HadUK-Grid is a dataset of gridded in situ climate data for the UK. It comprises monthly, seasonal, annual and long-term average values for 11 variables at 1km resolution. Three of these variables are also available at daily resolution. It also contains area averages for the UK, constituent countries, administrative regions and river basins. The first version was released in 2018 and there have been annual updates since then. The latest version is v1.3.1.0 which includes data to the end of 2024.

In previous work we investigated the uncertainties in the gridded data using a leave-one-out cross-validation approach to calculate an RMS error. This gave a more detailed understanding of the characteristics of gridded UK climate data than had been available previously. However, it also highlighted several unexpected features as well as some deficiencies. These included: seasonal cycles and step changes in the RMSE for maximum temperature (that were not visible in minimum temperature), outliers in the RMSE time series caused by small numbers of poorly predicted stations, and pronounced trends in the RMSE for individual stations.

In this presentation we will give an update on progress towards understanding and, where appropriate, correcting these issues, including: an intercomparison of the results for different station types (coastal/inland, lowland/upland); an audit of which individual stations are having the most influence on the RMSE; an assessment of whether changes in exposure, equipment type or reporting practice might explain some of the observed trends or step changes; an investigation into ways to improve the gridding process in areas where the network is sparse, especially on the edge of the domain; and some thoughts on the relative importance of data quality, station density and gridding method.

How to cite: Hollis, D., Carlisle, E., and Kendon, M.: Gridding uncertainties in UK climate data – progress and improvements, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-649, https://doi.org/10.5194/ems2025-649, 2025.

For several years now, MeteoSwiss has been developing and producing monthly grid datasets for temperature and precipitation that extend over 150 years in the past and are highly consistent over time. Therefore, they are particularly suitable for applications requiring high standards in temporal consistency, such as climate monitoring, trend calculation, and the study of long-term climate variations. Long-term climate datasets are generated using statistical reconstruction methods that explicitly avoid artefacts from variations in observational density over time.

In this contribution, we present two recent developments in our suite of long-term climate datasets: Firstly, a new dataset has been developed for relative sunshine duration. Despite the methodological experience from previous reconstructions, this development was particularly challenging due to the scarcity of the available measurements and the related complications for homogenization in the complex terrain of Switzerland. We outline the necessary methodological adjustments, illustrate results from interesting example cases, and present results on the temporal and spatial changes in sunshine duration in Switzerland since the beginning of the 20^th century. A brief excursus refers to the absolute sunshine duration, which is now also available as a grid product.

Secondly, we present an update of the pan-Alpine long-term precipitation dataset “LAPrec”, developed during the COPERNICUS C3S_311a_Lot4 Project. The update builds on a substantial extension of the period for calibrating the reconstruction method. This was possible thanks to the longer availability of one of the key data sources of LAPrec, the high-resolution dataset “APGD”, which newly extends from 1971 to 2019. The long-term station data used in the update was collected and homogenized at Geosphere Austria. Results from the updated LAPrec will be presented on long-term precipitation changes in the Alps. The dataset is available in the Copernicus climate data store.

How to cite: Isotta, F., Begert, M., Chimani, B., Hiebl, J., and Frei, C.: New developments in long-term spatial climate datasets for Switzerland and the European Alps, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-173, https://doi.org/10.5194/ems2025-173, 2025.

Ljubljana is the largest urban settlement in Slovenia, resulting in the most pronounced urban heat island (UHI) effect in the country. While numerous in-depth studies of the UHI effect on the surface and in the canopy have already been conducted in the Slovenian capital, our project, funded by the Municipality of Ljubljana, seeks to gain new insights into this phenomenon by applying advanced GIS and GeoAI methods.

As part of the project, a network of 32 urban meteorological stations was set up to record air temperature, relative humidity and air pressure at 10-minute intervals and display them in real time. The station locations were selected based on expert knowledge of the spatial variability of UHI in conjunction with the Local Climate Zone (LCZ) classification.

The air temperature data obtained from these stations were used to map canopy UHI at different times of the day, under optimal meteorological conditions for canopy UHI formation on November 1, 2024, characterised by anticyclonic weather and low wind speeds. A random forest regression model was used for the mapping.

To support the model, several spatial layers representing explanatory variables were derived from various remote sensing datasets, including LiDAR, satellite imagery and orthophotos. These variables included impervious area, mean height of surface features, average building height, mean tree canopy height, normalised difference vegetation index (NDVI), sky view factor, land surface temperature, surface albedo, and land use, among others.

After selecting a subset of explanatory variables, regression modelling was performed at 30 m and 20 m spatial resolution. The results for both resolutions showed a comparable level of accuracy; however, the 20 m resolution performed slightly better (explained variance: 62.04 %; RMSE: 0.34 °C; MAE: 0.30 °C).

Although the main objective of this presentation is to present preliminary methodological results based on a single case study day, the ultimate goal of the project is to develop an algorithm for the automatic generation of canopy UHI maps to be made accessible via a web platform.

How to cite: Gregorčič, T., Ogrin, M., Miklavcic, I., and Svetlin, D.: A new approach to studying the urban heat island in Ljubljana, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-335, https://doi.org/10.5194/ems2025-335, 2025.

