Spatio-Temporal Analysis of Land Use Change and Prediction in an Artisanal Mining Hotspot: A Multi-Decadal Study of the Kampene Mining Area, DRC
Espoir M. Birhenjira1,2*orcid, Hwaba Mambo3orcid, Isaac Bishikwabo4orcid, Christian Amani5,6, Jacques M. Batumike1,7orcid
1Department of Geology, Faculty of Science, Université Officielle de Bukavu, Bukavu, Democratic Republic of the Congo.
2Faculty of Geoscience, Geotechnics and Mining, Technische Universität Bergakademie Freiberg, Freiberg, Germany.
3Department of Geography and Environmental Management, Institut Supérieur Pédagogique de Kaziba, Kaziba, Democratic Republic of the Congo.
4Institute of Water and Energy Sciences Including Climate Change, Pan African University (PAUWES), University of Tlemcen, Tlemcen, Algeria.
5Center for International Forestry Research (CIFOR), Bogor, Indonesia.
6Department of Biology, Faculty of Science, Université Officielle de Bukavu, Bukavu, Democratic Republic of the Congo.
7Department of Geology, Faculty of Science, Université de Lubumbashi, Lubumbashi, Democratic Republic of the Congo.
DOI: 10.4236/jep.2025.1612067   PDF    HTML   XML   20 Downloads   168 Views  

Abstract

Artisanal and small-scale (ASM) gold mining activities contribute to the livelihoods of about 200,000 miners and their dependents in the eastern provinces of the Democratic Republic of the Congo (DRC). Despite their socio-economic importance, mining activities cause significant alterations to the environment, especially due to their impact on forest cover. Deforestation at mining sites, timber consumption, and pollution are among the observed direct impacts of mining activities. Land use and land cover (LULC) changes have been assessed for the period between 1984 and 2020 and predicted for the period between 2020 and 2040 using Landsat images in the Pangi Territory in general, and the Kampene mining area in particular, in the Maniema Province, known as an important mining province in the eastern part of the DRC. In this study, the Support Vector Machine algorithm, Post Classification Change Detection, and Markov Chain Analysis were used to classify, detect, and predict LULC changes, respectively. The results indicate a significant decrease in forest cover from 1984 to 2020, mainly due to the conversion of these areas into agricultural lands, while a significant loss is predicted for the period between 2020 and 2040 as the need for more croplands will increase mainly because of population growth. A transition probability of 0.25 is found for the conversion of primary forest to bare soils and mines, which is comparable with the greatest transition probability of 0.28 obtained in this study for the conversion of primary forest into agricultural lands. This shows the harmful effects of mining activities on the destruction of forest cover. The implementation of land protection and restoration policies and strong control of mining activities are required to reduce the impacts on the environment and forests.

Share and Cite:

Birhenjira, E.M., Mambo, H., Bishikwabo, I., Amani, C. and Batumike, J.M. (2025) Spatio-Temporal Analysis of Land Use Change and Prediction in an Artisanal Mining Hotspot: A Multi-Decadal Study of the Kampene Mining Area, DRC. Journal of Environmental Protection, 16, 1258-1289. doi: 10.4236/jep.2025.1612067.

1. Introduction

The pressure on natural resources is particularly strong in developing countries where demographic factors, including extreme poverty, push people to focus on short-term survival goals. Although there is a large consensus that tropical Africa is the region most affected by environmental problems [1] [2], the situation in the eastern part of the Democratic Republic of the Congo (DRC) is by far the worst. In this part of the globe, an endless series of political turmoils has led the population to rely on natural ecosystems for their living.

Although at the global scale, agriculture can be considered the main driver of deforestation, forest degradation, and even habitat fragmentation [3]-[6]; the situation in eastern DRC is exacerbated by other phenomena including mining activities, largely dominated in recent years by international mining companies. However, there is a clear contrast between the financial returns made by the direct beneficiaries of these activities and the real environmental challenges. Even the protected areas have been severely affected by mining and other extractive activities over the last decades [7]-[9].

