For urban hydrology, rainfall time series and especially design values with high temporal resolution are crucial. Since most climate scenarios offer daily resolution only, statistical downscaling in time seems a promising and computational effective solution. In the presented method, rainfall is first disaggregated to continuous 5min time series, and subsequently design values are derived from these time series.

The micro-canonical cascade model (MRC) is chosen as downscaling method since it conserves the daily rainfall amounts exactly, so the resulting 5 min time series are coherent with the daily time series used as starting point. Rainfall extreme values are often linked to temperature (especially convective events, which are crucial for e.g. urban hydrology or insurance companies). Therefore, a temperature-dependent MRC is introduced in this study. Temperature-dependency is tested for minimum temperature, mean temperature and maximum temperature, which all allow a physical interpretation of rainfall extreme values and provide deeper insights into their future changes.

For this study 45 locations across Germany are selected. To ensure spatial coherence with the climate model data (~∆l=5 km*5 km), the YW dataset (radar-gauge-merged data) from the German Weather Service (DWD) with originally ∆l=1km*1 km and ∆t=5 min was aggregated in space and used for the estimation of the MRC parameters. The DWD core ensemble with six combinations of global and regional climate models is applied for the climate change analysis, for both, RCP4.5 and RCP8.5 scenario.

For the temperature-dependency, class widths of 5 K are chosen to include a representative number of time steps in each class. No significant influence on continuous rainfall characteristics as wet spell amount, average intensity, wet and dry spell duration can be identified. To analyze the impact on rainfall extreme values peak-over-threshold series and 99.9 %-quantile q_99.9 are studied. While the reference model without temperature-dependency leads to higher overestimations for ∆t=5 min for ϑ<13 °C and underestimations for ϑ>18 °C, the temperature-dependency reduces the deviations over the whole range to a median overestimation of 1 mm/5 min (range of observations: 4 mm/5 min<q_99.9<6 mm/5 min). For peak-over-threshold, the overestimation of rainfall extreme values is reduced significantly by the introduction of the temperature-dependency.

Climate model data are disaggregated using both, MRC without and with temperature-dependent parameters. The rainfall extreme values are analyzed regarding their relative changes from the control period (1971-2000) to near-term (2021-2050) and long-term future (2071-2100). While extreme values from disaggregated time series without temperature-dependency indicate an increase of ‘only’ 12 % for the long-term future, the consideration of temperature shows an increase of 21 % (for duration D=1 h and return period T=2 yrs). Thus, neglecting the temperature impact leads to an underestimation of future rainfall extreme values.

How to cite: Müller-Thomy, H., Ebers, N., and Schröter, K.: High-resolution design rainfall estimation from climate model data, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-15880, https://doi.org/10.5194/egusphere-egu24-15880, 2024.

Analyzing the complex behavior of extreme precipitation events is essential for a better understanding of the effects of climate change on water resources and for forecasting extreme hydrologic events. In this study, complex network concepts are applied to investigate the synchronization patterns of extreme daily precipitation events across India, with an evaluation of how these patterns may vary in the future. Daily rainfall data provided by the India Meteorological Department (IMD) for the period 1961-2020 at a spatial resolution of 0.25ᵒ×0.25ᵒ are used to investigate the synchronization patterns of extreme rainfall during the historical period. To assess synchronization patterns in the future, rainfall projections from selected General Circulation Models under different Shared Socio-Economic Pathway Scenarios are employed. A day with precipitation greater than 1 mm is considered a wet day, and a wet day is then classified as an extreme precipitation event only if its precipitation exceeds the 95th percentile of all wet days. For the construction of the network, each grid is considered as a node, and the connections between them are identified using the event synchronization method. Both historical and future precipitation networks are analyzed for two different seasons: (i) Summer (June, July, August, and September); and (ii) Winter (December, January, and February). Two network measures, namely degree centrality and clustering coefficient, are determined for these networks. Changes in network measures, relative to the baseline period of 1961–2020, are analyzed across three different timeframes: the 2020s, 2040s, and 2070s. The findings from the network measures can reveal crucial geographic locations in terms of their connection patterns to other areas for both seasons.

How to cite: Bhadran, D. and Sivakumar, B.: Analysis of Extreme Precipitation Events in India under Shared Socioeconomic Pathway Scenarios: Application of Complex Networks, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-542, https://doi.org/10.5194/egusphere-egu24-542, 2024.

Projections from regional climate models are traditionally bias-corrected with ground observations before being used for hydrological modelling in order to improve the representation of local climate features. The choice of correction method affects hydrological projections, especially in mountain regions where the relationship between precipitation and temperature is a key property of the hydrological cycle as it controls the partitioning between solid and liquid precipitation. While several studies have investigated the sensitivity of hydrological projections (i.e., projections generated by feeding a hydrological model with climate projections) to the choice of bias correction technique, none have focused on single-model initial-condition large ensembles (SMILEs), which are a suitable tool for disentangling the response of hydrological extremes to climate change from their natural variability. In addition, the interaction between hydrological model choice and bias-correction method needs to be investigated in order to obtain more reliable hydrological projections.

