Juan Pedro Díaz
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Juan Pedro Díaz has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Albano González
Univ. de La Laguna (Spain)
4 shared publications - Francisco Javier Exposito
Univ. of La Laguna (Spain)
4 shared publications - Juan Carlos Pérez
Universidad de La Laguna
3 shared publications - Marcos Esquivel-González
Universidad de La Laguna
3 shared publications - Jose González-Abad
Instituto de Física de Cantabria (IFCA), CSIC-Universidad de Cantabria, Santander, Spain
2 shared publications Samuel Márquez-Cubas
2 shared publications
Latest Publications
- Open Access
- 3 August 2026
- Published by Springer Nature in Earth Systems and Environment
- Vol. 10 (10) , 13001-13025
ClimDiT: A Generative Latent Diffusion Transformer Framework for Multivariate Climate Downscaling
Accurate, high-resolution local climate projections are essential for climate adaptation and impact assessment. Most statistical downscaling approaches rely on regression-based models that produce single deterministic estimates, thereby limiting their ability to capture the inherent variability and multivariate dependencies of the climate system. To address this limitation, ClimDiT is introduced, a generative multivariate statistical downscaling framework based on latent diffusion transformer models. Thus, instead of a single expected result, it estimates the full conditional distribution of high-resolution climate fields. The proposed system operates within a compressed, autoencoded latent space, enabling computationally efficient sampling while implicitly capturing inter-variable dependencies. In this work, ClimDiT simultaneously downscales daily maximum temperature, minimum temperature, and accumulated precipitation to approximately 5 km resolution over the Iberian Peninsula. It is trained using ERA5 predictors and ROCIO-IBEB high-resolution targets and its performance is evaluated using metrics aligned with the VALUE framework. ClimDiT attains deterministic accuracy comparable to, or moderately superior to, state-of-the-art regression networks, while improving spatial coherence. Although specialised stochastic baselines, which suffer from high deterministic errors, retain an edge in univariate and multivariate probabilistic scores, a ClimDiT ensemble generated from 50 samples demonstrates reasonable probabilistic skill and joint calibration of the three target variables (evaluated with CRPS, Brier, Energy and Variogram scores). Overall, ClimDiT demonstrates that latent diffusion models enable a generative pathway toward coherent, probabilistic, and physically consistent regional climate downscaling, advancing beyond the limitations of traditional regression-based AI systems. Graphical This graphical abstract provides a concise visual overview of ClimDiT: A Generative Latent Diffusion Transformer Framework for Multivariate Climate Downscaling. It summarizes the study objectives, methods, and main findings. At the top, the Data section illustrates the statistical downscaling approach, by linking coarse-resolution ERA5 predictors with high-resolution ROCIO-IBEB climate targets over the Iberian Peninsula. It highlights the three downscaled variables—maximum temperature, minimum temperature, and precipitation—as well as the corresponding training and testing periods. The Methods panel presents the core innovation: the ClimDiT model. It combines an autoencoder, which compresses climate fields into a latent space, with a latent diffusion transformer that progressively reconstructs high-resolution fields from Gaussian noise, conditioned on large-scale predictors. This reflects the proposed use of generative models, rather than the widely used regression-based AI approaches, as they allow the generation of multiple outputs for the same synoptic state. A benchmark comparison box displays the univariate state-of-the-art (SOTA) methods (DeepESD and U-Net) used as references, along with the tested loss functions (MSE, ASYM, and STO). STO refers to stochastic models, which use a Bernoulli–Gamma distribution for precipitation and a Gaussian distribution for temperature, trained using the negative log-likelihood loss. Additionally, a deterministic multivariate baseline (DeepESD-MULT, trained with MSE) is included to directly evaluate the impact of joint variable modeling within a traditional regression framework. The Results section visualizes performance across spectral, deterministic, and probabilistic metrics (RAPSD, MAE, CRPS, Variogram Score), emphasizing ClimDiT superior spatial coherence and balanced probabilistic skill. The Highlights box succinctly reinforces these contributions: a generative, latent-space-based approach that enables coherent multivariate downscaling.
