Progress 03/01/05 to 02/28/10
Outputs
OUTPUTS: Artificial Neural Network (ANN) models were developed to predict hourly air temperatures and dew point temperatures for periods of one to twelve hours. The ANNs were incorporated into a web-based Decision Support System(DSS). The inputs are observed weather data over the past 24 hours obtained by the Georgia Automated Environmental Monitoring Network(AEMN). New predictions are generated every 15 minutes for each of the over 75 weather stations in the AEMN. Plots of the air temperature and dew point temperature are then generated for the subsequent 12 hours along with the observed temperatures for the previous 12 hours. Support Vector Machine Regression (SVMR) models were developed to predict hourly air temperatures for one to twelve hours. These models used the same inputs as the previously developed ANN models. The research objective was to determine if the SVMR models would be more accurate than the ANN models. A Fuzzy Logic based model was developed to incorporate the air temperature and dew point temperature predictions along with current wind conditions to assess the risk of frost. A team of agrometeorologists was assembled and created five categories of frost/freeze risk: (1) No Warning, (2) Possible Frost, (3) Mild Frost, (4) Severe Frost, and (5) Hard Freeze. Scenarios of possible air temperatures, dew point temperatures, and wind were established and the frost/freeze risk was assessed. The Fuzzy Logic model was developed with a portion of the scenarios and was evaluated with the remaining scenarios. The Fuzzy Logic model was developed to be included with the web-based DSS. A color shaded computer graphics map of the state of Georgia was developed to show the frost/freeze risk at all locations. PARTICIPANTS: Gerrit Hoogenboom, Ph.D. Professor of Biological and Agricultural Engineering. Brahm Verma, Ph.D. Professor of Biological and Agricultural Engineering. Ronald W. McClendon, Ph.D. Professor of Biological and Agricultural Engineering. Risk Management Agency of the USDA Presentations were made to various meetings in Georgia of horticultural and agricultural producers, industry professionals, USDA and universtiy professionals demonstrating the use of the website to make informed decisions to protect crops from frost/freeze. TARGET AUDIENCES: Target audiences: horticultural and agricultural producers, industry professionals, USDA and universtiy professionals. Allow producers to protect crops from frost/freeze by providing localized predictions of air temperature, dew point temperature along with frost/freeze risk assesssment. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.
Impacts
The ANNs for predicting air temperature and dew point temperature were incorporated into a website to allow access from any location. The user can go online to the site www.georgiaweather.net , select one of the available weather stations, and then prompt the system for temperature predictions for that location. These AEMN predictions are available in agricultural producing areas as opposed to the National Weather Service predictions, which tend to be provided for more urban locations. The ANN air temperature models were evaluated with a dataset not used in model development and the Mean Absolute Error (MAE) ranged from approximately 0.52 deg C at one hour to approximately 1.87 deg C at 12 hour predictions. The SVMR models were evaluated on the same dataset and the MAE values were comparable to the ANN model results. Frost/freeze conditions can be very damaging to agricultural and horticultural crops. Often the producer has the capability to mitigate the damaging effects of frost/freeze, however, they need time to take action. This system provides up to a twelve-hour warning and allows the user to take action to protect the crop. The Fuzzy Logic based DSS was evaluated on weather scenarios not used in model development and the system agreed with the assessments of the agrometeorologists for all of the scenarios.
Publications
- Ashish, D., G. Hoogenboom, and R.W. McClendon. 2009. Land-use classification of mutispectral aerial images using artificial neural networks. International Journal of Remote Sensing 30(8):1989-2004.
- Smith, B.A., G. Hoogenboom, and R. W. McClendon. 2009. Artificial neural networks for automated year-round temperature prediction. Computers and Electronics in Agriculture 68(1):52-61.
- Chevalier, R.F.G. Hoogenboom, R.W. McClendon, and J. A. Paz. 2010. Support Vector Regression With Reduced Training Sets For Temperature Prediction: A Comparison With Artificial Neural Networks. Neural Computing and Applications. (In Press)
|
Progress 01/01/08 to 12/31/08
Outputs
OUTPUTS: Artificial Neural Network prediction models for air temperature and dewpoint temperature have been implemented on the web site of the Georgia Automated Environmental Monitoring Network, www.Georgiaweather.net. Each model is automatically run every five minutes and graphical temperature predictions are presented for all 75+ weather stations. The location-specific predictions are for the subsequent 12 hour period. PARTICIPANTS: Gerrit Hoogenboom and Joel Paz, faculty members in Biological and Agricultural Engineering , UGA. Dr. Paz is a faculty member with an appointment in Cooperative Extension Service and frequently conducts training sessions for other exension service personnel and agricultural producers. Dr. Hoogenboom regularly conducts workshops on both crop growth simulation models and agrometeorology. The research to develop decision support models has been partially supported by the USDA-Federal Crop Insurance Corporation through the Risk Management Agency TARGET AUDIENCES: Agricultural producers, consultants, researchers, and cooperative extension personnel. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.
Impacts
The web-based decision support systems allow the users to make informed decisions about protecting crops from freezing temperatures and protecting livestock and personnel from heat stress conditions.