The objective of this work is to identify the most appropriate statistical techniques to adjust climate SMILEs for studying changes in hydrological extremes in mountainous terrain. For this purpose, we bias-corrected the climate projections of two high-resolution SMILEs (0.11°) under the Representative Concentration Pathway 8.5 for the domain of Switzerland. Specifically, we used a 2 km gridded reanalysis derived from ground observations and three techniques to correct the precipitation and temperature time series: (1) univariate quantile mapping, (2) trend-preserving univariate quantile mapping, and (3) trend-preserving multivariate quantile mapping. Then, we used the bias-corrected time series as inputs to an ensemble of 11 hydrological models to simulate streamflow at the outlet of 93 near-natural Swiss catchments for the period 1955 - 2099. We compared the performance of the three bias-correction techniques with respect to their ability to simulate historical floods and droughts. In addition, we determined the most fit-for-purpose method by examining both the robustness of the corrections (e.g. towards model choice, transferability in time) and the sensitivity of future hydrological projections to the choice of bias correction technique.

How to cite: Astagneau, P. C., Wood, R., and Brunner, M. I.: Which bias correction methods are suited to represent hydrologic extremes in the Alps?, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6118, https://doi.org/10.5194/egusphere-egu24-6118, 2024.

High-resolution precipitation and temperature projections are indispensable for informed decision-making, risk assessment, and planning. Here, we have developed an extensive database of high-resolution (0.1°) precipitation, maximum, and minimum temperature projections extending till 2100 at a daily scale for Canada. We employed a novel Semi-Parametric Quantile Mapping (SPQM) methodology to bias-correct the Coupled Model Intercomparison Project, Phase-6 (CMIP6) projections for four distinct Shared Socio-economic Pathways. SPQM is simple, yet robust, in reproducing the observed marginal properties, trends, and variability according to future scenarios, and maintaining a smooth transition from observations to projected simulations. The database encompasses a substantial collection of 759 simulations derived from 37 diverse climate models for precipitation. Similarly, for maximum and minimum temperature projections, our database comprises 652 simulations from 30 climate models. These meticulously curated projections carry immense value for hydrological, environmental, and ecological studies, offering a comprehensive resource for analyses within these domains. Furthermore, these projections serve as a valuable asset for the quantification of uncertainties arising from variant labels, climate models, and future scenarios.

How to cite: Rajulapati, C. R., Abdelmoaty, H. M., Nerantzaki, S., and Papalexiou, S. M.: Bias-corrected high-resolution temperature and precipitation projections for Canada , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14970, https://doi.org/10.5194/egusphere-egu24-14970, 2024.

In the 21st century, the Mekong River Delta (MRD), in Viet Nam, is projected to experience intensified extreme precipitation events due to global warming. General Circulation Models (GCMs) offer possible future climate estimations globally that adeptly capture large-scale features of precipitation extremes under different scenario settings. However, challenges persist in replicating detailed regional flood patterns within the MRD. This study focuses on the application of CMIP6 models, including European Centre for Medium-Range Weather Forecasts Reanalysis v5—ERA5, Situ-Based Data Set of Temperature and Precipitation Extremes—HadEX3, and Rainfall Estimates on a Gridded Network—REGEN. Before it can be applied for the future flood assessment for the study area, this study compared annual maximum daily precipitation, derived from CMIP6, Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS), and ground-based precipitation records, from 1978 to 2012, and identified the relationships between annual maximum daily precipitation and flooded area. Accordingly, this study projects precipitation and flooded areas for the near future (2026–2050), mid-future (2050–2075), and far future (2075-2100). The preliminary results will be presented in this meeting. In light of the persistent global warming trend, the expected rise in flooding within the MRD and variations in heavy precipitation patterns emphasize the importance of the findings in this study. These results play a crucial role in mitigating adverse effects and fortifying resilience to global warming and climate change in the MRD.

KEY WORDS: CMIP6, CHIRPS, flood, extreme precipitation, global warming.

How to cite: Ton Binh, M. and Chiang, S.-H.: Examining the application of CMIP6 General Circulation Models in regional flood projections for the Mekong River Delta, Viet Nam, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14714, https://doi.org/10.5194/egusphere-egu24-14714, 2024.

