Using Modulated On-farm Response Surface Experiments (MORSE) to develop evidence based, agronomic recommendations for precision
Term: 5 years, ending June 2025
Status: Complete
Researcher(s): Steven Shirtliffe, University of Saskatchewan
SaskCanola Investment: $66,700
Total Project Cost: $268,883
Funding Partners: Sask Wheat, WGRF
Objectives
Develop methodology that will allow crop input experiments to be performed on-farm using Modulated On-farm Response Surface Experiments.
To refine image-based technology as a tool to assess crop response variables, including yield.
To develop statistical tools to analyze field-based MORSE experiment.
Project Description
Conventional agronomic field trials, including small-plot and strip trials, often fail to reflect the spatial complexity and operational scale of commercial agriculture. Small-plot designs are limited by narrow treatment ranges, low replication, and artificial conditions, while conventional strip trials lack sufficient replication and statistical resolution. This project aimed to evaluate whether the Modulated On-farm Response Surface Experiments (MORSE) framework could overcome these limitations by implementing precise, replicated input-response assessments while accounting for field-scale spatial variability, even within a small-plot research context.
Field trials were conducted from 2022 to 2024 on wheat and canola at multiple research sites near Saskatoon, Saskatchewan. Nitrogen response trials included 10–11 rate treatments applied in alternating MORSE and randomized complete block designs (RCBD). Fungicide trials evaluated multiple products at 0.33× to 2× label rates. UAV multispectral imagery was collected biweekly to derive normalized difference vegetation index (NDVI) and normalized difference red edge index (NDRE). All yield data were analyzed using spatial exponential covariance models, and model fit (corrected Akaike Information Criterion- AICc) and precision were compared between designs. Pearson correlation coefficients were used to quantify the relationship between vegetation indices and final grain yield.
MORSE yielded comparable or improved model performance at several site-years, particularly in canola nitrogen trials. Treatment effects were statistically significant across both designs, with no systematic loss of power in MORSE. NDRE consistently outperformed NDVI as a yield predictor, with the strongest correlations observed at 70–80 days after planting—corresponding with podding in canola and flowering in wheat.
In conclusion, the MORSE design is a viable alternative to RCBD for conducting high-resolution, spatially aware agronomic trials under commercial conditions. The integration of UAV-derived NDRE further enhances the capacity to monitor crop response and support early agronomic decision-making. This approach offers a scalable framework for advancing precision agriculture through evidence-based, field-scale experimentation.
Grower Benefits:
The MORSE design is a viable alternative to traditional RCBD. While both designs produced statistically valid results, MORSE’s structured layout and spatial modeling allowed more consistent detection of treatment effects across variable field conditions—addressing key limitations of traditional small-plot and strip trial approaches.
Producers can use NDRE imagery to assess crop response before harvest. UAV imagery collected 70–80 days after planting showed strong correlation with final yield, offering a non-destructive tool to evaluate input responses earlier in the season.
The MORSE framework shows potential for broader application in precision agriculture. Although not yet tested at full commercial scale, MORSE demonstrated the ability to evaluate multiple treatment levels efficiently within operational constraints. Its compatibility with UAV data and spatial modeling positions it as a practical tool for developing detailed, site-specific input-response relationships in future on-farm research.