Convolutional conditional neural processes for local climate downscaling

TL;DR

This paper introduces convolutional conditional neural processes (convCNPs) for multisite climate downscaling, enabling flexible, accurate predictions at arbitrary locations and improving the representation of extreme events over existing methods.

Contribution

The paper presents a novel convCNP model that outperforms traditional downscaling techniques and Gaussian process interpolation, especially for extreme precipitation events.

Findings

ConvCNPs outperform existing methods in Europe for temperature and precipitation.

ConvCNPs provide accurate multisite predictions at arbitrary locations.

Significant improvement in modeling extreme precipitation events.

Abstract

A new model is presented for multisite statistical downscaling of temperature and precipitation using convolutional conditional neural processes (convCNPs). ConvCNPs are a recently developed class of models that allow deep learning techniques to be applied to off-the-grid spatio-temporal data. This model has a substantial advantage over existing downscaling methods in that the trained model can be used to generate multisite predictions at an arbitrary set of locations, regardless of the availability of training data. The convCNP model is shown to outperform an ensemble of existing downscaling techniques over Europe for both temperature and precipitation taken from the VALUE intercomparison project. The model also outperforms an approach that uses Gaussian processes to interpolate single-site downscaling models at unseen locations. Importantly, substantial improvement is seen in the representation of extreme precipitation events. These results indicate that the convCNP is a robust downscaling model suitable for generating localised projections for use in climate impact studies, and motivates further research into applications of deep learning techniques in statistical downscaling.