How can climate model simulations for the future generate high-resolution sub-daily precipitation that could be trusted for a range of hydrologic design applications? This is a question that often provokes contradicting responses from climate scientists and hydrologists. Our study attempts to unify the knowledge gained from these two disciplines to present the first ever alternative for generating multi-scale (low-to-high frequency temporal persistence), high resolution (down to 1km if needed), sub-daily precipitation for future climates that is dynamically consistent with concurrent climate forcings (including atmospheric moisture, circulation patterns) and local topography. This dynamical precipitation generator contains three components. The first component is the raw temporal resolution climate field simulated using global climate models on the lower and the lateral boundaries of the spatial domain precipitation simulations are needed for. The second component is SDMBC, an innovative alternative for correcting systematic biases at Sub-Daily (SD) time steps, using the Multivariate Bias Correction (MBC) approach which corrects multivariate dependence, persistence and distributional attributes across variables that form the lateral and lower boundaries of the domain of interest. The third and final component is a Regional Climate Model (RCM), chosen to be the Weather Research and Forecasting (WRF) model for the present study, which uses the corrected lateral and lower boundary forcings generated from the second component of our framework, and generates dynamically consistent sub-daily precipitation along with other physically consistent atmospheric variables at high-resolution. It is shown here that use of this framework simulates precipitation fields that exhibit features consistent with observations including observed extremes, and storm events that are consistent with our expectations of how precipitation extremes will evolve in future (warmer) climates. A Python software (named SDMBC) that simplifies the implementation of the bias correction process is presented, and results are shown for simulations across the Australian domain. This software is now available from (https://pypi.org/project/sdmbc/) and can be used for applications over any domain worldwide in conjunction with WRF models that have been formulated independently.

How to cite: Sharma, A., Kim, Y., and Evans, J.: A dynamical alternative for simulating multi-scale high-resolution sub-daily space-time precipitation for future climates, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-7076, https://doi.org/10.5194/egusphere-egu24-7076, 2024.

Radar composite products are essential for tracking and forecasting heavy storms over mountainous catchments where rain gauge information is scarce. The impact of radar beam blockage from an individual radar, resulting in low reflectivity data, can significantly contribute to the underestimation of radar rainfall estimates in such areas. The quality of rain radar composites is critical as these products will be used for near real-time forecasting of hydrometeorological hazards. This study aimed to develop a relative quality index scheme based on the radar reflectivity fraction of the compositing radars to improve the accuracy of heavy rainfall estimates in (partly) blocked areas. Three additional surrounding environmental quality indices, i.e., the distance to the radar station, the height of the beam above the ground, and the radar beam blockage fraction were integrated in the overall quality indices (QI) computation. Furthermore, we expanded the use of the QI to enhance the mean field bias adjustment in tracking high-intensity rainfall. To comprehensively assess the merits and drawbacks of the compositing methods with multiple quality indices, we compared our results with conventional and well-known maximum composite techniques. We have tested this scheme in the Khao Yai National Park, Lamtakong basin, and the surrounding areas. Two rain radar stations were selected: Sattahip, 220 kilometer southwest and Phimai, 140 kilometer North of the Lamtakong basin. Automatic rain gauges in the overlapping area were used to evaluate the radar composite product during storm events in 2020 and 2022. The results indicate that radar composite approach with multiple QIs can effectively identify areas with unreliable radar measurement. The radar reflectivity fraction was the most important quality index in the composite region, especially in the beam blockage area where the reflectivity from Phimai consistently registers lower values compared to that from the Sattahip radar. Combining this novel relative QI scheme with traditional quality indices (distance, height, and beam blockage fraction) increased the overall accuracy and reliability of heavy radar rainfall estimates. While the combined QIs and the maximum composite method resulted in composite products with similar overall accuracy, the proposed new QI method provides more coherent storm structure. Furthermore, a noteworthy finding is that heavy rainstorms in obstructed areas become visibly apparent with higher accuracy when applying thresholds to the quality index values for bias adjustment computation of the composite products. These final products of radar rainfall estimates represent a critical advancement of rain radar based Early Warning Systems for hydrometeorological hazard mitigation in mountainous regions.

How to cite: Methaprayun, M., Bogaard, T. A., and Mapiam, P. P.: Developing a relative quality index scheme for improving radar composite products and bias adjustment in mountainous regions, Thailand, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-409, https://doi.org/10.5194/egusphere-egu24-409, 2024.

The spatial variability of rainfall is difficult to measure due to the lack of ground weather rain gauges. As a result, meteorological radars have become crucial sources of information for estimating precipitation fields. However, radar cluttering, which refers to external factors of nature that affect radar data quality, poses a significant challenge, and contributes to errors and uncertainties in the estimation process. In this study, we focused on analyzing the impact of cluttering on the Barrancabermeja C-band weather radar, situated between the Eastern and Central Ranges in the Colombian Andes. The analysis was conducted using radar information collected between 2019 and 2020. A frequency analysis of reflectivity of rainless day records showed the topographic interferences caused by the surrounding radar Ranges. Afterwards, the radar quality index (RQI) for both rainy and dry conditions was estimated considering factors such as clutter frequency map, partial beam blockage, effects of range distance quality, radar noise, and attenuation. The evaluation revealed an approximate clutter area of 50% in a beam elevation of 0.5°, primarily associated with the topographical interferences, indicating a direct impact of the Andean region on radar data quality.

