Assessing future wave climate in the Ebro Delta using SWAN nested modelling and CMIP6 wind projections

Uhinak 2024 Revista de Investigación Marina, 2024, 30(1)| 38 Keywords: wave projections, CMIP6, SWAN, Ebro Delta, nested modelling Introduction Coastal communities and ecosystems around the world are currently facing serious threats including flooding and erosion, which will be exacerbated by climate change in the coming century. While many studies focus on sea level rise, altered wind-wave climates will also play an important role in the evolution of future coastal hazards. Projected changes in wave patterns vary by location, with some areas expecting more frequent and intense wave events and others expecting calmer wave climates. As a result, studies at local or regional scales focusing on specific vulnerable coastal areas are extremely important. Intense wave storms including Storm Gloria in 2020 have caused serious flooding and erosion in the Ebro Delta in southern Catalonia. Understanding the future evolution of the wind-wave climate near the Ebro Delta is, therefore, critical for planning future mitigation and adaptation strategies to protect the area’s population, infrastructure, and ecosystems. To project future wave climate, wind-wave models such as the Simulating Waves Nearshore (SWAN) model can be forced with wind projections from global climate models (GCMs). As GCMs are improved, additional data is collected, and new climate change scenarios are analyzed, incorporating the most recent wind projections into wave climate studies is important. Currently, research using projections from CMIP6, the latest phase of the Coupled Model Intercomparison Project, is lacking, particularly in the Western Mediterranean. In this study, wave projections were generated in three domains— the Western Mediterranean, the Catalan Coast, and the Ebro Delta—using SWAN nested models forced with wind projections for two climate scenarios, SSP2-4.5 and SSP5-8.5, from 20 different GCMs developed under CMIP6. The findings of this research provide insight into future changes in wave climate in the Western Mediterranean, with a particular focus on the Ebro Delta. Materials and methods Before generating wind projections, a SWAN nested model first needed to be configured and validated in the study area. In addition, to determine which models were most likely to produce reliable wave projections required assessing and comparing the performance of SWAN models forced with historical outputs of the 20 different GCMs. This study, therefore, consisted of three main parts: model validation, assessment of historical model performance, and generation of wind projections. A nested SWAN model composed of three domains with increasing computational grid resolution was used in this study (Figure 1). Simulated wave conditions in larger domains served as boundary conditions for smaller domains. Initial configuration of model parameters was based on previous work off the Catalan Coast (Pallares et al., 2014)fetch-limited conditions. This applies particularly to the wave period, in semi-enclosed domains with highly variable wind patterns as along the Catalan coast. The wave model SWAN version 40.91A is used here in three nested grids covering all the North-western Mediterranean Sea with a grid resolution from 9 to 1km, forced with high resolution wind patterns from BSC (Barcelona Supercomputing Center. Figure 1. Domain A (Western Mediterranean), Domain B (Catalan Coast), and Domain C (Ebro Delta) of the SWAN nested model. The validation model was forced with ERA5 wind reanalysis data and run for a historical period from 1999-2014. Model outputs were compared with measured data at seven wave buoys located in the Western Mediterranean considering three integral wave parameters: significant wave height (HS), mean period (Tm), and mean direction (θm). As discussed later in the Assessing future wave climate in the Ebro Delta using SWAN nested modelling and CMIP6 wind projections Houghton, Tobias1,2, Olmo, Matías2, Mestres, Marc1,3, Cos, Pep2, Soret, Albert2, Espino, Manuel1,3 1 Maritime Engineering Laboratory, Universitat Politècnica de Catalunya, Carrer de Jordi Girona 1–3, Campus Nord, 08034, Barcelona, Spain 2 Barcelona