Near-surface meteorological variables can be reconstructed using various methods, including geostatistical approaches like Kriging or techniques based on inverse problem theory, such as Optimal Interpolation. In recent years, machine learning methods have also been applied to this type of problem, including the reconstruction of meteorological fields near the surface.

Machine learning methods require large training datasets. At MET Norway, we have been collecting hourly temperature observations from personal weather stations (PWS) managed by private individuals. These data have been used operationally since 2018. Compared to conventional observation networks, PWS data increase the number of available training samples by a factor of 50 or more. However, their use also introduces specific challenges, such as data quality and representativeness. We will describe the steps taken to make use of PWS data in an effective way.

We apply a neural network model to estimate the hourly temperature at a location based on the 20 nearest observations. The aim is to estimate both the expected value and its associated uncertainty. The neural network is based on an approach developed for post-processing numerical weather prediction output, which provides probabilistic forecasts in the form of quantile functions. These functions are represented as linear combinations of Bernstein basis polynomials, with the coefficients predicted by the network.

The implementation is done in Python using the JAX library and training is run on a single Nvidia A30 GPU with 24 GB RAM. Further details about the training parameters used will be provided during the presentation.

The geographical domain covers Fennoscandia and the Baltic states. We use hourly temperature data from January 2020 to July 2024 and train a separate model for each month. The number of training samples per month ranges from approximately 145 million to 180 million.

The work is still at an early stage. We are currently addressing questions such as: What is the accuracy and precision of the predicted temperature values? Are the uncertainty estimates reliable? Does including temporal and spatial context in the training data improve the results?

How to cite: Lussana, C., Carrer, M., Båserud, L., Turin, G., and Bremnes, J. B.: Deep Neural Networks for the reconstruction of near-surface hourly temperature, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-219, https://doi.org/10.5194/ems2025-219, 2025.

Understanding sub-daily temperature variations is crucial in changing climate, especially as the diurnal temperature range (DTR) serves as a sensitive indicator of local and regional climatic shifts. DTR tends to be higher in arid regions, and its amplification directly affects human health – particularly cardiovascular and respiratory conditions – as well as agricultural productivity and energy demand. Rapid temperature changes within a day can stress crops, increase respiratory activity in plants, and result in spikes in heating or cooling requirements. 

This study investigates sub-daily temperature dynamics over South-East Europe using 3-hourly temperature data from the ERA5-Land reanalysis (1971–2024) and high-resolution (the atmosphere on 25 km) CMIP6 climate model simulations (1971–2050). The focus is on characterising temporal trends and potential climatic drivers behind observed and projected changes of at least 4-10 °C within 3 hours. 

Preliminary results reveal a notable increase in large positive sub-daily temperature changes – especially during transient months (March, April, September, October) – over the Pannonian Basin, predominantly during warm periods. Furthermore, an increase in large negative temperature shifts is detected from February to September, often associated with warm spells and precipitation events following warm spells. These findings suggest an evolving pattern of intra-day thermal variability, influenced by seasonal warming and atmospheric instability. 

The analysis highlights the role of cloud cover, soil moisture, relative humidity, and dew-point temperature as contributing factors. Changes in these parameters influence both the magnitude and direction of sub-daily temperature shifts. Understanding these mechanisms is essential for developing adaptation strategies in agriculture, urban planning and public health sectors.

Acknowledgements. This work has been implemented by the National Multidisciplinary Laboratory for Climate Change (RRF-2.3.1-21-2022-00014) project within the framework of Hungary's National Recovery and Resilience Plan supported by the Recovery and Resilience Facility of the European Union. In addition, this study has been supported by the European Climate Fund (G-2409-68866).

How to cite: Szabó, P. and Pongrácz, R.: Analysis of sub-daily temperature variations of past and future climates in South-East Europe, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-437, https://doi.org/10.5194/ems2025-437, 2025.

The latest generation of high-resolution and convection-permitting reanalyses, which allows the representation of the atmospheric processes at small spatial scales (<=4 km), are crucial to understand the evolution in time and space of phenomena such as mesoscale systems, convective storms, and orographic precipitation. Given the availability of long (>35 years) and continuum datasets of convection-permitting reanalysis data over Italy, this work aims to use them to investigate the occurrence of extreme events and to quantify their possible intensification over time in this area.

Previous studies have compared those convection-permitting reanalyses and validated against observations their precipitation fields from the climatological to the daily scale, demonstrating that they can capture fine-scale precipitation events, though spatial mismatches with observations sometimes occur. When transitioning to sub-daily precipitation fields, new challenges arise: the scarcity and inhomogeneities in observations increase, impacting the quality of the assimilation in the reanalyses; moreover, the representation of the small-scale physical processes increase in complexity, giving even more space to uncertainties in the results.

This work tackles these issues by employing an event-based approach when analysing sub-daily precipitations, focusing on the Italian domain, where extreme events cause every year many damages to the population and infrastructures. Using the MERIDA HRES and MOLOCH convection-permitting reanalyses over Italy (1986–2022), precipitation events are identified from hourly fields by applying a 1 mm threshold and a clustering technique. This resulted in a dataset of approximately 250.000 events, each of them characterized by its extent, timing, peak values, average intensity, and shape.