Finally, we focused on intense rainfall events (greater than 10mm per event) to determine the parameters of three Quantitative Precipitation Estimation (QPE) relationships (Marshall & Palmer (1948), Seliga & Bringi (1976), and Sachidananda & Zrinc (1987) methodologies). Records from 91 available rain gauges were used to obtain relationships for individual gauges and, for four individual rings spaced 50 km from the radar. By employing these relationships, we calculated uncertainty maps of the quantitative precipitation estimation, obtaining an uncertainty of 60% from cluttering in the QPE of the meteorological radar. Overall, our findings emphasize the significant role of cluttering in the estimation of precipitation fields from the Barrancabermeja radar. The study underscores the importance of addressing cluttering effects and accounting for the topographical interferences in radar data interpretation to enhance the accuracy of quantitative precipitation estimates in the Andean region.

How to cite: Ramírez Tamayo, J. I., Piña Fulano, A. P., and Ladino Rincon, A.: Radar cluttering incidence in the estimation of rainfall fields in the Colombian Andes, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-1507, https://doi.org/10.5194/egusphere-egu24-1507, 2024.

In the past decade, there has been an increment in the magnitude and frequency of severe flood events across Bangalore city due to rise in rainfall intensities, increase in urban population, and drastic changes in urban landscape. The effect of urban development in a rapidly growing city has a substantial impact on the urban environment, leading to frequent flooding during monsoon and post-monsoon. Hence, a reliable forecasting system for rainfall at an urban scale is of priority to enhance the preparedness for disaster management. In this regard, a framework is developed for Bangalore City to dynamically downscale the daily rainfall prediction from National Centers for Environmental Prediction-Global Forecast System (NCEP-GFS) to high-resolution rainfall predictions, 3 km and 1 km spatial resolutions at 15-minute interval, employing the Weather Research and Forecasting (WRF) model. The model utilizes the initial and the boundary conditions forced at 06 UTC, resulting in a 24-hour lead-time forecast. The primary objective of this study is to test the performance of high-resolution WRF forecast with respect to the observed rainfall, using qualitative and quantitative statistical skill scores for the monsoon of 2023. Rank probability score at the municipal administrative level and performance indices such as critical success index, bias score, heidke skill score, false alarm ratio, and probability of detection at grid level are used for qualitative analysis. Whereas, quantitative measures are coefficient of determination, correlation, root mean square error, mean bias, and mean absolute error at grid as well as station levels. These metrics are estimated for various rainfall events and for different lead times. The study found that, the grid level correlation coefficient values for heavy rainfall events in 2023 fall in the range of 0.6 – 0.8 for the northern part of Bangalore city for both the spatial resolutions. Overall, our findings suggest that the forecasting framework can efficiently issue rainfall prediction with a lead time of 24 hours. This forecast can be further coupled with 1D and 2D hydrological models to predict flood inundation.

How to cite: Kusuman, K., Pentakota, L., Kishore, N., Gaddam, N., Rishika, A., Mujumdar, P. P., and Bhowmik, R. D.: A Qualitative and Quantitative Evaluation of the WRF Model Simulations for High Resolution Urban Rainfall Forecasting , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14385, https://doi.org/10.5194/egusphere-egu24-14385, 2024.

This paper presents an evaluation and sensitivity analysis of km-scale simulations of the unprecedented extreme rainfall event of July 2021 over Belgium and Germany, with a specific focus on sub-hourly extremes, size distributions and kinetic energy (KE) of rain. These variables are critical for hydrological applications, such as flood forecasting or soil loss monitoring, but are rarely directly obtained from Numerical Weather Prediction (NWP) or climate models. We present an extensive set of simulations exploring sensitivities to realistic variations in a newly implemented double-moment microphysics parameterization in the UK Met Office Unified Model. Most simulations reproduce the overall characteristics of the event, but overestimate the extreme rain rates. The rain rate - KE relation is captured well, despite too large volume-mean drop diameters. Amongst the sensitivities investigated, the representation of the raindrop self-collection - breakup equilibrium and the raindrop size-distribution shape have the most profound impact on the rainfall characteristics. While extreme rain rates vary within 30 %, the rain KE varies by a factor of four between the realistic perturbations to the microphysical assumptions. Changes to the aerosol concentration and the rain terminal velocity have a relatively smaller impact on the extreme rainfall characteristics. However, larger aerosol loading produces slightly smaller domain total rainfall, for which we propose a mechanism involving dynamical impacts of warm-rain suppression. Given the large uncertainties, a continued effort to improve the model physics will be indispensable to reliably estimate sub-hourly rain intensities and KE for direct hydrological applications.

How to cite: Van Weverberg, K., Ghilain, N., Goudenhoofdt, E., Barbier, M., Kostinen, E., Doutreloup, S., Van Schaeybroeck, B., Frankl, A., and Field, P.: Sensitivity of simulated rain intensity and kineticenergy to aerosols and warm-rain microphysicsduring the extreme event of July 2021 in Belgium, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-3245, https://doi.org/10.5194/egusphere-egu24-3245, 2024.