Supercomputing Center, Plaça d’Eusebi Güell, 1-3, Les Corts, 08034, Barcelona, Spain 3 International Centre for Coastal Resources Research, Carrer de Jordi Girona 1–3, Campus Nord, 08034, Barcelona, Spain E-mail contact: [email protected] Uhinak 2024 39 | Revista de Investigación Marina, 2024, 30(1) Results and Discussion section, validation model results were acceptable, and the initial model configuration was used for generating wave projections. Following model validation, the performance of models forced with historical wind outputs for the period 1985-2014 from the 20 GCMs was assessed. Historical models were ranked based on performance across all buoys considering HS, Tm, and θm. Model performance was assessed by comparing average daily outputs from the historical models and validation model using the Distance between Indices of Simulation and Observation (DISO; Equation 1) index for HS and Tm and by comparing overall distribution of θm between the historical and validation models using the Kuiper test (Equation 2). Equation 1 DISOij= √((Rij-1)2 + NMBij 2+NRMSEij 2) Where Rij, NMBij, and NRMSEij are the correlation coefficient, normalized mean bias, and normalized root mean squared error for model i at buoy j (Hu et al., 2019). Equation 2 Vij =supθ (Fs,ij(θm) - FM,ij(θm)) + supθ (FM,ij(θm) - Fs,ij(θm)) Where Vij is the Kuiper statistic, Fs,ij(θm) is the empirical cumulative distribution function (CDF) of simulated θm, FM,ij(θm) is the empirical CDF of measured θm, and the supremum function denoted by sup determines the upper boundary of a set of numbers (Jammalamadaka & Sengupta, 2001). After assessing historical model performance, wind projections were generated between 2070-2099 using GCM data from two climate change scenarios, SSP2-4.5 and SSP5- 8.5. Wave projections were analyzed considering both mean changes in wave climate, and changes in extreme HS for 1-yr, 10-yr, and 100-yr return periods. Extreme wave heights were determined by fitting a generalized Pareto distribution (Equation 3) to a set of extreme events identified using a peaks above threshold approach for each dataset. Equation 3 F(x)=1-[1+ξ((x-u)/σ)](-1/ξ) where ξ is a shape parameter, σ is a scale parameter, and u is the sample mean (Wilks, 2019). Results Outputs from the validation model showed high correlation with measured data, with correlation coefficients above 0.9 for HS at almost all buoys. However, persistent negative biases were also obsereved for both HS and Tm, with normalized mean bias values of -26% and -2.3% respectively at the Tarragona buoy, closest to the Ebro Delta. While model results showed significant levels of error, they were comparable to a previous study off the Catalan Coast (Pallares et al., 2014)fetch-limited conditions. This applies particularly to the wave period, in semi-enclosed domains with highly variable wind patterns as along the Catalan coast. The wave model SWAN version 40.91A is used here in three nested grids covering all the North-western Mediterranean Sea with a grid resolution from 9 to 1km, forced with high resolution wind patterns from BSC (Barcelona Supercomputing Center. As a result, the initial model configuration was used when assessing historical model performance and generating wave projections. Historical models generally had even greater negative biases for HS and Tm when compared with the validation model. However, historical models were able to simulate seasonal trends in HS and Tm well, with high HS and Tm values in the winter and low values in the summer. Most models failed almost completely in simulating past distribution of θm, with the exception of a few models with relatively high resolution. In general, higher resolution models performed the best with the four top-performing models also being the four models with the highest resolution wind data. Wind projections indicated reductions in mean HS and Tm for the SSP2-4.5 scenario with even greater reductions in mean HS and Tm for SSP5-8.5 across all but three models. The only models showing increases in wave height in future scenarios also performed poorly in historical scenarios. Results showing reductions in HS in the Western Mediterranean are in agreement with earlier studies using wind projections based on