GeoPard Tutorials
Ask or search…
K

Trial Analytics

Agronomists employ Trial Analytics to assess the performance of various crop varieties, cultivation techniques, and input applications, including the results of Variable Rate Applications in Precision Agriculture. By collecting, analyzing, and interpreting data generated from Field Trials, researchers gain insights into the interactions between genetics, environment, and management practices. This knowledge informs the development of crop management strategies that optimize yield potential while minimizing input usage. Furthermore, Trial Analytics not only enables the evaluation of the effectiveness of Precision Farming practices but also helps in identifying resilient crop varieties that can thrive under diverse and challenging conditions, thereby contributing to food security.

Data Preparation

For effective trial analytics, a few essential datasets are required:
  1. 1.
    Yield Dataset: This dataset captures the yield data. We can import this from the JohnDeere Operation Center or manually upload it as shapefile or as machinery proprietary format.
  2. 2.
    Application Dataset: This is crucial for understanding the actual executed Application on the Field. At a minimum, it contains attributes like TargetRate, AppliedRate, and some machinery-related metrics. As with the Yield Dataset, we have options to import it from the JohnDeere Operation Center or manually upload it as shapefile or as machinery proprietary format.
  3. 3.
    Zones/Plots with Trials/Experiments: These show the planned Application rates for our Trials, giving insight into the experimental design. If such a data layer is available, we upload it as shapefile into the AsApplied/AsPlanted or Yield control. This ensures compatibility when building EquationMaps, streamlining your trial analytics experience. If such a data layer isn't available, the TargetRate attribute from the Application Dataset can serve as a substitute for Trial evaluations.
  4. 4.
    Historical Field Potential Zones: These zones are generated by GeoPard (details are HERE). They are useful for analyzing Trials with consistent historical productivity. This is particularly beneficial when Trials are distributed across regions with varied historical productivity.
Once we've gathered these datasets, the next step is starting the Trial evaluation process.

Data Overview

There are the following data for the 2023 agricultural season of winter wheat:
  • Yield Dataset highlighting Wet Mass distribution (Fig.1)
Fig.1 Original YieldDataset
  • Nitrogen (N34) VRA plan (150 kg/ha) with 2 Trial Plots (120kg/ha and 180 kg/ha)(Fig.2)
Fig.2 Nitrogen (N34) VRA plan with Trial Plots
  • Application Dataset showcasing applied statistics (Fig.3)
Fig.3 Application Dataset
  • Historical Field Productivity (Fig.4)
Fig.4 Historical Field Productivity
The YieldDataset hasn't been calibrated: multiple harvesters were operating there, there are turnaround and missing data traces, and noise is evident. It's recommended to apply the Yield Calibrate and Clean operations on top for optimal results. A step-by-step tutorial can be found at LINK.
The YieldDataset, after calibration and cleaning, is displayed in Fig.5, along with the updated statistics. This dataset will be utilized in subsequent steps.
Fig.5 Calibrated and Cleaned YieldDataset

Concept

Here the objective of Trial Analytics is to ascertain the most effective Nitrogen (N34) rate for the field. There are the pinpointed areas with Nitrogen rates of 120kg/ha, 150kg/ha, and 180kg/ha. This data is derived from the ApplicationDataset on one hand and the calibrated YieldDataset on the other.
We're focusing our analysis on three distinct zones:
  • 120kg/ha (designated as the trial zone)
  • 150kg/ha (considered the main zone)
  • 180kg/ha (another trial zone)
Our approach will include the following evaluations:
  1. 1.
    Plan-based: using the Planned Variable Rate Application (VRA) linked to the Calibrated Yield.
  2. 2.
    Applied-based: Comparing the Actual Applied Datasets against the Calibrated Yield.
  3. 3.
    Applied-based and Historical Productivity: Comparing the Actual Applied Datasets against the Calibrated Yield overlapped with Historical Field Potential Zones.
This methodical approach will allow for a comprehensive evaluation of Nitrogen's impact on Yield, based on both Planned and actual Applied application data.

Plan-based

The influence of applied planned Nitrogen (N34) on Yield distribution is visually captured in the subsequent screenshots (Fig.6, Fig.7, Fig.8). Here's a concise breakdown of the findings:
  • The main zone, with a Nitrogen rate of 150 kg/ha, spans 45.8 ha and averages a yield of 4.99 t/ha (Fig.6).
  • The first trial zone, utilizing a 180 kg/ha Nitrogen application, covers 1.76 ha, yielding an average of 6.5 t/ha (Fig.7).
  • The second trial zone, with 120 kg/ha of Nitrogen, encompasses 1.86 ha and produces an average yield of 6.39 t/ha (Fig.8).
The results prompt a significant query: Why does the lower application rate seem to be more efficient than the higher one? To gain deeper insights, the next phase involves evaluating the Trials using the actual Applied data.
Fig.6 Main Zone with N34 150kg/ha
Fig.7 Trial Zone with N34 180kg/ha
Fig.8 Trial Zone with N34 120kg/ha
Further down, you'll find an in-depth discussion of the formulas and configurations employed during the evaluation.
To delve deeper into the Equation approach and its execution, please consult our tutorials for both the User Interface and API.
Here are the Equations to run to reproduce the calculations.
  1. 1.
    Main with 150 kg/ha: Yield_Main = np.where(Zone==1, Yield_WetMass, np.nan)
  2. 2.
    Trial with 120 kg/ha: Yield_Zone = np.where(Zone==3, Yield_WetMass, np.nan)
  3. 3.
    Trial with 180 kg/ha: Yield_Zone = np.where(Zone==2, Yield_WetMass, np.nan)
It is important to activate Numpy (Fig.9) and turn off Interpolation (Fig.10).
Fig.9 Activate "Numpy"
Fig.10 Turn off using of "Interpolated" data