Models like the conceptual hydrological model COSERO and the physically-based mountain surface process model Alpine3D are highly sensitive to meteorological inputs, especially precipitation. Gridded precipitation data sets usually originate from spatially interpolated weather station data, which are not corrected for precipitation undercatch. The term precipitation undercatch describes the deviation of measured precipitation in rain gauges to the actual amount in a given area due to several factors like instrument design or effects of splash, evaporation and especially wind. Specifically solid precipitation is prone to wind drag. Because of these effects, models or model chains fail to simulate observations for discharge, reservoir inflow, snow and ice melt as well as glacier mass balance due to the lack of realistic precipitation input into the system in high-alpine regions. However, correcting the undercatch directly within the gridded data set leads to an overestimation of precipitation, which has two main reasons: First, undercatch correction functions are not derived for alpine temperatures and wind speeds. Second, stations at lower elevations, where the undercatch is comparatively small, are usually over-represented in gridded data sets.

Therefore, we composed a method to perform a precipitation undercatch correction in high-alpine areas by using a gridded precipitation data set and quality-controlled, representative station data in the vicinity of snow-dominated and glacierized catchments as well as their altitude and exposure to generate spatial undercatch correction fields for three selected catchments in Austria (Maltatal, Zillertal and Vernagtferner) on a monthly basis. These correction factors are a function of elevation and the month and result from a stepwise linear interpolation with elevation, whereas the highest factors are obtained in the winter months due to low temperatures. Using the topography and averaging over whole catchments, the highest (lowest) correction factors are obtained in February (August), ranging from 2.16 to 1.04, depending on the catchment and season.

The meteorological data (with and without the undercatch corrected precipitation) was used as an input for a coupled snow-glacier-discharge simulation with the models COSERO and Alpine3D on the selected catchments. The output was validated against reservoir inflow, observed glacier mass balances and satellite derived snow depth maps. With the undercatch corrected precipitation, the models perform substantially better in simulating observations for glacier mass balance as well as reservoir inflow.

Acknowledgements: We thank VERBUND AG for fruitful discussions and providing us with data.

How to cite: Maier, P., Ehrendorfer, C., Lücking, S., Lehner, F., Koch, F., Herrnegger, M., and Formayer, H.: On the improvement of runoff and glacier mass balance modelling by performing an undercatch correction on gridded precipitation data sets based on independent station data, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-2047, https://doi.org/10.5194/egusphere-egu24-2047, 2024.

The estimation of extreme rainfall based on short records is of considerable interest, above all in the context of rapidly changing rainfall regimes. Regionalization techniques, by trading space for time, allow us to partially overcome the lack of long observational records. The recently introduced Metastatistical Extreme Value Distribution (MEVD), a non-asymptotic extreme-value model, accounting for all observed rainfall events to infer the probability distribution of annual maxima, also contributed towards improving our ability to determine large quantiles based on short observational time series. Here we combine established regionalization techniques, aggregating data from multiple adjacent stations complying with set homogeneity criteria, with MEVD-based methodologies to explore how their joint use may further reduce the predictive uncertainty associated with the estimates of the probability of large events. In this work, we use precipitation data sets from a selection of worldwide regional station networks (Europe, USA, Middle East, and Asia) deployed in a wide range of elevations and different rainfall regimes. The temporal data resolution varies according to country ranging from sub-daily to daily scales. We analyze different event durations, between 5 minutes and 24 hours for the sub-daily scale, 1 day and 2 days for the daily one, and we implement a cross-validation procedure to evaluate predictive uncertainty. To evaluate possible improvements with respect to regionalization techniques based on traditional extreme value theory, such as the Generalized Extreme Value (GEV) distribution, we comparatively apply them and the proposed MEVD-based regionalization approach. The results show the benefits arising from the regionalization technique, which enhances the robustness of the models by increasing the consistency of the observed data population within the stations of the same cluster, particularly in the lowlands, where homogeneous regions can be more trivially identified. The proposed regionalization approach based on the metastatistic distribution brings a significant reduction of the estimation uncertainty for very high ratios between the forecasting return period value and the length of the calibration sample when compared to traditional methods.

How to cite: Devò, P., Caruso, M. F., Borga, M., and Marani, M.: A regionalized framework for the Metastatistical Extreme Value Distribution applied to daily and sub-daily rainfall, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-17329, https://doi.org/10.5194/egusphere-egu24-17329, 2024.

In mountains, topography-atmosphere interactions generate orographic effects which make windward slopes usually wetter than leeward ones, and highlands wetter than lowlands. The transitions between wet and dry areas can occur within few kilometers, which creates strong horizontal gradients of rainfall statistics such as frequency of occurrence, daily mean accumulation, or extreme intensities. This spatial variability of rainfall statistics breaks the hypothesis of stationarity on which rely most geostatistical models that are used for the spatial analysis of rainfall data. Using stationary models to process non-stationary data can lead to a degraded performance in spatial prediction (e.g., mapping rainfall by interpolation of sparse rain gauge observations) and to unrealistic rainfall features in simulations (e.g., emulation of synthetic rain fields using a stochastic rainfall generator). 
 
