Determining degree-days thresholds from field observations

This paper compares several methods for determining
degree-day (°D) threshold temperatures from
field observations. Three of the methods use the mean
developmental period temperature and simple equations
to estimate: (1) the smallest standard deviation in °D, (2)
the least standard deviation in days, and (3) a linear regression
intercept. Two additional methods use iterations
of cumulative °D and threshold temperatures to determine
the smallest root mean square error (RMSE). One
of the iteration methods uses a linear model and the other
uses a single triangle °D calculation method. The
method giving the best results was verified by comparing
observed and predicted phenological periods using 7
years of kiwifruit data and 10 years of cherry tree data.
In general, the iteration method using the single triangle
method to calculate °D provided threshold temperatures
with the smallest RMSE values. However, the iteration
method using a linear °D model also worked well. Simply
using a threshold of zero gave predictions that were
nearly as good as those obtained using the other two
methods. The smallest standard deviation in °D performed
the worst. The least standard deviation in days
and the regression methods did well sometimes; however,
the threshold temperatures were sometimes negative,
which does not support the idea that development rates
are related to heat units.