- 3 August 2026
- Published by Springer Nature in Earth Systems and Environment
- Vol. 10 (10) , 13001-13025
Open AccessClimDiT: A Generative Latent Diffusion Transformer Framework for Multivariate Climate Downscaling
Accurate, high-resolution local climate projections are essential for climate adaptation and impact assessment. Most statistical downscaling approaches rely on regression-based models that produce single deterministic estimates, thereby limiting their ability to capture the inherent variability and multivariate dependencies of the climate system. To address this limitation, ClimDiT is introduced, a generative multivariate statistical downscaling framework based on latent diffusion transformer models. Thus, instead of a single expected result, it estimates the full conditional distribution of high-resolution climate fields. The proposed system operates within a compressed, autoencoded latent space, enabling computationally efficient sampling while implicitly capturing inter-variable dependencies. In this work, ClimDiT simultaneously downscales daily maximum temperature, minimum temperature, and accumulated precipitation to approximately 5 km resolution over the Iberian Peninsula. It is trained using ERA5 predictors and ROCIO-IBEB high-resolution targets and its performance is evaluated using metrics aligned with the VALUE framework. ClimDiT attains deterministic accuracy comparable to, or moderately superior to, state-of-the-art regression networks, while improving spatial coherence. Although specialised stochastic baselines, which suffer from high deterministic errors, retain an edge in univariate and multivariate probabilistic scores, a ClimDiT ensemble generated from 50 samples demonstrates reasonable probabilistic skill and joint calibration of the three target variables (evaluated with CRPS, Brier, Energy and Variogram scores). Overall, ClimDiT demonstrates that latent diffusion models enable a generative pathway toward coherent, probabilistic, and physically consistent regional climate downscaling, advancing beyond the limitations of traditional regression-based AI systems. Graphical This graphical abstract provides a concise visual overview of ClimDiT: A Generative Latent Diffusion Transformer Framework for Multivariate Climate Downscaling. It summarizes the study objectives, methods, and main findings. At the top, the Data section illustrates the statistical downscaling approach, by linking coarse-resolution ERA5 predictors with high-resolution ROCIO-IBEB climate targets over the Iberian Peninsula. It highlights the three downscaled variables—maximum temperature, minimum temperature, and precipitation—as well as the corresponding training and testing periods. The Methods panel presents the core innovation: the ClimDiT model. It combines an autoencoder, which compresses climate fields into a latent space, with a latent diffusion transformer that progressively reconstructs high-resolution fields from Gaussian noise, conditioned on large-scale predictors. This reflects the proposed use of generative models, rather than the widely used regression-based AI approaches, as they allow the generation of multiple outputs for the same synoptic state. A benchmark comparison box displays the univariate state-of-the-art (SOTA) methods (DeepESD and U-Net) used as references, along with the tested loss functions (MSE, ASYM, and STO). STO refers to stochastic models, which use a Bernoulli–Gamma distribution for precipitation and a Gaussian distribution for temperature, trained using the negative log-likelihood loss. Additionally, a deterministic multivariate baseline (DeepESD-MULT, trained with MSE) is included to directly evaluate the impact of joint variable modeling within a traditional regression framework. The Results section visualizes performance across spectral, deterministic, and probabilistic metrics (RAPSD, MAE, CRPS, Variogram Score), emphasizing ClimDiT superior spatial coherence and balanced probabilistic skill. The Highlights box succinctly reinforces these contributions: a generative, latent-space-based approach that enables coherent multivariate downscaling.
