Event Title
Start Date
28-8-2017 9:30 AM
End Date
28-8-2017 9:45 AM
Description
Abstract:
The cranberry industry in MA has long been served by predictive formulas for cranberry frost protection, diligently created by Dr. Franklin in the 1940s. Recently, climate patterns and grower winter management practices have changed, and in several of the last few years, a need for frost prediction as early as the last week of March has emerged. Likewise, in the fall, late harvesting has become a more regular practice, so that prediction for the first two weeks of November is also needed. Because the Franklin formulas were developed for specific seasonal periods, their use outside of these intervals yields unreliable results. To mitigate unreliable predictions outside the time periods used by Franklin, I. DeMoranville developed supplementary formulas that have not been adopted by the Cape Cod Cranberry Growers Association (CCCGA) for cranberry frost predictions. This project is developing new frost prediction models based upon the current body of agroclimatic research and information. Frost prediction models are being developed using logistic regression and decision tree techniques. Hit Rate (HR), Probability of Detection (POD), and False Alarm Rate (FAR) from both models were calculated and compared. Threshold values for the logistic regression models were selected to maximize HR and POD and minimize FAR, and the split for the decision tree models was stopped when change in entropy was relatively small. Although this approach is somewhat risky, it has the potential to unearth modern knowledge in this area and to develop new empirical formulas that bear no resemblance to the old. Models will be validated against historical data; however, emphasis will be placed on the period during which cranberry culture dramatically changed and where weather patterns seem to have shifted.
Recommended Citation
Jeranyama, Peter, "Spring Frost Prediction Models in Cranberry" (2017). North American Cranberry Researcher and Extension Workers Conference. 6.
https://scholarworks.umass.edu/nacrew/2017/papers/6
Included in
Spring Frost Prediction Models in Cranberry
Abstract:
The cranberry industry in MA has long been served by predictive formulas for cranberry frost protection, diligently created by Dr. Franklin in the 1940s. Recently, climate patterns and grower winter management practices have changed, and in several of the last few years, a need for frost prediction as early as the last week of March has emerged. Likewise, in the fall, late harvesting has become a more regular practice, so that prediction for the first two weeks of November is also needed. Because the Franklin formulas were developed for specific seasonal periods, their use outside of these intervals yields unreliable results. To mitigate unreliable predictions outside the time periods used by Franklin, I. DeMoranville developed supplementary formulas that have not been adopted by the Cape Cod Cranberry Growers Association (CCCGA) for cranberry frost predictions. This project is developing new frost prediction models based upon the current body of agroclimatic research and information. Frost prediction models are being developed using logistic regression and decision tree techniques. Hit Rate (HR), Probability of Detection (POD), and False Alarm Rate (FAR) from both models were calculated and compared. Threshold values for the logistic regression models were selected to maximize HR and POD and minimize FAR, and the split for the decision tree models was stopped when change in entropy was relatively small. Although this approach is somewhat risky, it has the potential to unearth modern knowledge in this area and to develop new empirical formulas that bear no resemblance to the old. Models will be validated against historical data; however, emphasis will be placed on the period during which cranberry culture dramatically changed and where weather patterns seem to have shifted.