Progress 07/01/99 to 06/30/04
Outputs
Late frosts during the end of Winter and early Spring can have a detrimental impact on fruits and vegetables. Flowers on blueberry bushes and peach trees can be permanently damaged and young vegetable plants can be frozen during these late frosts, causing, in some cases, a permanent loss in production. Growers and producers have several options for freeze protection, including irrigation and wind machines. However, critical for implementing frost and freeze protection is accurate and timely weather forecasts. Recent changes in the federal law prohibit the National Weather Service from providing agricultural-specific weather forecasts. The University of Georgia, therefore, has explored the use of artificial intelligence techniques to predict temperature based on local temperature, relative humidity and other weather information. The College of Agricultural and Environmental Sciences of the University of Georgia currently operates a network of automated weather stations
that are located across the state of Georgia. These weather stations have been sited in regions where agriculture is the dominant economic sector. Each automated station monitors local weather conditions, including air temperature, relative humidity, dewpoint, wind speed and direction, solar radiation, precipitation, and other variables. The weather data are transmitted to a computer in Griffin and then disseminated via the world wide web at www.Georgiaweather.net. Current weather conditions are updated at least hourly. Based on the historical weather data that have been collected for each site, neural network models have been developed that can predict temperature for up to 12 hours. An artificial neural network is a complex computer model that emulates the operation of neurons in the human brain. An ANN can determine relationships between complex data structures and based on past experience, can predict what will happen in the future. In this implementation the ANN model is provided
with the current weather conditions, as well as with the conditions during the previous eight hours to predict the temperature for up to 12 hours. The average prediction error was 1oF for the one-hour model and 4.5 oF for the 12-hour model. A graphical user interface has been developed to display both the current weather conditions as well as the predicted temperatures for each site where an automated weather station is operational. Further research is being conducted to refine the use of neural network and other artificial intelligence techniques for temperature prediction.
Impacts
The ANN model performed well in an evaluation on an independent set with an r2=0.924 and mean absolute error of 0.015 mg/g/hr. This model could be used to manage a composting facility to maximize the microbial activity. Artificial Neural Networks (ANNs) were developed to classify the following 4-hour, 8-hour, and 12 -hour periods as having a frost event (Temp. < 0 deg. C), near frost event ( 0 deg. C< Temp.< 3 deg. C), or no frost (Temp. > 3 deg. C). The networks were developed for multiple locations in Georgia. Nine years of daily weather data were used in the study. As part of this effort, the preferred weather data inputs, ANN architecture and the duration of prior weather data needed to produce the most accurate predictions were determined. It was found that the following weather data variables were important in predicting frost: air temperature, relative humidity, wind speed, rainfall, and solar radiation. Site-specific models were compared to models developed
using weather data from multiple locations. It was found that models developed using weather data from nine locations performed as well as models developed for a particular site. This will allow the use of this model to predict frost for locations without historical weather data
Publications
- Liang, C., K.C. Das and R.W. McClendon. 2003. The Influence of Temperature and Moisture Content Regimes on the Aerobic Microbial Activity of a Biosolids Composting Blend. Bioresource Technology 86(2):131-137
- Li, B., R.W. McClendon, G. Hoogenboom. 2004. Spatial Interpolation of Weather Variables for Single Locations using Artificial Neural Networks. Transactions of the ASAE 47(2):629-637.
- Ashish, D., G. Hoogenboom, and R.W. McClendon. 2004. Land-use Classification of Gray-scale Aerial Images using Probabilistic Neural Networks, Transactions of the ASAE 47(5) (In Press)
|
Progress 01/01/03 to 12/31/03
Outputs
Microbial activity during biosolids composting was predicted using artificial neural networks (ANNs). Currently no model is available to predict this process. The inputs to the ANN were temperature, moisture content, and the duration of the process (time). The microbial activity was indicated by O2 uptake rate (mg kg/hr). The biosolids were obtained from a water treatment facility in Athens, GA.
Impacts
The ANN model performed well in an evaluation on an independent set with an r2=0.924 and mean absolute error of 0.015 mg/g/hr. This model could be used to manage a composting facility to maximize the microbial activity.
Publications
- Kastner, J.R., K.C. Das, C. Hu, and R.W. McClendon. 2003. Effect of pH and Temperature on the Kinetics of Odor Oxidation Using Chlorine Dioxide. Journal of the Air & Waste Management Association 53:1218-1224.
- Liang, C., K.C. Das and R.W. McClendon. 2003. Prediction of Microbial Activity During Biosolids Composting Using Artificial Neural Networks. Transactions of the ASAE 46(6):1713-1719.
|
Progress 01/01/02 to 12/31/02
Outputs
Effective decision making for agricultural crop production and natural resource management is highly dependent on accurate and timely localized weather data. Currently the Georgia Automated Environmental Monitoring Network (AEMN) consists of more than 50 automated weather stations located across the state of Georgia. However, given the size of Georgia, the nearest weather station could be 50 miles away. Artificial Neural Networks (ANNs) were developed to estimate maximum and minimum temperature and solar radiation for specific locations in Georgia.
Impacts
The model was evaluated for Tifton Georgia giving a mean absolute error (MAE) of 0.61? C for maximum temperature, 0.74? C for minimum temperature, and 1.24 MJ/m2 for solar radiation. These results were better than obtained using traditional methods of interpolation.
