This project utilizes co-located eddy-covariance flux sensors as “training data”to allow machine learning (neural networks and support vector machines) to estimate actual evapotranspiration using simple and low-cost on-farm weather stations. On-farm trials have been conducted on a wide range of commercial crops (alfalfa, snap beans, hazelnut orchards, vineyards, pasture) at sites ranging across the western states, including the Willamette Valley and Harney County Oregon, Rifle CO, Parlier, Dixon, and Biggs California.
Field trials and analysis indicate that neural networks can make robust estimates of actual evapotranspiration with as little as one week of training with the eddy-covariance or other control estimate. Results indicate that a minimum set of sensors including air temperature, humidity, and solar radiation are sufficient to faithfully reproduce daily irrigation demand estimates.
Grants
UI-ARES Research Support “Idaho Center for Center for Agriculture, Food and the
Environment – Irrigation, Water use, and Atmospheric science” PI Kelley, Co-PI Rick Allen
Project Award $96,220. Completed in 2020.
USDA-NIFA Award # 2017-67012-26125 “Integrating On-Farm Information to Optimize
Water Management” Sole PD. Project Award $116,931. Completed in 2018.
Publications and Presentations
Kelley, Jason. “Machine Learning Model of Soil Heat Flux and Delayed Closure of the Surface
Energy Budget.” Submitted to Agricultural and Forest Meteorology September 2022.
Kelley, Jason. 2020. Assessment and Correction of Solar Radiation Measurements with Simple
Neural Networks. Atmosphere 11, 1160. DOI: 10.3390/atmos11111160
Kelley, Jason, McCauley, D., Alexander, A., Gray, W., Siegfried, R., Oldroyd, H.J., 2020.
Using Machine Learning to Integrate On-Farm Sensors and Agro-Meteorology Networks into
Site-Specific Decision Support. Transactions of the ASABE 63. DOI: 10.13031/trans.13917
Kelley, Jason, and Pardyjak E.R., 2019 “Using Neural Networks to Estimate Site-Specific Crop
Evapotranspiration with Low-Cost Sensors.” Agronomy 9, no. 2: 108. DOI:10.3390/agronomy9020108
Kelley, Jason. “Research Advances in ET based water management”. Quarterly meeting of the
California Specialty Crop Commission, November 9 2022
Jome, Mathilde, Lohou F., Lothon M., Kelley J., Pardyjak E., “Using Artificial Neural Network
to Estimate Surface Convective Fluxes.”, European Geophysical Union, May 2022.
Kelley, Jason “Using Machine Learning to Integrate On-Farm Sensors and Ag-Weather
Networks into Site-Specific Decision Support” ASABE/IA 6th Decennial National Irrigation
Symposium, San Diego California, 06-08 December 2021.
Kelley, Jason. “Using neural networks for data assimilation and analysis”, Annual UI Data
Science Symposium (Invited), Moscow Idaho 16May2019.
Kelley, Jason. ” The Importance of Meta-Data in Interdisciplinary Collaborations” (Invited
Seminar), Department of Plant Sciences, UC Davis, 14 March 2019.
Kelley, Jason. Supporting Site-Specific Agriculture with ‘Medium Data'” (Invited Seminar),
WSU Crop & Soil Sciences Seminar Series, 03 March 2019.
Kelley, Jason. “Using Machine Learning to Evaluate Site-specific Crop Coefficients.”, Int’l
Meeting of ASABE, Detroit MI, 01 August 2018.
Kelley, Jason. “Measuring Site Specific Evapotranspiration using Neural Networks” Invited
Presentation for Special Session on Agriculture and ET Measurement at American Water
Resources Association Annual Meeting, 07 November 2017.
Kelley, Jason. “Neural Networks and Low Cost Sensors to Estimate Site-Specific
Evapotranspiration” Amer. Society of Agricultural and Biological Engineers Annual Meeting,
Spokane WA, USA, 19 June 2017.