Get the job done with a reputable company and technology.
Every core is taken at the exact depth and accurate location to ensure quality samples.