Improving agroecosystem input management with on-farm experimentation

Background/Question/Methods

In semi-arid regions, soil water and nitrogen (N) are generally limiting factors for wheat (Triticum aestivum L.) production; hence, implementation of efficient N fertilization strategies is needed to optimize profitability and to maintain agro-ecosystem sustainability.  Site-specific technologies (yield monitors, protein sensors, weed density monitors and spectral reflectance monitors) may improve within field spatial prediction of crop response to N application and thus increase N use efficiency and decrease N leaching. These technologies also allow for local parameterization of predictive models through on-farm experimentation. The spatial dynamic of yield and protein content in a wheat crop was studied in three fields of approximate 60 hectares over seven years in Montana under a crop/fallow annual rotation. Several driving variables (previous yield, previous protein content, field topography, weed density, nitrogen rates, normalized difference vegetation index (NDVI), soil electric conductivity, soil N, etc.) were evaluated to develop practical models (easily reproducible by farmers) with sufficient precision to predict wheat yield and wheat quality in different areas of each field. Based on those models variable nitrogen rate maps were created for each field and compared for efficiency with the current nitrogen recommendations for the region.

Results/Conclusions

Although N savings with site specific management was field specific, a significant quantity of N could be saved with the use of precision agriculture practices in all fields studied in the semi-arid regions. The use of site-specific technologies contributed to improved distribution of N fertilizer due to the variability wheat productivity and quality (protein) in response to fertilizer in different areas of each field. The specific quantity of N savings per field and level of certainty in outcomes will depend mainly on the magnitude of field variability, and on the quality of the input data.  Best models for predicting current wheat yields and protein included yield and protein from previous years, weed density, elevation over the sea level, N rates and NDVI. These driving variables were relatively stable over time in a given field, but tend to vary more for different fields accentuating the requirement for site-specific parameterization of management focused predictive models. On farm experimentation is a logical component of adaptive management in agroecosystems and should lead to higher sustainability.