Publications
- Shank, D.B., G. Hoogenboom, and R.W. McClendon. 2008. Dew point temperature prediction using artificial neural networks. Journal of Applied Meteorology and Climatology 47(6):1757-1769.
- Shank, D.B., R.W. McClendon, J.O. Paz, and G. Hoogenboom. 2008. Ensemble artificial neural networks for prediction of dew point temperature. Applied Artificial Intelligence 22(6):523-542.
|
Progress 01/01/07 to 12/31/07
Outputs
Artificial Neural Networks(ANNs) were developed to predict the likelihood of rainfall during a 24-hour period. The prediction models were created with the conditions that the prediction would be made at 6:00 PM and the prediction period would be midnight to midnight. The ANNs used inputs of weather variables obtained from the University of Georgia Automated Environmental Monitoring Network(AEMN). ANNs were developed for each season and for year-round predictions. The ANNs will be used with data from any of the 70+ weather stations which make up the AEMN system. Research was also started on developing ANN models to predict rainfall amounts for the same periods
Impacts
ANN models predicting rainfall probablitiy and rainfall amount will be valuable for decision-makers related to most areas of agricultural production. In particular decisions related to the management of irrigation systems would be greatly enhanced with the availability of rainfall predictions localized to rural areas of Georgia.
Publications
- No publications reported this period
|
Progress 01/01/06 to 12/31/06
Outputs
Dew point temperature is the temperature at which water vapor in the air will condense into liquid. This temperature can be useful in estimating frost, fog, rain, snow, dew, evapotranspiration, and other meteorological variables. Artificial neural networks (ANNs) were developed to predict dew point temperature from one to 12 hours ahead using prior weather data as inputs. It was found that in addition to dew point temperature, important weather related ANN inputs included relative humidity, solar radiation, air temperature, wind speed, and vapor pressure. The evaluation of the final models with weather data from 20 separate locations and a different year showed that the one-hour prediction had a mean absolute error (MAE) of 0.550 degrees C, the four-hour prediction model had a MAE of 1.234 degrees C, the eight-hour prediction had a MAE of 1.799 degrees C, and the 12-hour had a MAE of 2.280 degrees C.
Impacts
Dew point temperature and air temperature are important factors related to frost formation and in intensifying the effects of a heat wave. Predictions of dew point temperature and air temperature can be used in decision support to allow steps to be taken to prevent losses to crops and livestock and the loss of human lives.
Publications
- Jain, A., R.W. McClendon, G. Hoogenboom. 2006. Freeze Prediction for Specific Locations Using Artificial Neural Networks, Transactions of the ASABE 49(6):1955-1962
- Smith, B.A., R.W. McClendon, and G. Hoogenboom. 2006. Improving Air Temperature Prediction with Artificial Neural Networks, International Journal of Computational Intelligence 3(3):179-186
|
Progress 01/01/05 to 12/31/05
Outputs
Research was focused on developing Artificial Neural Network(ANN) models with reduced average prediction errors by increasing the number of distinct observations used in training, adding additional input terms that describe the date of an observation, increasing the amount of prior weather data to include in each observation, and reexamining the number of hidden nodes used in the network. Models were created to predict air temperature at hourly intervals from one to 12 hours ahead. Each ANN model, consisting of a network architecture and set of associated parameters, was evaluated by instantiating and training 30 networks and calculating the mean absolute error (MAE) of the resulting networks for some set of input patterns. The inclusion of seasonal input terms, up to 24 hours of prior weather information, and a larger number of processing nodes were some of the improvements that reduced average prediction error compared to previous research across all horizons. For
example, the four-hour MAE of 1.40 deg.C was 0.20 deg.C, or 12.5%, less than the previous model. Prediction MAEs eight and 12 hours ahead improved by 0.17 deg.C and 0.16 deg.C, respectively, improvements of 7.4% and 5.9% over the existing model at these horizons.
Impacts
Frost damage is a significant concern for horticultural producers when bud formation and flowering occur during late-winter and early-spring. Unseasonably cold temperatures during early 1996 and 2002 damaged floral buds and were responsible for reduced fruit harvests. Growers can take steps to lessen the effects of frost ,but these methods require advance warning of freezing conditions. The proposed system will provide these warnings.
Publications
- McClendon, R.W. and G. Hoogenboom. 2005. New non-insurance risk management tools: Decision support for freeze protection using artificial neural networks. Speech Booklet 2, Thursday, February 24, 2005. Agricultural Outlook Forum 2005, Arlington, Virginia.
- McClendon, R.W., G. Hoogenboom, A. Jain, R. Ramyaa, B. Smith. 2005. Temperature for Frost Prediction. Proceedings of the 2005 Southeast Regional Vegetable Conference. Savannah, GA. January 6-9, 2005. P. 97.
- Smith, B.A., R. W. McClendon, and G. Hoogenboom. 2005. An Enhanced Artificial Neural Network for Air Temperature Prediction. International Enformatica Society, Prague, Czech Republic, August 26-28, 2005. Enformatika 7: 7-12.
|
|