Mineral exploitation contributes significantly to economic growth and development in most world economies [10]. The current contribution of mining activities to the GDP of many African countries and the rising demand and recent spikes in mineral prices indicate that both social and ecological effects of the industry are likely to grow in the forthcoming years [11]-[13]. Despite their important contribution to GDP in the DRC (nearly 25%), mining activities cause significant land alterations [14]. The expansion of mining concessions threatens the DRC’s forests, even though nearly 12% of the forests are under some form of protection [14]. In the eastern provinces of the DRC (where most of the Congolese gold is mined), artisanal and small-scale (ASM) gold mining contributes to the livelihood of about 200,000 miners and their dependents. In this region where mining activities started in the twentieth century, the environmental consequences of artisanal mining exploitation are largely negative, including chemical pollution of water tables, deforestation, diversion of rivers, leveling of hills and reduction of arable land, as well as intensive poaching in adjoining national parks [15]. In gold and base metal mining areas, sulfide oxidation resulting from chemical and biogeochemical processes leads to the production of low pH groundwater that induces the dissolution of trace metals into the groundwater system in very high concentrations. The groundwater, thus, becomes dangerous for human consumption [16]. Mining activities, especially illegal small-scale mining, deplete environmental resources such as water, soil, landscape, vegetation and ecosystems, among others [10].

Although environmental degradation is persistent in most of the DRC’s mining areas for each mining activity and its phases, the law has provided corresponding environmental plans, including the Mitigation and Rehabilitation Plan (PAR, Plan d’Atténuation et Remediation) during the research phase and the Project Environmental Management Plan (PGEP, Plan de Gestion Environnementale du Projet), which are included in the Mining Code [17]. The PGEP is associated with the Environmental Impact Study (EIE, Etude d’Impact Environnemental) required during the feasibility study phase before the start of mining activities. The principle of environmental protection is enshrined in the constitution of the DRC (articles 53, 54 and 55). Articles 203 and 204 of the Mining Code specify the environmental protection measures during the two major phases of a given mining project, namely the research phase and the exploitation phase. All environmental plans follow the same procedural regime for obtaining environmental permits. The practical modalities are fixed by the Mining Regulations [18], especially in articles from 430 to 436 and from 454 to 456. However, these studies are expensive; thus, most artisanal miners do not have the means to conduct them.

2. Land Use and Land Cover

Understanding the distribution and dynamics of land cover is challenging, but it is critical for better understanding the Earth’s fundamental characteristics and processes, including the productivity of the land, the diversity of plant and animal species, and biogeochemical and hydrological cycles [19] [20]. Land use and land cover (LULC) is one of the most important domains of human-induced environmental transformation [21]. Land use and land cover change (LULCC) is the conversion of different land use types as a result of complex interactions between humans and the physical environment [22].

The mapping of LULCC is an important activity of land management and monitoring [23]. Recent advancements in remote sensing, GIS, and computer technology allow the assessment and monitoring of LULCC at multiple spatial and temporal scales [20] [22] [24]. Although fine-resolution data have been used to examine changes in surface mining extent, many studies are based on Landsat imagery due to its global coverage, medium resolution (30 m), and data acquisition at regular intervals [25]. The monitoring of LULCC and the building of spatio-temporal patterns of change can be done using change detection methods. This use of multi-date images allows for a better understanding of the causes and consequences of the change [26]. Principal component analysis and post-classification comparison are common methods used for change detection for image differentiation in remote sensing change detection [27] [28].

The prediction of future disturbances can be based on different scenarios during the modeling of LULC. The commonly used models for the estimation of future land cover changes include analytical equation-based models, statistical models, evolutionary models, cellular models, Markov models, hybrid models, expert system models, and multi-agent models [22]. Markov chains harness the advantages of long-term predictions, the simplicity of the logic, and efficiency in computation [22] [29] [30]. The assemblage of Landsat spatial, spectral, and temporal resolutions, over a reasonably-sized image extent, results in imagery that can be processed to represent land cover over large areas with an amount of spatial detail that is absolutely unique and indispensable for monitoring, management, and scientific activities [31].

The main objective of the present study is to assess the dynamics of LULCC, in time and space, in Kampene mining areas using Landsat images. Specifically, this study focuses on: 1) the determination of the LULCC over the last 40 years, 2) the prediction of future LULC dynamics from 2020 to 2040 using historical records, 3) the investigation of the relationship between the LULCC and the mining activities.

3. The Study Area

Kampene is located in the Maniema Province in the centre-eastern part of the DRC (Figure 1). The climate in the province is predominantly equatorial, characterised by continuous precipitation throughout the year (mean annual rainfall of 1600 mm), but the southern area is affected by a humid tropical climate with alternating dry and wet seasons. The vegetation also displays features as per climatic patterns, with dense forests in the north and savannahs to the south. The hydrographic network is dense and three-quarters of the rivers belong to the Congo River basin.

