Convolutional conditional neural processes for local climate downscaling

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.

Federated Condition Monitoring Signal Prediction With Improved Generalization

“…For example, studies aim to capture global uncertainties in the encoding process [7,27], mitigate underfitting to the context points [13,28,29], account for correlations across functions [30,31] or across target output points [32], ensure invariant predictions under input shifts or transformations [33]. Built upon the methodological advances above, NPs have been widely used in a range of domains, including clinical data analysis [34], climate science [35], image classification [36], robotics [37,38], and so on. Despite that, their applications in the predictive modeling of CM signals have yet to be investigated.…”

Reply on RC2

“…Finally, we investigate current state-of-the-art downscaling models: Convolutional Conditional Neural Processes (ConvCNP) (Gordon et al, 2019;Vaughan et al, 2021) and Convolutional Gaussian Neural Processes (ConvGNP) (Markou et al, 2022;Andersson et al, 2023). These models offer similar advantages to the MFGP model, including: capturing extreme values and spatiotemporal structure, generalising to multiple locations, predicting at arbitrary locations and overcoming gridding biases.…”

“…Traditional and state-of-the-art statistical downscaling techniques are used to address these problems but present their own issues. These methods generally struggle to simultaneously solve the following problems: 1/ capturing extreme values and spatiotemporal structure, 2/ generalising to multiple locations, 3/ predicting at arbitrary locations, 4/ overcoming gridding biases and 5/ working effectively with sparse and 'small' datasets (King et al, 2013;Maraun and Widmann, 2018;Baño-Medina et al, 2020;Vaughan et al, 2021;Andersson et al, 2023). We propose Multi-Fidelity Gaussian Processes (MFGPs) as an alternative to other statistical downscaling and bias-correction methods.…”

Downscaling precipitation over High Mountain Asia using Multi-Fidelity Gaussian Processes: Improved estimates from ERA5

“…Finally, we investigate current state-of-the-art downscaling models: Convolutional Conditional Neural Processes (ConvCNP) (Gordon et al, 2019;Vaughan et al, 2021) and Convolutional Gaussian Neural Processes (ConvGNP) (Markou et al, 2022;Andersson et al, 2023). These models offer similar advantages to the MFGP model, including: capturing extreme values and spatiotemporal structure, generalising to multiple locations, predicting at arbitrary locations and overcoming gridding biases.…”

“…Traditional and state-of-the-art statistical downscaling techniques are used to address these problems but present their own issues. These methods generally struggle to simultaneously solve the following problems: 1/ capturing extreme values and spatiotemporal structure, 2/ generalising to multiple locations, 3/ predicting at arbitrary locations, 4/ overcoming gridding biases and 5/ working effectively with sparse and 'small' datasets (King et al, 2013;Maraun and Widmann, 2018;Baño-Medina et al, 2020;Vaughan et al, 2021;Andersson et al, 2023). We propose Multi-Fidelity Gaussian Processes (MFGPs) as an alternative to other statistical downscaling and bias-correction methods.…”