The resulting event-based dataset was used to calculate climatological averages and trends over time of these characteristics, both at the annual and seasonal aggregation. Subsequently, from the whole dataset, filtering criteria were applied to subset the most extreme events. Since there is no single criterion to define what constitutes an extreme event, multiple filtering criteria were applied to capture different types of extremes. Preliminary results, based on thresholds using the average of the annual maxima of hourly values, reveal a significant increase in extreme precipitation events in the pre-Alpine region and Western Alps during summer. In autumn, an increasing trend emerges along the southern and insular Italian coastlines. This pattern aligns with regions and seasons where convective phenomena drive intense precipitation at small spatial scales.

Further analyses will consider different filtering criteria, to also capture events that are anomalous for a given time of the year and synoptic events that typically occur in winter and spring. The results of this work will not only shed light on the changing sub-hourly precipitation patterns over Italy but also provide guidance to reanalysis users and stakeholders in planning resilient infrastructure to prevent damage from extreme precipitation events.

How to cite: Cavalleri, F., Viterbo, F., Lussana, C., Manara, V., Bonanno, R., Brunetti, M., Lacavalla, M., and Maugeri, M.: Characterizing the increase of extreme precipitation through an event-based analysis of hourly convection-permitting reanalyses fields over Italy, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-90, https://doi.org/10.5194/ems2025-90, 2025.

High-resolution precipitation data are critical for addressing a wide range of hydrological and climatological challenges, including flood modelling, precipitation-runoff behaviour analyses, risk assessments, and climate trend studies. Despite their importance, no existing dataset combines the spatial resolution of 100 meters with hourly temporal resolution for large areas over extended historical periods. This gap severely limits the accuracy and applicability of current models and analyses. Our study introduces a innovative method to generate such datasets, covering the period from 1961 onward - the start of the modern climate reference period - thereby addressing this long-standing need.

The main challenge lies in the lack of suitable radar or station-based precipitation data before the early 2000s. While reanalysis products like ERA5 and ERA5-Land provide hourly data, their spatial resolutions (approximately 31 km and 9 km, respectively) are too coarse for applications requiring fine-scale insights. Existing disaggregation methods attempting to refine these datasets to higher resolutions often fall short in terms of consistency and accuracy. Our approach bridges this gap by combining machine learning techniques (gradient boosting decision trees), spatial interpolation methods, and meteorological station data at both hourly and daily resolutions.

The proposed deterministic method has been rigorously validated using spatial and temporal independent cross-validation with over 1,100 stations across Germany. The results demonstrate robust performance, achieving a correlation coefficient (R) exceeding 0.6 and a Heidke Skill Score (HSS) also above 0.6. These metrics underscore the method's reliability in accurately distinguishing between precipitation events and non-events while capturing precipitation intensity variations.

As a proof of concept, we applied our method to generate a high-resolution precipitation dataset for Saxony, Germany, which is freely available for analysis and application. This dataset represents a significant advancement in hydrological and climatological research tools, enabling unprecedented precision in modelling and analysis across diverse domains.

How to cite: Körner, P.: High-Resolution Hourly Precipitation Grids Since 1961: A Novel Deterministic Approach, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-342, https://doi.org/10.5194/ems2025-342, 2025.

A more accurate understanding of climate and its changes requires spatially representative climate databases. However, weather stations are not evenly distributed, consequently, the station network consists of both densely and sparsely covered subregions. In order to estimate the values of meteorological variables at points where no measurements are available, a spatial interpolation method must be used. Our gridded climate datasets are generated using the MISH (Meteorological Interpolation based on Surface Homogenized Data Basis) method (MISHv1.03 software). Using the MISH interpolation, we have spatially representative climate database.

In general, the spatial distribution of meteorological elements depends to a large extent on topography, e.g. altitude. For this reason, if a more accurate gridded database is to be created, then a more accurate elevation model is required. For Hungary, the elevation model used for MISH modelling was unchanged for the last more than 20 years. This year, a new and more accurate elevation model has been introduced, which resulted in a complete renewal of the climate statistical parameters per meteorological element for MISH interpolation. At the same time, new model variables have been introduced in addition to those previously used for modelling, further improving our gridded climate data series. In this presentation, we will present the results obtained for temperature and precipitation interpolation and, as it is particularly important for decision-makers to have as detailed information as possible on a given area, we will also present the ~1 km resolution climate database for Hungary.

Acknowledgements:

The development presented was carried out within the framework of the Széchenyi Plan Plus Program with the support DIMOP Plusz-2.3.1-23-2023-00001 project, and the EKÖP-KDP-24 University Excellence Cooperative Doctoral Program of the Ministry for Culture and Innovation from the source of the National Research, Development and Innovation fund.

How to cite: Szentes, O., Lakatos, M., and Pongrácz, R.: Innovations of the spatially representative climate data series for Hungary, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-329, https://doi.org/10.5194/ems2025-329, 2025.