To overcome these limitations, we present in this work a fully non-stationary trans-Gaussian geostatistical model dedicated to the spatial analysis of daily rainfall over complex topography. This model allows not only for a non-stationary marginal distribution of daily rainfall accounting for rainfall intermittency and non-Gaussian intensity, but also for a non-stationary covariance structure of Matérn type that models the spatial dependencies.

The model is tested for the Island of Hawai‘i (State of Hawaii, USA) where rainfall gradients are amongst the strongest on Earth and can reach 1000 mm.year^-1/km. To make our model operable in practice, we designed a procedure to infer model parameters from rain gauge observations that are freely available in near-real-time on the Hawai‘i Climate Data Portal. Model assessment demonstrates good skills at reproducing the spatial variability of daily rainfall occurrence, intensity distribution and spatial dependencies.

How to cite: Benoit, L., Lucas, M., Allard, D., and Giambelluca, T.: Non-stationary geostatistical modeling of daily rainfall over complex topography, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-7137, https://doi.org/10.5194/egusphere-egu24-7137, 2024.

Climate change is changing the intensity and frequency of extreme precipitation. Understanding the impact of climate change on extreme precipitation quantiles is fundamental for managing flood risk and taking adaptation measures. Convection-Permitting Models (CPM), run at spatial resolutions for which deep convection is resolved (≤ 4 km), have been demonstrated to be more accurate than Regional Climate Models (RCM, ~10 km resolution) in describing the intensity of extremely short-duration events.

This study uses the projections of a CPM to evaluate quantiles of precipitation extremes at the national scale (Italy) with a high spatiotemporal resolution. Indeed, VHR-PRO_IT, a recent downscale product of the CMCC model at a convection-permitting scale of 2.2 km, with 1h temporal resolution, is used as a dataset. So far, this is the only CPM projection that covers the entire Italy in both emission scenarios (RCP 4.5 and RCP 8.5) and for a temporal coverage of 90 years (1981-2070).

A non-stationary implementation of the Simplified Metastatistical Extreme Value (SMEV) non-asymptotic approach is used to evaluate continuous changes in precipitation quantiles for different durations (1h, 3h, 6h, 12h and 24h) over the period 1981-2070 (1981-2005 historical + 2006-2070 emission scenarios). We adopt a two-parameter Weibull distribution to model the marginal distribution of the ordinary precipitation events. Three different models are compared: i) a stationary SMEV, with the two parameters constant over the entire time series; ii) a non-stationary model in which the higher-order parameter is kept constant; iii) a fully non-stationary model in which both parameters are allowed to change linearly in time.

The results show a clear geographical organization of the projected changes, with both increases and decreases in precipitation quantiles depending on the zone, the emission scenario, the precipitation duration and the return period of interest. The non-asymptotic approach allows us to discuss the results in terms of dynamic and thermodynamic drivers.

The research is carried out within the RETURN – multi-Risk sciEnce for resilienT comUnities undeR a changiNg climate Extended Partnership and received funding from the  European Union Next-GenerationEU (National Recovery and Resilience Plan – NRRP, Mission 4, Component 2, Investment 1.3 – D.D. 1243 2/8/2022, PE0000005).

How to cite: Lompi, M., Marra, F., Dallan, E., Deidda, R., Caporali, E., and Borga, M.: Non-stationary frequency analysis of extreme precipitation over Italy using projections from a Convection Permitting Model, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-10004, https://doi.org/10.5194/egusphere-egu24-10004, 2024.

Daily rainfall records are the most common form of rainfall information. These records are the ones normally used to characterize the extremes of rainfall. However, in many situations, sub-daily information is required, normally to characterize the extreme response of small watersheds. Different methods exist to extrapolate the daily information to smaller time scales -normally, hourly time scales-, which tend to be based on a limited number of finer than daily observations.

In this work, we will deal with two common downscaling problems that the hydrologist faces: transforming daily rainfall observations into estimates of sub-daily rainfall statistics and incorporating climate change information into these estimates. Although generally, these two procedures are different, both conceptually and mechanically, we will combine stochastic generators and machine learning to create a unified framework where both problems are connected and solved in a similar manner.

We will use NEOPRENE (Diez-Sierra et al., 2023), a Python-based open source library that implement the Nyeman-Scott, or Cox and Isham, stochastic model of rainfall (Cox & Isham, 1988) to characterize the rainfall process, and random forests to relate daily and hourly rainfall statistics (del Jesus & Diez-Sierra, 2023). The model assumes a geometric description of the rainfall process, that allows to decompose observed time series and reproduce several statistics at different levels of aggregation.

