Frost depth prediction

The purpose of this research project is to develop a model that is able to accurately predict frost depth on a particular date, using available information. Frost depth prediction is useful in many applications in several domains. For example in agriculture, knowing frost depth early is crucial for farmers to determine when and how deep they should plant. In this study, data is collected primarily from NDAWN(North Dakota Agricultural Weather Network) Fargo station for historical soil depth temperature and weather information. Lasso regression is used to model the frost depth. Since soil temperature is clearly seasonal, meaning there should be an obvious correlation between temperature and different days, our model can handle residual correlations that are generated not only from time domain, but space domain, since temperatures of different levels should also be correlated. Gupta's research [1] "A note on the asymptotic distribution of Lasso estimator for correlated data" is used in this project. Furthermore, root mean square error (RMSE) is used to evaluate goodness-of-fit of the model.