previous CMIP phases (Aznar Lecocq et al., 2016; Ramírez Pérez et al., 2019). Figure 2. Mean change in HS between historical and SSP2-4.5 and SSP5- 8.5 scenarios averaged across the four top-performing models. When considering extreme HS, model results showed significantly more disagreement. However, for the four topperforming models, results still indicated decreases in extreme HS for all three return periods along the Catalan Coast. At buoys located in other parts of the Western Mediterranean, model results indicated smaller decreases or even increases in extreme HS for some return periods. Changes in extreme HS generally resemble results from Ramírez Pérez et al. (2019) who found large decreases in 99th percentile HS off the Catalan Coast when forcing windwave models with CMIP5 based wind projections, but smaller reductions in HS in other locations in the Western Mediterranean. Uhinak 2024 Revista de Investigación Marina, 2024, 30(1)| 40 Conclusions This study provides valuable insight into future wave climate in the Western Mediterranean, Catalan Coast, and Ebro Delta using a large ensemble of CMIP6 wind projections. Results indicate decreases in both mean and extreme HS off the Catalan Coast. While results indicating reductions in mean HS are consistent across almost all models, more variation is observed when considering extreme HS, with several models showing increases in extreme HS near the Tarragona buoy. Wind data resolution was low for both the validation model (0.25°) and the historical models (0.5-2.82°). Results from the historical model evaluation indicate the importance of wind data resolution with the models forced with the highest resolution wind data performing best. Dynamic or statistical downscaling of GCM data could significantly improve wave projections. In addition, results from both the validation and historical models exhibit significant negative biases that would likely persist even if higher resolution data were used. Bias correction should be applied to model outputs in the future to improve wave projections. Acknowledgements This research has received funding from the European Union’s Horizon 2020 research and innovation programme under the REST-COAST project (grant agreement No 101037097). References Aznar Lecocq, R., Padorno Prieto, M. E., Pérez Gómez, B., Gómez Lahoz, M., García Sotillo, M., Álvarez Fanjul, E., Pérez Jordán, G., Marcos Moreno, M. I., Martínez Asensio, A., Llasses, J., Gomis, D., Sánchez Perrino, J. C., Rodríguez González, J. M., Rodríguez Camino, E., Somot, S., Sevault, F., Adloff, F., & Compte, A. (2016). Vulnerabilidad de los puertos españoles ante el cambio climático. Vol. 1: Tendencias de variables físicas oceánicas y atmosféricas durante las últimas décadas y proyecciones para el siglo XXI. https://repositorio.aemet. es/handle/20.500.11765/8809 Hu, Z., Chen, X., Zhou, Q., Chen, D., & Li, J. (2019). DISO: A rethink of Taylor diagram. International Journal of Climatology, 39(5), 2825– 2832. https://doi.org/10.1002/joc.5972 Jammalamadaka, S. R., & Sengupta, A. (2001). Topics in Circular Statistics (Vol. 5). World Scientific. Pallares, E., Sánchez-Arcilla, A., & Espino, M. (2014). Wave energy balance in wave models (SWAN) for semi-enclosed domains– Application to the Catalan coast. Continental Shelf Research, 87, 41–53. https://doi.org/10.1016/j.csr.2014.03.008 Ramírez Pérez, M., Menéndez García, M., Camus Braña, P., & Losada Rodríguez, I. (2019). Tarea 2: Proyecciones de Alta Resolución de Variables Marinas en la Costa Española (Elaboración de La Metodología y Bases de Datos Para La Proyección de Impactos de Cambio Climático a Lo Largo de La Costa Española). Ministerio para la Transición Ecológica y el Reto Demográfico. https://www. miteco.gob.es/content/dam/miteco/es/costas/temas/proteccion-costa/ tarea_2_informe_pima_adapta_mapama_tcm30-498855.pdf Wilks, D. S. (Ed.). (2019). Front Matter. In Statistical Methods in the Atmospheric Sciences (Fourth Edition) (pp. i–ii). Elsevier. https://doi. org/10.1016/B978-0-12-815823-4.09987-9 Figure 3. Change in extreme HS between historical and SSP2-4.5 and SSP5-8.5 scenarios averaged across the four top-performing models. The heatmap to the left shows results for all 20 models at the Tarragona buoy.