We will also demonstrate how downscaling can be carried out to generate plausible hourly rainfall distributions from daily ones, and how this process serves to characterize the uncertainties of the estimates.

Cox, D. R. & Isham, V., 1988. A Simple Spatial-Temporal Model of Rainfall. Proceedings of the royal society a: Mathematical, physical and engineering sciences, 415 ​(1849), 317–328.

Diez-Sierra, J., Navas, S. & Jesus, M. del., 2023. NEOPRENE v1.0.1: A Python library for generating spatial rainfall based on the NeymanScott process. Geoscientific model development, 16 (17), 5035–5048.

Jesus, M. del & Diez-Sierra, J., 2023. Climate change effects on sub-daily precipitation in Spain. Hydrological sciences journal, 68 (8), 1065–1077.

How to cite: del Jesus, M., Diez-Sierra, J., and Navas, S.: Rainfall downscaling using stochastic generators and machine learning, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-6450, https://doi.org/10.5194/egusphere-egu24-6450, 2024.

Precipitation events are one of the most crucial processes governing the water cycle and therefore acts as a major input in majority of water resource studies. Furthermore, the modern era is witnessing an unprecedented increase in the frequency of extreme events highlighting the importance of understanding and predicting the precipitation events. Current precipitation prediction methods utilise complex physics-based models that require large number of input parameters as well as powerful computational facilities, making precipitation prediction a complex task. This scenario has not improved much over the years as advancements are often limited to either improving the model’s physics or input data quality. Hourly precipitation prediction is even more challenging due to increasing complexity and non-linearity with decreasing scale and therefore studies on understanding hourly precipitation is limited. Recent trends have shown a shift towards utilizing deep learning models in weather prediction owing to the ability of neural networks to capture complex patterns leading to high accuracy predictions. The current research introduces a Long Short Term Memory (LSTM) neural network adept at forecasting fine-scale hourly precipitation patterns up to two hours ahead, a critical development for real-time rain predictions. The Bi-LSTM's architecture, with its forward and backward processing capabilities, is particularly suited to capture the dynamic temporal relationships among the limited meteorological variables helping in effective precipitation prediction. Granger Causality analysis is done to capture relevant information for improving model performance. The model's performance is evaluated on its ability to accurately forecast weather conditions by learning from the historical inter-variable influences that were clearly detailed in the causality diagram. The findings from this study and the interlinks observed is expected to enhance our understanding of variable impact and improve the predictive power of precipitation models for future weather forecasting

How to cite: Sreedhar, R., Sunil, A., and L Murthy, R.: Hourly Precipitation Prediction: Integrating Long Short-Term Memory (LSTM) Neural Networks with Granger Causality, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-10291, https://doi.org/10.5194/egusphere-egu24-10291, 2024.

The accurate estimation of precipitation (P) at a high spatio-temporal resolution is vital in various applications such as climatic modelling, water resources management, drought and flood assessment, and climate change adaptation, among others. However, an adequate representation of P products in space and time remains challenging, particularly over regions with sparse or non-existent gauge-reference observations. Jordan, ranked among the top four driest countries in the world, urgently requires reliable P in good spatio-temporal resolutions to enable decision-makers and researchers to manage water resources effectively. In this study, seven state-of-the-art P products (MSWEPv2.8, ERA5, CHIRPSv2, CMORPHv1, PERSIANN-CDR, IMERG-FR, and ERA5 LAND) were evaluated against 124 gauge stations over the region using point-to-pixel evaluation at daily, monthly, annual, and seasonal temporal scales. Kling-Gupta efficiency as a continuous index with its three components (temporal dynamic r, bias ratio β, and variability ratio) was used to identify the systematic errors and uncertainties of the P products. Additionally, four categorical indices (probability of detection (POD), frequency bias (fbias), false alarm ratio (FAR) and the Critical success index (CSI) were used to assess the ability of the P products to capture different P intensities. The best performing daily scale P products were then resampled to a finer resolution of 0.05° (5 km) and merged with the gauge station observations to improve the representation of P over the region using two distinct approaches: i) machine learning approach, the Random Forest based MErging Procedure (RF-MEP), and ii) geostatistical approach, Kriging with External Drift (KED). We applied RF-MEP and KED over Jordan for the period 2001 - 2017 with a focus on its arid and climatic conditions; thus, we also applied the models to each climatic zone using daily observations of 80% of the gauge stations as a training dataset, and 20% were used for the verification of the merged P products. The results revealed that MSWEPv2.8 emerged as the top-performing P product. For this reason, and already being a merged dataset, MSWEPv2.8 was used as a benchmark in evaluating the merged products. For RF-MEP, The remaining datasets, excluding ERA5-LAMD and IMERGE-FR due to their poor performance, were merged with gauge observations, while KED was merged with the second-top performance product, ERA5. Both merged products demonstrated significant improvements in P patterns, linear correlation, bias, and variability at different temporal scales and in capturing different precipitation intensities. RF-MEP showed superior performance across Jordan compared to KED. However, KED outperformed RF-MEP in elevated terrains. Subsequently, a practical application of the newly merged P products was tested through simple drought assessment, using the Standardized Precipitation Index (SPI) specifically, SPI-12. The outcomes demonstrated that RF-MEP showed promising results in the detection of extreme long-term dry spells, highlighting its ability for practical application in drought assessment.

