Do you think you are a good forecaster? A bad one? There is only one way to find out: a competition! We are going to try to forecast forage production in the annual grasslands of California based on weather covariates. The idea is pretty simple:
To make it easy for me to calculate the accuracy of your forecast, please follow these formatting guidelines. Create a data frame with two to four columns with the following names: Year, Forecast, LowerCI, UpperCI. The Year column should contain the integers 2009 to 2018, in order. The “Forecast” column contains your point forecasts for each year. The point forecasts should be in the original units! If you fit on a transformed scale, please back transform.
The last two (CI) columns are optional. If you do report confidence intervals, please calculate the 95% intervals. Again, these should be on the same scale as the observations.
Write your data frame to a .csv file using the following line of code, substituting in the name of your data frame and the filename you want to use (your group name?):
write.csv(your_data_frame, your_file.csv, header=T)
Email the .csv file to Peter as an attachment.
As we’ve discussed, mechanistic knowledge can improve forecasts. The forage data come from the San Joaquin Experimental Range, in the Sierra Nevada foothills of California. Here is the citation for the data, which USDA NRCS and University of California Extension have generously made public for our use:
Dennis Dudley, USDA NRCS Rangeland Specialist, Madera County; Neil McDougald, UCCE Livestock, Range, and Natural Resources Advisor Emeritus, Madera County
This is an annual grassland, so measuring aboveground annual production (forage production) is straightforward: biomass is clipped to ground level, dried, and weighed. This is usually done in June at the end of the spring growing season. For more information about the factors that determine productivity in these grasslands, see this report.
I downloaded the weather data from the PRISM Explorer. You can find some metadata here. “ppt” refers to precipitation, “t” to temperature, and “vpd” to vapor pressure deficit.