COMPREHENSIVE EVALUATION OF GULLY EROSION PRONE ZONES IN AKWA

IBOM NORTH WEST REGION USING GEOSPATIAL APPROACH

Background of the study

Gully erosion is an erosive process of natural disaster characterized by land loss mechanisms
posing a serious threat to the shape of the earth’s surface and its ecosystem.
When gully erosion is not properly studied and monitored, the control of such gully will be
difficult. This study will provide qualitative information on gullying which extends historical
record and fills the existing gap through remotely sensed imagery of the study area.
Comprehensive evaluation of Gully erosion and inventory Gully map developed to enhance
scientific knowledge of gully-driving processes and reveals gully dynamics, knowledge of
interactions of the dominant drivers and other predisposing factors underlying gully formation.
This study aims to delineate the areas with the most severe gully erosion susceptibility (GES)
prone zones using geospatial approach and machine learning (ML) modeling principle for
analysis. Nevertheless, accurate identification of gullies is a function of the selected ML
algorithms, the image and number of classes used for the study. We applied Linear Discriminant
Analysis (LDA), Support Vector Machine (SVM), and Random Forest (RF) on Sentinel 2
satellite imagery to extract gullies and investigate it qualitative information.
Developed modeling framework that uses remote sensing environmental data to predict
gully erosion susceptibility more accurately have now been deployed as a scientific tool to
analyze its spatial existence for better studies of gullies generation. This predictive model allows
landowners, conservation agencies and other stakeholder to direct management resources to the
most vulnerable areas.

Research problems
Gully erosion is a basic source of soil deterioration in many regions of the world. It is
now generally recognized as a global indication of desertification and land degradation, as it is
viewed as a potentially destructive process that poses a threat to life and property.
In the past, comprehensive analysis of gullies has often been neglected by the scientific
community. For example, a review by Casali et al (2009) shows that less than 10% of soil
erosion studies around the world address gullies directly. However, in order to develop strategies
to control, prevent and rehabilitate gullies, the spatial extent of the problem has to be established
and monitored (Le Roux et al., 2007; Kakembo et al., 2009). Gully erosion as the dominant
environmental problem in the southeast region of Nigeria has led to the isolation of villages,
severance of communication lines such as roads, as well as loss of homes, schools, human and
material resources (Egboka et al., 1990).

Geographic Information Systems (GIS) and Remote Sensing (RS) technologies have been
frequently used to assess soil erosion features. However, in Akwa Ibom North West region there
has been lack of information regarding the spatial extent of gullies both at the local and regional
scale. The evaluation of this process and the elements that contribute to it generation are
becoming increasingly essential. Hence the need for this study.

Justification of study

The growing interest in gully erosion research stems from a desire to better understand its
impacts and as a result try to mitigate their negative consequences. This is achieved through the
spatial modeling of the gullies, as well as evaluation of each variable that influence their
development. Conventional methods of studying gully erosion susceptibility pose a difficult task
that requires a lot of money, effort and resources. With the advancement of geographic
information systems (GIS), remote sensing, and the evolution of computational and intelligent
solution algorithms, geospatial modeling has rendered the process of studying susceptibility to
gully erosion simpler, affordable and novel. The techniques enabled the study of gully erosion
susceptibility at large scale and developed maps to indicate areas more vulnerable to erosion for
future protection. This research will be of great importance to the communities in the study area,
urban planners, agriculturalist and environmentalists as it will boost information database to
counter the threat of land degradation, loss of farmlands, damages to buildings, roads and other
infrastructure including life.

Aim
The aim of the study is to identify Gully Erosion prone zones and it Spatial Distribution
in North West Region of Akwa Ibom State using Remote Sensing and GIS.

