Distributional Regression Forests for Probabilistic Precipitation Forecasting in Complex Terrain

TL;DR

This paper introduces distributional regression forests that combine tree-based models with classical distributional approaches, enabling automatic variable and interaction selection for probabilistic weather forecasting in complex terrains.

Contribution

It develops a novel distributional regression forest framework that integrates trees with GAMLSS distributions, improving flexibility and performance in probabilistic precipitation forecasting.

Findings

Distributional regression forests outperform GAMLSS in many cases.

The method automatically selects relevant variables and interactions.

It provides accurate probabilistic precipitation forecasts in mountainous regions.

Abstract

To obtain a probabilistic model for a dependent variable based on some set of explanatory variables, a distributional approach is often adopted where the parameters of the distribution are linked to regressors. In many classical models this only captures the location of the distribution but over the last decade there has been increasing interest in distributional regression approaches modeling all parameters including location, scale, and shape. Notably, so-called non-homogeneous Gaussian regression (NGR) models both mean and variance of a Gaussian response and is particularly popular in weather forecasting. Moreover, generalized additive models for location, scale, and shape (GAMLSS) provide a framework where each distribution parameter is modeled separately capturing smooth linear or nonlinear effects. However, when variable selection is required and/or there are non-smooth dependencies or interactions (especially unknown or of high-order), it is challenging to establish a good GAMLSS. A natural alternative in these situations would be the application of regression trees or random forests but, so far, no general distributional framework is available for these. Therefore, a framework for distributional regression trees and forests is proposed that blends regression trees and random forests with classical distributions from the GAMLSS framework as well as their censored or truncated counterparts. To illustrate these novel approaches in practice, they are employed to obtain probabilistic precipitation forecasts at numerous sites in a mountainous region based on a large number of numerical weather prediction quantities. It is shown that the novel distributional regression forests automatically select variables and interactions, performing on par or often even better than GAMLSS specified either through prior meteorological knowledge or a computationally more demanding boosting approach.