Penalty CRPS Model And Its Application In Ground Temperature Prediction

Appropriate temperature is one of the important environmental conditions for human survival,at the same time,temperature has a huge impact on agriculture,medicine,public health and human production.In particular,the low tempera-ture prediction accuracy cannot provide good guidance for people's production and work plans for some regions with large temperature differences.In order to effectively improve the accuracy of temperature prediction and guide people's pro-duction and life,this paper explores a statistics post-processing ensemble model of continuous ranked probability score?CRPS?based on₁regular term.In par-ticular,based on the truncated normal distribution,the integrated prediction is converted into a probabilistic prediction and the integrated prediction result is corrected by constructing CRPS and adding a regular term to improve the pre-diction accuracy.Taking the Lanzhou city of Gansu province and the New Mexico of US as examples,we collected hourly ground temperature data of the two re-gions.Firstly,we used lagrange interpolation and singular spectral analysis?SSA?to complete and preprocess the original time series data.Then,a regular term₁is added to the original correct method of CRPS to reduce calculation and avoid overfitting.Finally,numerical results indicate that the penalty CRPS model has higher accuracy and basically fitted out the fluctuations of future temperature.