Objectives of the study

1. To evaluate gully erosion hotspots, drainage and lineament patterns as well as creating a
database to the development of gullies.
2. To Identify and Evaluate the impact of gully erosion in the study area through
Benchmark and Machine Learning Models for the characterization of gullying
vulnerability.
3. To produce relevant thematic maps for the conditioning factors and establish the
correlation between predisposing factors on gully occurrence.
4. To Produce a Gully erosion susceptibility maps (GESM) as well as developing a
prediction framework for severity of gullies in the region.
 Research Questions

The study seeks to address the following research questions;

1. What is the distribution and spatial extent of prone gully sites in the study area?
2. What are the gully dynamics over a period of time, and the degree of land degradation (gully
density) in the study area?
3. Which method employed in the study yield the most reliable prediction framework model for
gully erosion analysis?

RESEARCH METHODOLOGY
i. Research Design:
This study is on two basic approaches; A Benchmark model employed to identified and predict
areas prone to gully Erosion and Machine Learning Algorithms in the analyses of gullies to
establish land cover changes over the time epoch of 2015 to 2023. Remote sensing imagery of
Sentinel 2 archive, were used as a low-cost data source to allow analyses of gullies over the time
frame.
In this study, the Benchmark model employed a GIS-based Frequency Ratio (FR) and
Analytical Hierarchy Process (AHP) models to predict gully prone areas. This was achieved by a
detailed field exercise, where over fifty gullies were identified with captured coordinates. Google
EarthPro of high-resolution imagery serves as a tool to authenticate the location and spatial
extents of the gullies in line with land truthing exercise. A practically relevant methodology of
three Machine Learning algorithms (RF, SVM, LDA), two approaches of class numbers (binary
and multiclass) and ten combinations of study areas as train and test sets were used to evaluate
the gullies.
The ensemble accuracies of the Benchmark (FR and AHP) models and Machine Learning (LDA,
SVM, and RF) models in the gully prediction and feature extraction were analyzed using
AUROC, SCAI, Statistical measures to includes PPV, FDR, Accuracy, MAE and RMSE.
ii. Area of the Study
The study area covered North Western region of Akwa Ibom State. It constitutes Ten
(10) Local Government area. A total area of 2539 sq.km with the population of about
1,449,921 (NPC,2006). It lies between latitude 169 0 58’86’’to 1690 78’76’’and
longitude 040 68’52’’ to 040 40’09’’in the tropical rain forest zone of Nigeria. The
region is bounded on the East by Cross River State, on the West by Rivers State and
on the North by Abia State. The study area has mean maximum temperature of
27.35°C and average amount of monthly rainfall per year is 247mm3. Average
monthly relative humidity ranges from 51 percent to 89 percent annually. There is
 usually early rainfall in January/February with full commencement of rainy season in
March and stopping in November of each year. The dry season lasts between four to
five months. The highest rainfall is recorded from July to October with little break in
rainfall in August popularly described as “August break”. (Source: Nigerian
Meteorological Agency, Akwa Ibom State Meteorological Inspectorate, Uyo.)

iii. Reasons for Choosing Geospatial Approach, Machine Learning and Benchmark
Models:
The selected method for evaluation of gullies were adopted based on the following
research findings:
(a) It is more accessible, widely acceptable as an efficient and realistic methods to
gully erosion susceptibility study.
(b) It gives a novel regression approaches to Machine Learning Models where true
representation, evaluation and optimization principles are ensemble for
comprehensive study.
(c) It posed a versatile and veritable research tools for gullies assessment and
prediction visualization.
(d) Its ability to mapped predisposing factors underlying gully formation and
providing a reliable soil information index.
(e) Its enhanced validation of results through the following test modules;
- Area under the receiver operating characteristics (AUROC) curve.
- Seed cell area index (SCAI).
- Prediction rate curve
- Statistical measurement such as Precision (PPV), false discovery rate (FDR),
Accuracy, mean absolute error (MAE), Root mean squared error (RMSE).

iv. DATA ANALYSIS AND PRESENTATION

Data will be analyzed and presented as Maps, charts, statistical tables and statements.