Sea level modelling contribution to flooding risk reduction planning in coastal urban areas

Papadopoulos, Nestoras; Gikas, Vassilis

dc.contributor.author	Papadopoulos, Nestoras
dc.contributor.author	Gikas, Vassilis
dc.date.accessioned	2026-07-20T17:01:27Z
dc.date.available	2026-07-20T17:01:27Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/65547
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.33241
dc.rights	Default License
dc.subject	astronomical tide	el
dc.subject	storm surges	el
dc.subject	Deep Learning	el
dc.subject	LSTM	el
dc.subject	timeseries forecast	el
dc.title	Sea level modelling contribution to flooding risk reduction planning in coastal urban areas	en
heal.type	conferenceItem
heal.classification	Oceanography	en
heal.classification	Geodesy	el
heal.contributorName	Papadopoulos, Nestoras
heal.contributorName	Gikas, Vassilis
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2025-06-18
heal.bibliographicCitation	Papadopoulos N, Gikas V. Sea Level modelling Contribution to Flooding Risk Reduction Planning in Coastal Urban Areas. Joint Conference with UNECE WPLA & REM, FIG Com 3 & FIG Com 9, EgoS, World Bank. Athens – Santorini, Greece, 2025.	el
heal.abstract	Coastal urban regions are susceptible to flooding events led by the complex interplay of extreme sea level events resulting from meteorological disturbances and high astronomical tides, especially when these phenomena coincide. This risk is amplified further by the increased challenge of long-term sea level rise as a result of climate change. To mitigate such escalating risks, the development and testing of interdisciplinary robust risk reduction plans are of paramount importance. Such plans necessitate the integration of meticulously curated datasets, sophisticated predictive models capable of forecasting future sea level values, geospatial information leading to reliably defined sea level thresholds that can effectively trigger early warning systems to allow for proactive measures for communities and infrastructure. This study contributes to this need through the development of a framework proposal for sea level prediction based solely on sea level recordings and meteorological information. Analyses have focused on the strategically important harbor of Piraeus, one of the largest and busiest ports in the Mediterranean Sea, located near the metropolitan city of Athens. The research leverages a database comprising tide gauge recordings and meteorological variables, spanning the period from 1999 to 2019. The raw data recordings have been under a thorough quality control process whereas statistical analyses have been applied to determine tidal datums of the study period. Building upon this robust data foundation, we developed an analytical model designed to predict future sea level values given a meteorological forecast. This model adopts a component-based approach, wherein the total sea level estimate is decomposed into its constituent parts; namely, the predictable astronomical tide, the meteorologically induced storm surge and a residual function of time, which accounts for longer-term trends and other unmodeled influences. Furthermore, recognizing the increasing tolerance of data-driven computational approaches, this study explores AI-based, deep learning methodologies through the implementation and exhaustive evaluation of alternative Long Short-Term Memory (LSTM) recurrent neural network architectures. The proposed LSTM models were specifically designed to predict high (hourly) resolution sea level estimates at scalable forecasting times ranging from 7 to 28 consecutive days. To account for different scenarios of data availability and to maximize predictive capability, multiple LSTM networks were constructed and trained using a diverse array of input variables. We explored and cross-compared the performance of various LSTM network configurations including univariate, multivariate and multivariate-to-univariate models. Both static and dynamic forecasting strategies were investigated. The results obtained proved to be highly promising, demonstrating a strong fit between the forecasted sea level values and the corresponding data recording. Quantitative evaluation of the model performance revealed Root Mean Squared Error (RMSE) values varying within a range of 1 to 6 cm, depending on specific LSTM neural network architecture and the input variables adopted. This level of accuracy underscores the potential use of these hybrid modelling approaches for accurate sea level forecasting contributing to coastal flooding studies. The findings of this research offer the basis of a robust and adaptable methodology that can be readily applied to other coastal urban environments facing similar challenges, contributing to improved preparedness and resilience against the growing threat of coastal inundation in a changing climate.	en
heal.fullTextAvailability	false
heal.conferenceName	Joint UNECE WPLA & REM, FIG Com3 & Com9, EGoS,WBConference, 18-22 June 2025, Athens & Santorini	el
heal.conferenceItemType	tutorial