Keywords: Precipitation; Gridded Precipitation products; Point to pixel evaluation; KGE; Categorical indices; Merging; RF-MEP; KED; SPI; Jordan

How to cite: Al-Saeedi, B. A., M. Baez-Villanueva, O., and Ribbe, L.: An optimized representation of precipitation in Jordan: Merging gridded precipitation products and ground-based measurements using machine learning and geostatistical approaches, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-11510, https://doi.org/10.5194/egusphere-egu24-11510, 2024.

Southern South America (SSA) covers the extratropical part of South America (20–60°S) and presents a wide variety of climates. To the west of the Andes mountain range, annual precipitation increases southward from very dry conditions along the Atacama Desert in northern Chile to more than 3000 mm in the south. Conversely, east of the Andes, it increases from the Argentinian Patagonia in the south towards southeastern South America (southern Brazil, northeastern Argentina and Uruguay) where severe thunderstorm environments are typical. Global climate models project that the observed negative (positive) trends in precipitation over the subtropical central Andes (Southeastern South America) are expected to be more intense causing concern about water availability, ecosystems and socio-economic activities. However, the regional-to-local information that can be obtained by downscaling over GCMs outputs and needed for adaptation and mitigation policies, is still scarce over SSA. Unlike other parts of the world, limited studies analyzing the statistical downscaling (ESD) potential to simulate daily precipitation are available over the region and deep learning-based models have not been tested for downscaling daily precipitation over the region up to now.

In this context, this work presents a comprehensive assessment of Convolutional Neural Networks (CNNs) to downscale daily precipitation at a continental-scale, building on the validation framework of the European project VALUE. To this end, we conduct a sensitivity analysis to the domain size as well as to the selection of the loss function on the modeling of precipitation in both present and future climates. Overall, the CNNs show skilful performance in modeling daily precipitation characteristics, including the extremes, over the different climatic regions of SSA. Nevertheless, we find the selection of the loss function to be a source of uncertainty over the arid regions of northern Chile and northwestern Argentina for both present and future climates by projecting different climate change signals. Regarding the domain size, the CNNs show to be effective in selecting informative predictors and their area of influence demonstrating their self-learning skill and their efficiency to be applied on a continental scale. These results encourage the construction of ensembles of deep learning models based on different loss functions in SSA to account for this type of uncertainty in the modeling of precipitation, especially in future climates.

How to cite: Bettolli, M. L., Baño-Medina, J., Olmo, M., and Balmaceda-Huarte, R.: Modeling local precipitation in Southern South America using Deep Learning: A sensitivity analysis on the choice of input features and loss function in the climate change signal, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-786, https://doi.org/10.5194/egusphere-egu24-786, 2024.

We introduce a new version of the gridded near real-time Multi-Source Weighted-Ensemble Precipitation (MSWEP) product, developed to address the urgent need for accurate precipitation (P) data in the face of escalating climate change challenges. The product has an hourly 0.1° resolution spanning 1979 to the present, and is continuously updated, with a latency of approximately one hour. The development process involves two stages. Firstly, baseline P fields are generated from multiple satellite and (re)analysis P products, along with several static P-related variables, using random forest models trained on 3-hourly and daily P observations from gauges across the globe (n=17,322). Subsequently, these baseline P fields are locally corrected using available daily P observations, employing a procedure that accounts for the reporting times of gauges. To assess the accuracy of the product, we conducted the most comprehensive global evaluation of P products to date, using daily observations from independent P gauges as a reference (n=15,184). The new P product (prior to gauge corrections) outperformed all 18 other evaluated products, attaining a mean daily Kling-Gupta Efficiency (KGE) value of 0.65. In contrast, widely used products such as CHIRP, ERA5, GSMaP, and IMERG achieved mean KGE values of 0.31, 0.57, 0.37, and 0.40, respectively. Furthermore, our P product consistently ranked first or second across various metrics, including correlation, overall bias, peak bias, wet days bias, and critical success index. Notably, the new product also outperformed several gauge-based products like CHIRPS and CPC Unified, which had mean KGE values of 0.37 and 0.54, respectively. Set for release in late 2024, we anticipate that the new product will be useful for climate research, water resource assessment, and flood management, among numerous other potential applications.

How to cite: Beck, H., Wang, X., and Alharbi, R.: Hourly 0.1° Gridded Near Real-Time Precipitation (1979–Present) via Machine Learning Fusion of Satellite, Model, and Gauge Data, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-19191, https://doi.org/10.5194/egusphere-egu24-19191, 2024.