This study was undertaken within the chiefdom of Babene, in Pangi Territory, and focused on the Kampene mining area located in the southern portion of the Babene chiefdom, at approximately 110 km southeast of the city of Kindu (capital city of Maniema Province, Figure 1). In this study, Pangi Territory provides the territorial context, while the Kampene mining area is analyzed as a focused hotspot of artisanal mining. Territory-wide LULC assessments were conducted first, followed by a site-scale analysis of Kampene to quantify LULC change within the mining area.

This study region is a forested area known for its cassiterite, diamond, and gold mineralisation hosted in the Kivu Supergroup formations [32]. These resources have been mined since the early 20th century by different companies (including SYMETAIN and COBELMINES Mining Companies) for industrial extraction [15]. Since 1974, the Société Minière et Industrielle du Kivu (SOMINKI, now Société Aurifère du Kivu et du Maniema, SAKIMA since 1997), a privately held

Figure 1. Study area location.

mining company of former Zaire, took over the exploitation of gold and tin mines. Since SAKIMA does not have the means to develop these mines, gold extraction in the area is mainly performed by artisanal miners. The common methodology used to extract gold is panning and amalgamation using chemicals, especially mercury, which is not health-nor environment-friendly.

4. Methodology

4.1. Data Acquisition

Landsat (30-m resolution) level-1 images were downloaded from the United States Geological Survey (USGS) Earth Explorer website (https://earthexplorer.usgs.gov/). The details of the images used in this study are presented in Table 1. The selection of the years’ interval used a quasi-decadal approach. Three sets of ortho-rectified satellite images, namely Landsat Thematic Mapper (TM) image of 1984, Landsat Enhanced Thematic Mapper (ETM) image of 2000, and Landsat OLI-TIRS of 2020, were analyzed to determine the past and current LULCC in the whole Pangi Territory.

Table 1. Details of the acquired Landsat satellite images.

N˚

Date of collection (YY/MM/DD)

Satellite and Sensor

Path - Row

1

1984/6/2

Landsat Thematic Mapper (TM) image

174 - 62

2

1984/6/2

174 - 63

3

1984/9/13

175 - 62

4

1984/6/9

175 - 62

5

1984/11/6

175 - 63

6

2000/5/5

Landsat Enhanced Thematic Mapper (ETM) image

174 - 62

7

2000/7/24

174 - 63

8

2000/5/12

175 - 62

9

2000/4/10

175 - 62

10

2000/5/18

175 - 63

11

2020/4/2

Landsat OLI-TIRS

174 - 62

12

2020/5/4

174 - 63

13

2020/2/21

175 - 62

14

2020/5/27

175 - 63

Scene selection was based on two criteria: the acquisition date and the cloud cover. Data from the same season give uniform spectral and radiometric characteristics and minimise the seasonal variation in spectral reflectance of land cover types [26]. Images captured between May and August (dry season) were chosen to best distinguish the spectral signatures of the different types of land cover, especially the uncovered areas [24]. The acceptable cloud cover threshold was less than 10%. Some images were downloaded and mosaicked with other images of the same scene and year to reduce their cloud cover. A total of 4 to 5 different images per year were required to cover the entire study area.

The projection of images into the GIS platform uses the WGS-1984 and UTM 35S-Zone Coordinate System. The digital elevation model (DEM, 30-m resolution) was downloaded from the USGS Earth Explorer website, while all administrative shapefiles and road networks were downloaded from the “Référentiel Géographique Commun” (RGC) website (http://www.rgc.cd).

4.2. Image Pre-Processing

The following steps were adopted, as per [33]:

  • Radiometric calibration: image data calibration from digital number values to top-of-atmosphere (TOA) reflectance values (ranging from 0 to 1). The gains, offsets, solar irradiance, sun elevation, and acquisition time information were used, and they are presented in the satellite image metadata.

  • QUick Atmospheric Correction (QUAC): an automated atmospheric correction method for multispectral and hyperspectral imagery that works with the visible and near-infrared through shortwave infrared (VNIR-SWIR) wavelength ranges. The QUAC determines the atmospheric correction parameters directly from the observed spectra pixels in a scene, without ancillary information [34].

  • Dark Subtraction: removal of residual atmospheric scattering effects from an image by subtracting a pixel value that represents a background signature from each band.

  • Image pan sharpening: an image fusion method in which high-resolution panchromatic data is fused with lower resolution multispectral data to create a colourised high-resolution fused dataset. The principal component pan sharpening method was used to increase the image resolution from 30 m to 15 m [35].