- Open Access
- 3 July 2026
- Published by Springer Nature in Research Square
ClimDiT: A Generative Latent Diffusion Transformer Framework for Multivariate Climate Downscaling
Accurate, high-resolution local climate projections are essential for climate adaptation and impact assessment. Most statistical downscaling approaches rely on regression-based models that produce single deterministic estimates, thereby limiting their ability to capture the inherent variability and multivariate dependencies of the climate system. To address this limitation, ClimDiT is introduced, a generative multivariate statistical downscaling framework based on latent diffusion transformer models. Thus, instead of a single expected result, it estimates the full conditional distribution of high-resolution climate fields. The proposed system operates within a compressed, autoencoded latent space, enabling computationally efficient sampling while implicitly capturing inter-variable dependencies. In this work, ClimDiT simultaneously downscales daily maximum temperature, minimum temperature, and accumulated precipitation to approximately 5 km resolution over the Iberian Peninsula. It is trained using ERA5 predictors and ROCIO-IBEB high-resolution targets and its performance is evaluated using metrics aligned with the VALUE framework. ClimDiT attains deterministic accuracy comparable to, or moderately superior to, state-of-the-art regression networks, while improving spatial coherence. Although specialised stochastic baselines, which suffer from high deterministic errors, retain an edge in univariate and multivariate probabilistic scores, a ClimDiT ensemble generated from 50 samples demonstrates reasonable probabilistic skill and joint calibration of the three target variables (evaluated with CRPS, Brier, Energy and Variogram scores). Overall, ClimDiT demonstrates that latent diffusion models enable a generative pathway toward coherent, probabilistic, and physically consistent regional climate downscaling, advancing beyond the limitations of traditional regression-based AI systems.
- 3 July 2026
- Published by Springer Nature in Research Square
Open AccessClimDiT: A Generative Latent Diffusion Transformer Framework for Multivariate Climate Downscaling
Accurate, high-resolution local climate projections are essential for climate adaptation and impact assessment. Most statistical downscaling approaches rely on regression-based models that produce single deterministic estimates, thereby limiting their ability to capture the inherent variability and multivariate dependencies of the climate system. To address this limitation, ClimDiT is introduced, a generative multivariate statistical downscaling framework based on latent diffusion transformer models. Thus, instead of a single expected result, it estimates the full conditional distribution of high-resolution climate fields. The proposed system operates within a compressed, autoencoded latent space, enabling computationally efficient sampling while implicitly capturing inter-variable dependencies. In this work, ClimDiT simultaneously downscales daily maximum temperature, minimum temperature, and accumulated precipitation to approximately 5 km resolution over the Iberian Peninsula. It is trained using ERA5 predictors and ROCIO-IBEB high-resolution targets and its performance is evaluated using metrics aligned with the VALUE framework. ClimDiT attains deterministic accuracy comparable to, or moderately superior to, state-of-the-art regression networks, while improving spatial coherence. Although specialised stochastic baselines, which suffer from high deterministic errors, retain an edge in univariate and multivariate probabilistic scores, a ClimDiT ensemble generated from 50 samples demonstrates reasonable probabilistic skill and joint calibration of the three target variables (evaluated with CRPS, Brier, Energy and Variogram scores). Overall, ClimDiT demonstrates that latent diffusion models enable a generative pathway toward coherent, probabilistic, and physically consistent regional climate downscaling, advancing beyond the limitations of traditional regression-based AI systems.
- Open Access
- 27 March 2026
- Published by Elsevier in Sustainable Energy Technologies and Assessments
- Vol. 88
Exploring optimized solutions for wind and solar power deployment in a case study for the Canary Islands
- 27 March 2026
- Published by Elsevier in Sustainable Energy Technologies and Assessments
- Vol. 88
Open AccessExploring optimized solutions for wind and solar power deployment in a case study for the Canary Islands
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
New CORDEX Flagship Pilot Study on Island processes in the Macaronesian Archipelagos (FPS-I-Mac): Objectives and challenges
