SEA LEVEL PROJECTIONS WITH MACHINE LEARNING USING ALTIMETRY AND CLIMATE MODEL ENSEMBLES - Abstract

Aug 4, 2026

Upload by Patrick

Satellite altimeter observations retrieved since 1993 show that global mean sea level is rising at an unprecedented rate (3.4 mm/year). With nearly three decades of records, the study investigates how anthropogenic climate-change signals—greenhouse gases, aerosols, and biomass burning—contribute to sea level rise. A machine-learning framework combines satellite observations and climate model simulations to produce 2-degree spatial projections up to 30 years ahead. Fully connected neural networks fuse climate model hindcasts (1993–2019) and then are applied to future model projections. Spatial clustering is introduced to improve predictive performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages6

Reading Time15 min

AccessFree to download

Abstract

Introduction

arXiv:2308.02460v1 [[physics. ao-ph](physics. ao-ph)] 2 Aug 2023 SEA LEVEL PROJECTIONS WITH MACHINE LEARNING USING ALTIMETRY AND CLIMATE MODEL ENSEMBLES Saumya Sinha˚ University of Colorado Boulder, CO, USA John Fasullo National Center for Atmospheric Research Boulder, CO, USA R. Steven Nerem University of Colorado Boulder, CO, USA Claire Monteleoni University of Colorado Boulder, CO, USA ABSTRACT Satellite altimeter observations retrieved since 1993 show that the global mean sea level is rising at an unprecedented rate (3.4mm/year) . With almost three decades of observations, we can now investigate the contributions of anthropogenic climate-change signals such as greenhouse gases, aerosols, and biomass burning in this rising sea level. We use machine learning (ML) to investigate future patterns of sea level change. To understand the extent of contributions from the climate-change signals, and to help in forecasting sea level change in the future, we turn to climate model simulations. This work presents a machine learning framework that exploits both satellite observations and climate model simulations to generate sea level rise projections at a 2-degree resolution spatial grid, 30 years into the future. We train fully connected neural networks (FCNNs) to predict altimeter values through a non-linear fusion of the climate model hindcasts (for 1993-2019) . The learned FCNNs are then applied to future climate model projections to predict future sea level patterns. We propose segmenting our spatial dataset into meaningful clusters and show that clustering helps to improve predictions of our ML model. 1 INTRODUCTION With melting ice sheets and the growing warmth of the ocean water, the global mean sea level is rising at an extraordinary rate and is accelerating (0 .08mm/year2 ) [17] . While, on average, the sea level has risen 10 cm over the last 30 years, there is a considerable variation in the regional rates of the sea level change [11] . These regional patterns are richer in information and can be very useful in examining the impact of climate-driven factors including greenhouse gases, industrial aerosols, and biomass burning in the sea level change. Studies in [5; 6; 7] have found the regional variations to be linked to aerosols and greenhouse gas emissions. Identifying how regional trends will evolve in the future is also beneficial for socioeconomic planning. Our work investigates the sea level at a 2-degree or 180 x 90 (longitude x latitude) spatial resolution. Now that there are almost three decades of satellite altimeter records or observations 1 , we want to explore the rising rate of the sea level further and investigate how much of this rise can be attributed to climate-change signals. To do this, we turn to the climate models in order to provide a better understanding of the altimeter data and to help in forecasting future sea level change by learning more about the extent of the causal contributions from such factors. A climate model uses mathematical equations to simulate complex physical and chemical processes of Earth systems such asthe atmosphere, land, ocean, ice, and solar energy [8] . A recent study by Fasullo and Nerem (2018) with two climate model large ensembles showed that the forced responses of greenhouse gas and ˚ Corresponding author: Saumya Sinha, [saumya.sinha@colorado.edu](saumya.sinha@colorado.edu) 1We obtain our altimeter data for the period 1993-2019 from [https://www.aviso.altimetry.fr/](https://www.aviso.altimetry.fr/)[ ](https://www.aviso.altimetry.fr/)en/data/products/ocean-indicators-products/mean-sea-level .html aerosols have begun to emerge in the altimeter data patterns. This motivates us to include climate models in designing our machine learning (ML) pipeline. Our goal is to generate sea level projections 30 years into the future and at a 2-degree spatial resolution utilizing the altimeter data as well as climate model simulations. Some past works have

FAQs

What data sources are used to build the sea level projections?

The method uses satellite altimeter observations spanning 1993–2019 and climate model simulations. Climate model hindcasts train the machine-learning model and future projections are used for prediction.

How does the machine learning model generate future sea level patterns?

Fully connected neural networks learn a non-linear fusion of climate model hindcasts and then predict altimeter values for the learned patterns. The trained networks are applied to future climate model projections to forecast sea level changes 30 years ahead at 2-degree resolution.

Why is clustering applied to the spatial dataset?

The study segments the spatial data into meaningful clusters and shows that clustering improves prediction quality for the machine-learning model, helping it capture regionally relevant behavior.

Aug 4, 2026

Upload by Patrick