Applied-based

A notable observation is that the actual Applied Rate during the Trial doesn't consistently align with the Planned (Target) Rate. Specifically, a distribution ranges from 120 kg/ha to as high as 189 kg/ha (Fig.11). Given this variability, it became crucial to set a benchmark for error tolerance. Thus, a ±5% accuracy was determined to be an acceptable threshold to deem the trial suitable for evaluation.
Presented in the subsequent screenshots (Fig.12, Fig.13, Fig.14) is the statistical distribution of the Yield, focusing on the Nitrogen (N34) actually Applied numbers. Here are the summarized statistics, keeping in mind the ±5% accuracy acceptance:
  • The main zone at 150 kg/ha had an applied area of 43.5 ha, yielding an average of 4.9 t/ha (Fig.12).
  • The first trial zone at 180 kg/ha covered an area of 1.47 ha and produced an average yield of 6.5 t/ha (Fig.13).
  • The second trial zone set at 120 kg/ha spanned an area of 1.44 ha, with an average yield of 6.3 t/ha (Fig.14).
Fig.11 Actual Applied Rates in Trials
Fig.12 Main Zone with N34 150kg/ha ±5%
Fig.13 Trial Zone with N34 180kg/ha ±5%
Fig.14 Trial Zone with N34 120kg/ha ±5%
For a deeper understanding of the methodology and the specifics of these results, the used Equations are below:
  1. 1.
    Trial actual Applied Nitrogen: Applied_Trial = np.where((Zone == 3) | (Zone == 2), Applied_Value, np.nan)
  2. 2.
    Main with 150 kg/ha incorporating 5% acceptance: Yield_Main = np.where((Zone == 1) & (Applied_Value >= 142.5) & (Applied_Value <= 157.5), Yield_WetMass, np.nan)
  3. 3.
    Trial with 120 kg/ha incorporating 5% acceptance: Yield_Trial = np.where((Zone == 3) & (Applied_Value >= 114.0) & (Applied_Value <= 126.0), Yield_WetMass, np.nan)
  4. 4.
    Trial with 180 kg/ha incorporating 5% acceptance: Yield_Trial = np.where((Zone == 2) & (Applied_Value >= 171.0) & (Applied_Value <= 189.0), Yield_WetMass, np.nan)

Applied-based and Historical Productivity

The Yield figures from the Trials consistently surpass the average Yield across the entire Field. A key factor driving this disparity appears to be the historically high productivity zone where the Trials took place, as visualized in Fig.15 and Fig.16. For a more nuanced evaluation of the Trials, it's crucial to factor in the productivity zones when analyzing results.
Fig.15 Historical Field Potential Zones
Fig.16 Historical Field Potential Zones as YieldDataset
Presented in the subsequent screenshots (Fig.17, Fig.18, Fig.19) is the statistical distribution of the Yield, focusing on the Nitrogen (N34) actually Applied numbers overlapped with Historical Productivity Zones (created in GeoPard). Here are the summarized statistics, keeping in mind the ±5% accuracy acceptance for Applied numbers:
  • The main zone at 150 kg/ha had an applied area of 2.65 ha, yielding an average of 6.34 t/ha (Fig.17).
  • The first trial zone at 180 kg/ha covered an area of 1.08 ha and produced an average yield of 6.41 t/ha (Fig.18).
  • The second trial zone set at 120 kg/ha spanned an area of 1.78 ha, with an average yield of 6.33 t/ha (Fig.19).
Fig.17 Main Zone with N34 150kg/ha overlapped with Historical Productivity
Fig.18 Trial Zone with N34 180kg/ha ±5% overlapped with Historical Productivity
Fig.19 Trial Zone with N34 120kg/ha ±5% overlapped with Historical Productivity
For a deeper understanding of the methodology and the specifics of these results, the used Equations are below:
  1. 1.
    Main with 150 kg/ha incorporating 5% acceptance overlapped with Historical Productivity: Yield_Main = np.where((Application_Zone == 1) & (Productivity_SubZone == 51) & (Applied_Value >= 142.5) & (Applied_Value <= 157.5), Yield_WetMass, np.nan)
  2. 2.
    Trial with 120 kg/ha incorporating 5% acceptance overlapped with Historical Productivity: Yield_Trial = np.where((Application_Zone == 3) & (Productivity_SubZone == 51) & (Applied_Value >= 114.0) & (Applied_Value <= 126.0), Yield_WetMass, np.nan)
  3. 3.
    Trial with 180 kg/ha incorporating 5% acceptance overlapped with Historical Productivity: Yield_Trial = np.where((Application_Zone == 2) & (Productivity_SubZone == 51) & (Applied_Value >= 171.0) & (Applied_Value <= 189.0), Yield_WetMass, np.nan)
where
  • the part Productivity_SubZone == 51 reflects the high Productivity Zones with the applied experiments,
  • the parts (Applied_Value >= 142.5) & (Applied_Value <= 157.5) , (Applied_Value >= 114.0) & (Applied_Value <= 126.0), (Applied_Value >= 171.0) & (Applied_Value <= 189.0) incorporate ±5% accuracy from the rates 150, 120, 180 kg/ha.

Summary

The Yield results from the Trials closely align with the average Yield observed across the high Historical Productivity Zone of the Field. In other words, the experimental application of N34 product at rates of 120 kg/ha - 150 kg/ha - 180 kg/ha, resulted in average Yields of 6.33 t/ha - 6.34 t/ha - 6.41 t/ha respectively, does not have a significant impact on the harvested Yield within the high Productivity Zone.