Small islands, particularly those with complex topography and fragile economies reliant on climate-sensitive sectors like tourism and agriculture, are highly vulnerable to climate change. Challenges such as rising sea levels, tropical and extratropical cyclones, increasing temperatures, and shifting rainfall patterns significantly affect these territories. However, current global climate models (GCMs) lack the resolution needed to capture critical local processes essential for these regions. The “FPS on Macaronesian Archipelagos. Convection Permitting projections focused on island processes (FPS-I-Mac)” is a Flagship Pilot Study (FPS) initiative focuses on Macaronesia, an Atlantic region comprising the Azores, Madeira, Cape Verde, and the Canary Islands archipelagos.Each archipelago exhibits unique climatic characteristics, influenced by its topography and geographical location. For instance, the Azores are the rainiest, while the Canary Islands are more influenced by continental air masses. All four archipelagos experience distinct annual cycles, with heavier rainfall during autumn and winter. Unlike other regions, intense summer convective rain is uncommon.This communication presents the main objectives and challenges of this FPS: Investigate multiscale climatic processes and their interactions to improve climate projections in Macaronesia. Assess the influence of sea surface temperature (SST) and aerosols in these oceanic regions. Compare high-resolution simulations with standard-resolution models to evaluate their added value in extreme event analysis. Create a shared database and produce ensembles of climate projections generated by different dynamical techniques (as models, empirical-statistical downscaling, hybrid techniques as emulators) tailored for Vulnerability, Impacts, and Adaptation (VIA) communities. Foster collaboration between scientific communities and end-users. Expected Impact are, between others: Enhanced understanding and modeling of climate in islands with complex topographies. Reduction of uncertainties in 21st-century climate projections for the Macaronesian region. Provision of accurate data for key sectors such as agriculture, health, energy meteorology, and risk management. Strengthened synergies among stakeholders, promoting a fair ecological transition in these vulnerable regions. The FPS aligns with CORDEX objectives by improving regional climate projections for small islands, which are underexplored in current studies. We will use different regional climate models (RCMs) at convection-permitting resolution, but also empirical-statistical downscaling (ESD), and hybrid strategies for high-resolution simulations, enabling detailed analysis of extreme events. Local and satellite observations associated with reanalysis products will validate these processes. By contributing to global discussions and integrating end-user needs from the outset, this initiative will support both local decision-making and broader climate research in the area during the next 5 years. The current team includes research groups from Portugal, Switzerland, Italy, and Spain, but it is open to new participants in different topics.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessNew CORDEX Flagship Pilot Study on Island processes in the Macaronesian Archipelagos (FPS-I-Mac): Objectives and challenges
Small islands, particularly those with complex topography and fragile economies reliant on climate-sensitive sectors like tourism and agriculture, are highly vulnerable to climate change. Challenges such as rising sea levels, tropical and extratropical cyclones, increasing temperatures, and shifting rainfall patterns significantly affect these territories. However, current global climate models (GCMs) lack the resolution needed to capture critical local processes essential for these regions. The “FPS on Macaronesian Archipelagos. Convection Permitting projections focused on island processes (FPS-I-Mac)” is a Flagship Pilot Study (FPS) initiative focuses on Macaronesia, an Atlantic region comprising the Azores, Madeira, Cape Verde, and the Canary Islands archipelagos.Each archipelago exhibits unique climatic characteristics, influenced by its topography and geographical location. For instance, the Azores are the rainiest, while the Canary Islands are more influenced by continental air masses. All four archipelagos experience distinct annual cycles, with heavier rainfall during autumn and winter. Unlike other regions, intense summer convective rain is uncommon.This communication presents the main objectives and challenges of this FPS: Investigate multiscale climatic processes and their interactions to improve climate projections in Macaronesia. Assess the influence of sea surface temperature (SST) and aerosols in these oceanic regions. Compare high-resolution simulations with standard-resolution models to evaluate their added value in extreme event analysis. Create a shared database and produce ensembles of climate projections generated by different dynamical techniques (as models, empirical-statistical downscaling, hybrid techniques as emulators) tailored for Vulnerability, Impacts, and Adaptation (VIA) communities. Foster collaboration between scientific communities and end-users. Expected Impact are, between others: Enhanced understanding and modeling of climate in islands with complex topographies. Reduction of uncertainties in 21st-century climate projections for the Macaronesian region. Provision of accurate data for key sectors such as agriculture, health, energy meteorology, and risk management. Strengthened synergies among stakeholders, promoting a fair ecological transition in these vulnerable regions. The FPS aligns with CORDEX objectives by improving regional climate projections for small islands, which are underexplored in current studies. We will use different regional climate models (RCMs) at convection-permitting resolution, but also empirical-statistical downscaling (ESD), and hybrid strategies for high-resolution simulations, enabling detailed analysis of extreme events. Local and satellite observations associated with reanalysis products will validate these processes. By contributing to global discussions and integrating end-user needs from the outset, this initiative will support both local decision-making and broader climate research in the area during the next 5 years. The current team includes research groups from Portugal, Switzerland, Italy, and Spain, but it is open to new participants in different topics.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Probabilistic Postprocessing of Hourly Precipitation Ensemble Forecasts Using UNet