  • Mosaicking: the Seamless Mosaic workflow was used to merge georeferenced images into one image. This workflow applies color balancing and edge feathering to create high-quality mosaics.

The pre-processed image after these steps (Figure 2(b)) looks better than the raw image (Figure 2(a)).

(a) (b)

Figure 2. Images before preprocessing (a) and after preprocessing (b).

4.3. Image Processing

4.3.1. Reference Data Collection

The reference data were collected using the Collect Earth Online (CEO) tool, which is a free, open-source, and user-friendly tool for land monitoring. It uses a Google Earth interface in conjunction with an HTML-based data entry form and facilitates land use assessment through a sampling approach rather than wall-to-wall mapping. However, land use data (point vector files) generated with CEO can be used as training sites for wall-to-wall image classification [36].

A total of 100 ground control points were collected as reference data for the year 2020, and 80 points per year for each of the 1984 and 2000 classifications, using a stratified approach of 20 points per class.

4.3.2. DRC Land Use and Land Cover Type Requirements

The operational guide on the forest classification standards of DRC was used for the definition of different classes. There are 3 major subdivisions: non-forest land (land for mining, areas of human occupation, agricultural plantations, and other lands), productive forest land (secondary forests, primary forests on dry land, gallery forests, forest on hydromorphic soil) and non-productive forest land (barren land and savannahs) [37].

Thus, for this study, primary forests on dry land are considered as primary forests, agricultural plantations as croplands or agroforestry, zones of human occupation as settlements, humid denuded lands as wetlands, and bare soils and mining sites as bare soils and mines.

4.3.3. Supervised Classification

Generally, a good classifier should be able to discriminate pixels into desirable land covers. The factors taken into consideration when selecting a classification method include accuracy, processing speed, and practicality [38]. The maximum likelihood classification (MLC) and artificial neural network (ANN) are among the frequently used methods in the classification of land covers. However, ANN has been associated with overfitting and local minima problems [39], while MLC needs a large training area and the assumption that data are normally distributed. The support vector machine (SVM) is one of the more reliable classification methods developed recently [38] [40].

The SVM is a supervised classification method derived from statistical learning theory that often yields good classification results from complex and noisy data. It separates the classes with a decision surface called a hyperplane that maximizes the margins between the classes. The data points closest to the hyperplane are called support vectors, which are the critical elements of the training set [41] [42].

The SVM can be transformed into a non-linear classifier using non-linear kernels. While SVM is a binary classifier in its simplest form, it can function as a multi-class classifier by combining several binary SVM classifiers (creating a binary classifier for each possible pair of classes) [41]-[43]. The pairwise classification strategy for multi-class classification was used in this study. The basic idea for understanding the SVM classification is presented in Figure 3.

Figure 3. Basic idea of SVM [44], showing the support vectors in bold.

The use of CEO with Google Earth has led to five main LULC classes in Pangi Territory: the Primary Forest, the Croplands/Agroforestry, the Settlements, the Wetlands and the Bare Soils/Mines. Mines have been detected by searching for features that look similar to the ones shown in Figure 4 on Google Earth based on a contextual knowledge of the study area.

(a) (b)

Figure 4. View of mines as detected on Google Earth: (a) small scale and (b) large scale.

4.3.4. Land Use and Land Cover Change Detection

The timely and accurate change detection of Earth’s surface features is extremely important for understanding the relationships and interactions between human and natural phenomena in order to promote better decision making [27].

In this study, the Change Analysis tool [44] was used for LULCC detection. This tool is a component of the Land Change Modeler embedded in the software. It uses a post-classification change detection method and requires two (earlier and later) land cover images. The LULCC has been assessed for two different periods: from 1984 to 2000 and from 2000 to 2020.

In post-classification change detection, the images from each time period were classified using the same classification scheme into a number of discrete categories (e.g., land cover types). The two (or more) classifications are compared, and the area that is classified the same or different is tallied [45].

4.3.5. Land Use and Land Cover Change Prediction

The land cover change modeling means a time interpolation or extrapolation when the modeling exceeds the known period. The Markov Chain model treats the LULCC as a stochastic process; the later state (land cover type) of a pixel is only related to its immediately preceding state, but not to any other previous states. A transition probability is the direct outcome of the Markov Chain model. LULCC being a very complex process, the model includes social, economic, historical, and biophysical factors as change drivers to prevent its success [29].

