Validating Weather and Sensor-based N Prediction Models for Michigan Corn Production


01 Apr 2015

Project Description

Due to seasonal and geographic variables impacting N cycling and corn growth, accurately predicting the economic optimum nitrogen rate (EONR) remains a challenge. One approach to provide N recommendations, referred to as MRTN (Maximum Return To Nitrogen), has been utilized by seven Midwestern states and provides N rate recommendations based on yield response to applied N while simultaneously accounting for fertilizer and grain prices. MRTN is considered a regional approach to N recommendations as the research trials from which they are derived are conducted within individual states. Research trials occur over multiple sites and years allowing growers to adjust N rates within a profitable range based on crop rotation, soil productivity potential, N fertilizer to corn price ratios, and application timing. However spatial variability in N response is known to exist from site to site as well as within a single location from year-to-year. Variability creates additional risk when determining the N rate that will be economically optimal for a field-specific site.

The MRTN is a pre-season general N recommendation model (as compared to a fieldspecific or site-based model) that does not account for individual site variability nor variable inseason weather trends which can affect corn N response. Incorporating weather data may result in a more precise N rate decision. One specific site-based model-based approach Adapt-N ( was developed to provide field-specific locally-adjusted sidedress N recommendations for corn based on soil and management factors in addition to accounting for early-season weather patterns impacting N dynamics. The Adapt-N program has the potential to adjust side-dress N rates and increase the accuracy of site-specific N recommendations. There is potential risk for post-application fertilizer N losses which may impact the site economic optimum yield if N becomes a limiting factor. Previous work by Laboski, Camberato, and Sawyer (2014) found Adapt-N to under recommend N to a greater extent than MRTN in Indiana and Iowa, and the model failed to adequately account for excessive rainfall and mineralization.

Plant canopy sensors (i.e., Minolta SPAD 502 chlorophyll meter (CM)) have been used to detect N stress and guide in-season fertility decisions. Scharf et al. (2006) related CM readings with EONR and corn yield response to N. Other studies have utilized remote sensors such as the Greenseeker to develop in-season yield potential estimates and the likely yield response to added N fertilizer when the N recommendation was based on additional obtainable yield (Raun et al., 2005). Few data exist in Michigan for utilizing canopy sensors to determine in-season N requirements but may offer another opportunity for site-specific estimation of in-season N needs as influenced by early-season weather.

1) Evaluate site- and year-specific model-based approaches to identify strengths and weaknesses for predicting economic N-fertilization rates for corn. Compare pre-plant to sidedress application for achieving equivalent yield potential. Utilizing model-based approaches may begin to validate effectiveness and provide opportunities for improvement. (Incorporating weather data into site recommendations may impact N rate accuracy, but could limit economic potential where postapplication weather patterns are extreme.)

Hypotheses: 1) Application of N at V4 can produce yields as high as those achieved with
a non-limiting rate applied at planting, and 2) Adapt-N rate using site-specific soil and weather
information up to V8 is closer to EONR than MRTN prediction at planting.

2) Utilize Minnesota-based N rate calculator to develop economic optimum in-season N recommendation for Greenseeker sensor-based treatment at multiple growth stages. Optimum growth stage for sensing is estimated at V8 based on previous literature. Externally-derived data from similar latitude may produce similar recommendations to site yield-based EONR.

Hypothesis: Greenseeker rate based on Minnesota algorithm using sensor data at V8 is
closer to EONR than MRTN prediction at planting.

Experimental Design: Two or three site-year study. Eight (8) treatments arranged in a
randomized complete block design with four replications; plot size = 15 x 40 feet; 30” row

Rate (lbs N/A)
Placement + Timing
Non-limiting NPPI
N response40 unites 2x2, V4 Sidedress
33% MRTN
N response40 unites 2x2, V4 Sidedress
67% MRTN
N response40 unites 2x2, V4 Sidedress
133% MRTN
N response40 unites 2x2, V4 Sidedress
167% MRTN
N response40 unites 2x2, V4 Sidedress
Model Dependent
Adapt-N40 unites 2x2, V8 Sidedress
Sensor Dependent
Greenseeker40 unites 2x2, V8 Sidedress
Sensor Dependent
Greenseeker – season long data collection40 unites 2x2

    1. Greenseeker readings at growth stages V4, V6, V8, V10, V12, V14, R1 (treatments 9 and 1).
    2. Greenseeker readings at growth stage V8 for sidedress N rate calculation (treatments 8 and 1).
    3. Weather recorded from Michigan Automated Weather Network (MAWN)
    4. Harvest: Yield (bu/a), test weight (lbs/bu), % moisture
    5. Preplant 0-8” base soil sample (field composite) and 0-12” (per block).
    6. Pre-sidedress soil nitrate (PSNT) 0-12” at V6 (treatment 9).

    1. Regression analysis to fit corn yield response and calculate EONR
    2. Compare distributions of sensor-based N rates calculated at multiple growth stages relative to Adapt-N based N rates.
    3. Compare distribution of MRTN and Adapt-N based N rates vs. EONR.