Title: Probabilistic Postprocessing of Hourly Precipitation Ensemble Forecasts Using UNet Authors: Marcos Esquivel-González, Albano González, Juan Carlos Pérez, Juan Pedro Díaz, Pierre Simon TondreauAffiliation of authors: Grupo de Observación de la Tierra y la Atmósfera (GOTA), Avenida Astrofísico Francisco Sánchez, s/n, La Laguna, 38200, Canary Islands, Spain Reliable precipitation forecasting is crucial in sectors like public safety, agriculture and water management. Numerical Weather Prediction (NWP) models, which form the backbone of modern forecasting, are prone to errors due to their limitations and the chaotic behavior of equations, requiring postprocessing to improve accuracy and quantify uncertainties. Thus, this study evaluates probabilistic postprocessing models tailored for the Canary Islands, with the aim of enhancing Weather Research and Forecasting (WRF) ensemble forecasting accuracy in hourly precipitation forecast. UNet-based models were explored using two approaches, one incorporating the full set of km-scale convection-permitting ensemble forecast simulations (25) and another applying dimensionality reduction via Principal Component Analysis (PCA) and feature selection methods. These models were compared to traditional benchmarks like the Censored Shifted Gamma Distribution (CSGD) with Ensemble Model Output Statistics (EMOS) and the Analog Ensemble method. In the analysis of the results, not only the reliability of the predictions for the set of available meteorological stations was considered, but also the generalization capacity of the UNet models to obtain precipitation predictions for the whole region.In general, UNet models outperformed traditional approaches. The UNet with PCA excelled in probabilistic and deterministic metrics but struggled in regions without weather station data. Conversely, the UNet with feature selection, while slightly less accurate overall station locations, showed better generalization to unseen locations, maintaining consistent performance across the region and reducing computational demand. Additionally, the Integrated Gradients technique, an interpretability method that quantifies the contribution of each input feature to a model’s predictions by analyzing gradients, was employed to evaluate the impact of input variables on model performance. This analysis revealed that the integration of digital terrain elevation data significantly contributed to the UNet's outputs, underscoring the importance of topographic data in rainfall prediction.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessProbabilistic Postprocessing of Hourly Precipitation Ensemble Forecasts Using UNet
Title: Probabilistic Postprocessing of Hourly Precipitation Ensemble Forecasts Using UNet Authors: Marcos Esquivel-González, Albano González, Juan Carlos Pérez, Juan Pedro Díaz, Pierre Simon TondreauAffiliation of authors: Grupo de Observación de la Tierra y la Atmósfera (GOTA), Avenida Astrofísico Francisco Sánchez, s/n, La Laguna, 38200, Canary Islands, Spain Reliable precipitation forecasting is crucial in sectors like public safety, agriculture and water management. Numerical Weather Prediction (NWP) models, which form the backbone of modern forecasting, are prone to errors due to their limitations and the chaotic behavior of equations, requiring postprocessing to improve accuracy and quantify uncertainties. Thus, this study evaluates probabilistic postprocessing models tailored for the Canary Islands, with the aim of enhancing Weather Research and Forecasting (WRF) ensemble forecasting accuracy in hourly precipitation forecast. UNet-based models were explored using two approaches, one incorporating the full set of km-scale convection-permitting ensemble forecast simulations (25) and another applying dimensionality reduction via Principal Component Analysis (PCA) and feature selection methods. These models were compared to traditional benchmarks like the Censored Shifted Gamma Distribution (CSGD) with Ensemble Model Output Statistics (EMOS) and the Analog Ensemble method. In the analysis of the results, not only the reliability of the predictions for the set of available meteorological stations was considered, but also the generalization capacity of the UNet models to obtain precipitation predictions for the whole region.In general, UNet models outperformed traditional approaches. The UNet with PCA excelled in probabilistic and deterministic metrics but struggled in regions without weather station data. Conversely, the UNet with feature selection, while slightly less accurate overall station locations, showed better generalization to unseen locations, maintaining consistent performance across the region and reducing computational demand. Additionally, the Integrated Gradients technique, an interpretability method that quantifies the contribution of each input feature to a model’s predictions by analyzing gradients, was employed to evaluate the impact of input variables on model performance. This analysis revealed that the integration of digital terrain elevation data significantly contributed to the UNet's outputs, underscoring the importance of topographic data in rainfall prediction.