Publications
- Shahin, M.A., E.W. Tollner, R.W. McClendon, and H.R. Arabnia. 2002. Apple Classification Based on Surface Bruises using Image Processing and Neural Networks. Transactions of the ASAE 45(5):1619-1627.
- Liang, C., K.C. Das, and R.W. McClendon. 2002. The Influence of Temperature and Moisture Content Regimes on the Aerobic Microbial Activity of a Biosolids Composting Blend. Bioresource Technology 86(2):131-137.
- Bartley, P.G., Jr., S.O. Nelson, and R.W. McClendon. 2002. Dimensional Analysis of a Permitivity Probe. IEEE Transactions on Instrumentation and Measurement 51(6):1312-1315.
|
Progress 01/01/01 to 12/31/01
Outputs
A land use classification system was developed to use artificial neural networks to analyze aerial grayscale images of Georgia. Once training was complete, the best model had an overall accuracy of 92%. A neural network model was developed to predict microbial activity during biosolids composting. The model used temperature, moisture content and time as inputs and predicted O2 uptake with an r2 = 0.92.
Impacts
A land use classification system will aid in determining cropping patterns in the state and these can be used in making agricultural and water use policy. An accurate prediction of biosolids composting rate and be incorporated into controls to make the process more efficient.
Publications
- Shahin, M.A., E.W. Tollner, R.W. McClendon. 2001. Artificial Intelligence Classifiers for Sorting Apples based on Watercore. Journal of Agricultural Engineering Research 79(3):265-274.
- Thomas, D.L., K.A. Harrison, J.E. Hook, G. Hoogenboom, R.W. McClendon, I. Wheeler, W.I. Segars, J. Mallard, G. Murphy, M. Lindsay, D.D. Coker, T. Whitley, J. Houser, and C. Myers-Rocher. 2001. Status of Ag. Water Pumping: a program to determine agricultural water use in Georgia. p. 101-104. In: [K.J. Hatcher, editor] Proceedings of the 2001 Georgia Water Resources Conference. Institute of Ecology, The University of Georgia, Athens, Georgia (ISBN 0-935835-07-5).
- Bartley, P.G. Jr., S.O. Nelson, and R.W. McClendon. 2001. Dimensional Analysis of an Open-ended Coaxial-Line Probe. Conference Proceedings, IMTC/2001, IEEE Instrumentation and Measurement Technology Conference, May 21-23, 2001, Budapest, Hungary.
|
Progress 01/01/00 to 12/31/00
Outputs
Daily pan evaporation has been shown to be an important variable in making crop management decisions and in modeling crop response to weather conditions. However, daily pan evaporation is difficult to measure accurately and consistently over longer time periods. Automated data collection systems were compared to manual systems to determine their accuracy. The automated systems were found to closely follow the results of the manual systems.
Impacts
Pan evaporation data is useful in decision making regarding the management of irrigation systems. Having complete and accurate data could allow producers to protect crop fields with minimal water use. With the current drought and water use limitations, efficient water use would increase producer profits.
Publications
- Bruton, J.M., G. Hoogenboom, and R.W. McClendon. 2000. A Comparison of Automatically and Manually Collected Pan Evaporation Data. Transactions of the ASAE 43(5):1097-1101.
- Bruton, J.M., R.W. McClendon, and G. Hoogenboom. 2000. Estimating Daily Pan Evaporation with Artificial Neural Networks. Transactions of ASAE 43(2): 491-496.
- Lacey, B., T.K. Hamrita, and R.W. McClendon. 2000. Prediction of Poultry Deep Body Temperature Responses to Changes in Ambient Temperature using Neural Networks. Applied Engineering in Agriculture 16(3):303-308.
|
Progress 01/01/99 to 12/31/99
Outputs
Progress Report Narrative A hybrid artificial intelligence approach was used to estimate the level of aflatoxin contamination in preharvest peanuts. A genetic algorithm routine was used to search for the weights and parameters in an artificial neural network (ANN). Accurate models of aflatoxin contamination could allow producers to manage their crops to minimize contamination. This would allow the producers to protect profits and reduce risks to the consumer. An ANN model was also developed to estimate daily pan evaporation using other measured weather variables. Pan evaporation is a difficult weather variable to measure yet correlates well with crop water requirements. The best model was able to predict pan evaporation with a mean squared error of 1.11 mm for a dataset not used in model development.
Impacts
Aflatoxin contamination in peanut crops is a problem of significant health and financial importance. Predicting aflatoxin levels prior to crop harvest would allow a produce to make field level management decisions to minimize the losses due to aflatoxin contamination. The pan evaporation model can be used to estimate daily values for sites where the measuring system failed or where it is not being observed. With recent droughts and water restrictions being considered, this advanced technology could improve the economic conditions facing the grower by improving their decision making skills.
Publications
- Henderson, C.E., W.D. Potter, R.W. McClendon, G. Hoogenboom. 1999. Predicting Aflatoxin Contamination in Peanuts: A Genetic Algorithm/Neural Network Approach. Applied Intelligence 12(3): 183-192.
- Pabico, J.P., G. Hoogenboom, and R.W. McClendon. 1999. Determination of Cultivar Coefficients of Crop Models using a Genetic Algorithm: A Conceptual Framework. Transactions of the ASAE 42(1): 223-232.
|
|