FIG Peer Review Journal

Exploring Factors Inducing Vertical Surface Deformation in Ifewara Zungeru fault zone, Southwest Nigeria Through Machine Learning Approach (13049)

Ola Abdulganiy Shittu, Odera Patroba (South Africa), Godah Walyeldeen (Poland) and Joseph Odumosu (Nigeria)

Ola Abdulganiy Shittu
University of Cape Town, South Africa
South Africa

Corresponding author Ola Abdulganiy Shittu (email: ganishsurveyors[at]gmail.com)

[ abstract ] [ paper ] [ handouts ]

Published on the web 2025-03-16
Received 2024-12-02 / Accepted n/a

This paper is one of selection of papers published for the FIG Working Week 2025 in Brisbane, Australia PEER REVIEW in Brisbane, Australia and has undergone the FIG Peer Review Process.

FIG Working Week 2025 in Brisbane, Australia PEER REVIEW
ISBN n/a ISSN 2307-4086
URL n/a

Abstract

Understanding the underlying causes of vertical surface deformation (VSD) is crucial to planning for monitoring, prediction and mitigation of natural hazards. An assessment of the factors inducing VSD across Ifewara / Zungeru fault zone is carried out in this research using Ridge Regression (RR) and Random Forest (RF) machine learning (ML) algorithms. Based on SBAS-InSAR and using Sentinel-1 images, the Line of Sight (LoS) deformation in the study area at the rate of ±30mm/year was used to categorize the region into subsiding regions of moderately deforming tropical rain forest (MDTRF), low deforming tropical rain forest (LDTRF), Low deforming tropical savanna (LDTS) and uplift regions. Temporal variations of equivalent water thickness (𝛥EWT) and VSD from 2002 to 2023 were obtained from Gravity Recovery and Climate Change Experiment and Follow On satellite missions (GRACE/GRACE-FO) using IGiK-TVMGF V1.0 software. Rainfall data for the period was retrieved from the Center for Hydrometeorology and Remote Sensing (CHRS), while precipitation, earth skin temperature (TS), temperature(T2M), and relative humidity (RH2M) were obtained from Modern-Era Retrospective Analysis for Research and Applications Version-2 (MERRA 2). The data were modelled using RR and RF ML algorithms at the selected points of interest across the study area to determine the relationship between VSD and other factors (𝛥EWT, TS, T2M, RH2M, Rainfall and Precipitation). The development of the ML models was done using the extracted data from 2002 to 2020 at a ratio of 80:20 for training and testing, followed by factor analysis on the resulting model. The model was further validated by a Leave-Out (LO) validation approach using the selected ML model to predict VSD for the years 2021 to 2023, and the prediction output compared with the actual data. Obtained correlation coefficient, RMSE, R2 and SMAPE for RR are 0.80, 0.05,0.61, 5.04% and for RF are 0.86, 0.019, 0.63, 5.26% respectively which revealed a strong fit. Factor analysis performed at all the stations consistently revealed that 𝛥EWT is the highest contributor to VSD across the study area with contributions exceeding 90% at all the stations in MDTRF, LDTRF, LDTS and 34% in uplift areas followed by temperature with RF. While with RR, 𝛥EWT has an average contributory factor of 74.2% in 𝛥EWT MDTRF, LDTRF, LDTS and 8.87% in uplift areas followed by RH2M and T2M.The study recommends that adequate monitoring of 𝛥EWT within the study area should be embarked upon, since it could be used as precursor for monitoring VSD.

Keywords: Deformation measurement; Vertical surface deformation, Equivalent water thickness, Machine learning, Random Forest, Ridge Regression, factor analysis.

Back