- Open Access
- 25 January 2024
- Published by Elsevier in Chemosphere
- Vol. 352, 141303
Transport and deposition of radionuclides from northern Africa to the southern Iberian Peninsula and the Canary Islands during the intense dust intrusions of March 2022
- 25 January 2024
- Published by Elsevier in Chemosphere
- Vol. 352, 141303
Open AccessTransport and deposition of radionuclides from northern Africa to the southern Iberian Peninsula and the Canary Islands during the intense dust intrusions of March 2022
- Open Access
- 12 April 2023
- Published by Springer Nature in npj Climate and Atmospheric Science
- Vol. 6 (1) , 1-11
The uneven impact of climate change on drought with elevation in the Canary Islands
Climate models project an increase in drought and aridity in many regions in response to greenhouse gas concentrations in the atmosphere. In areas with complex topography, such as the Canary Islands, elevation gradients may play an important role in future changes. Convection-permitting climate simulations driven by data from three global climate models included in the Coupled Model Intercomparison Project (CMIP5) have been performed for the Canary Islands. A significant increase in the duration and severity of drought is projected by the end of the twenty-first century (2070–2099), relative to the recent past (1980–2009), under intermediate and high emissions scenarios. In addition, the percentage of land affected by droughts, on average, would increase considerably, covering up to 96% in the higher elevations, in the business-as-usual scenario. These changes and the increase in aridity are more pronounced at higher altitudes due to a clear dependence of temperature rise as a function of elevation and a substantial decrease in precipitation.
- 12 April 2023
- Published by Springer Nature in npj Climate and Atmospheric Science
- Vol. 6 (1) , 1-11
Open AccessThe uneven impact of climate change on drought with elevation in the Canary Islands
Climate models project an increase in drought and aridity in many regions in response to greenhouse gas concentrations in the atmosphere. In areas with complex topography, such as the Canary Islands, elevation gradients may play an important role in future changes. Convection-permitting climate simulations driven by data from three global climate models included in the Coupled Model Intercomparison Project (CMIP5) have been performed for the Canary Islands. A significant increase in the duration and severity of drought is projected by the end of the twenty-first century (2070–2099), relative to the recent past (1980–2009), under intermediate and high emissions scenarios. In addition, the percentage of land affected by droughts, on average, would increase considerably, covering up to 96% in the higher elevations, in the business-as-usual scenario. These changes and the increase in aridity are more pronounced at higher altitudes due to a clear dependence of temperature rise as a function of elevation and a substantial decrease in precipitation.
- 1 July 1997
- Published by Taylor & Francis in International Journal of Remote Sensing
- Vol. 18 (10) , 2247-2252
First intercomparison between the aerosols optical depth data obtained by the NOAA satellite sensors and the Optronic OL752 spectroradiometer in the Canary Islands
The aerosol optical depth data obtained by two methods are compared. The first method uses solar radiance measurements with the spectroradiometer Optronic OL752 installed inside the Santa Cruz de Tenerife Harbour (28.48 N, 16.24 W). The second one takes the solar radiation scattered by the atmosphere using the Advanced Very High Resolution Radiometer (AVHRR) radiometer channels on board the NOAA meteorological satellites. A 91 per cent correlation between the two procedures was obtained for channel 1 (630nm).
- 1 July 1997
- Published by Taylor & Francis in International Journal of Remote Sensing
- Vol. 18 (10) , 2247-2252
First intercomparison between the aerosols optical depth data obtained by the NOAA satellite sensors and the Optronic OL752 spectroradiometer in the Canary Islands
The aerosol optical depth data obtained by two methods are compared. The first method uses solar radiance measurements with the spectroradiometer Optronic OL752 installed inside the Santa Cruz de Tenerife Harbour (28.48 N, 16.24 W). The second one takes the solar radiation scattered by the atmosphere using the Advanced Very High Resolution Radiometer (AVHRR) radiometer channels on board the NOAA meteorological satellites. A 91 per cent correlation between the two procedures was obtained for channel 1 (630nm).