In this study, the LULCC has been predicted up to the year 2040, using the Markov Chain Analysis (MCA). The LULC prediction has been assessed in two main steps:

1) Creation of Transition Potentials

First, the different land cover changes have been grouped into Transition Sub-Models, each of which includes all the changes from different classes to the same class (e.g., all the changes transitioning from different classes to the Settlement class were grouped into the Settlement Sub-Model). Then, five change drivers are added, including slope, elevation, distance from roads, distance from settlements, and distance from bare soils/mines. Roads were considered one of the main factors influencing LULCC because, after different years’ classifications, obvious changes were observed along the roads. These drivers were selected because they are widely used in existing literature on tropical regions, where they have demonstrated a strong influence on LULC changes [46] [47].

Finally, the Transition Potential is created for each Transition Sub-Model.

2) Change prediction

The Markov Chain model simulation process produces mainly a land use area transfer matrix and a probability transfer matrix to predict land use change trends. The Markov Chain model can be described as a set of states, S = {S0, S1, S2, …, Sn}, assuming that the current state is St, and then, it changes to state Sj at the next step with a probability denoted by transition probabilities Pij. Thus, state St+1 in the system could be determined from the former stage St in the Markov Chain using the formula in Figure 5 [22]. The transition probability matrix from 1984 to 2020 and the predicted LULC map for 2040 were obtained using MCA in TerrSet software [44].

Figure 5. Matrix for state transition probability (Pij) calculation. S is land use status, t and t + 1 are the time points.

4.3.6. Classification Validation and Accuracy (Kappa Coefficient)

The error or confusion matrix is the standard method used to assess the classification accuracy and measure the quality of the classification system of an image [48] [49]. It is typical to extract several statistics from the error matrix, including overall accuracy, Kappa coefficient, producer’s accuracy, and user’s accuracy [49]. In this study, the overall accuracy and the Kappa coefficient were used to assess the accuracy of the different years’ classifications, and the omission and commission errors were used to evaluate the class-specific classification errors.

A total of 50 ground control points were collected to assess the accuracy of the 2020 LULC classification, while 40 points were collected for the 2001 classification and 40 other points for the 1984 classification. Thus, 10 points were collected per class for accuracy assessment. Figure 6 summarizes all the methodological steps.

Figure 6. Methodology workflow flowchart.

5. Results and Discussions

5.1. Land Use and Land Cover Change in Pangi Territory from 1984 to 2020

5.1.1. Land Use and Land Cover Mapping

The results indicate a decrease in forest cover and an increase in other LULC classes such as settlements, croplands and agroforestry, bare soils, mines, and wetlands between 1984 and 2020 (Table 2). The observed considerable decrease in forest cover from 13338.02 km2 (in 1984) to 11080.26 km2 (in 2020) (Table 2) is mostly due to human activities such as agriculture, building, and mining (Figure 7). Mining activities increased significantly from 1984 (10.15 km2) to 2020 (50.94 km2), as well as agricultural activities (from 871.95 km2 in 1984 to 2885.24 km2 in 2020) and settlements (from 81.49 km2 in 1984 to 208.21 km2 in 2020) (Figure 7). This is mainly due to the increase in human population in Maniema Province from 1984 to 2020. This province has registered around 148,000 internally displaced people (IDPs), including 90% due to clashes and armed attacks in the neighbouring Sud-Kivu Province. The territory of Pangi has registered 32,755 IDPs from 2009 to June 2016 [50].

Table 2. Summary of the area covered by different land use and land cover classes from 1984 to 2020.

Class

1984

2000

2020

Area (km2)

%

Area (km2)

%

Area (km2)

%

Bare soils/Mines

10.15

0.07

66.36

0.46

50.94

0.36

Wetlands

0

0

0

0

119.96

0.84

Croplands/Agroforestry

871.95

6.1

2179.41

15.24

2885.24

20.17

Primary Forest

13338.02

93.26

12008.15

83.96

11080.26

77.48

Settlements

81.49

0.57

47.69

0.33

165.23

1.16

Total

14301.62

100

14301.62

100

14301.62

100

Figure 7. Illustration of land use and land cover changes from 1984 to 2020: (a) 1984, (b) 2000, and (c) 2020.

The supervised classification of Landsat images using the Support Vector Machine algorithm produced the LULC maps with overall high accuracy (83.3%, 86.8%, and 87.7% for 1984, 2000, and 2020, respectively). The Kappa coefficients were 0.77, 0.82, and 0.84 for 1984, 2000, and 2020, respectively (Table 3). The Bare Soils and Mines classes, which cover smaller areas mainly inside the forests, could be easily confused with other classes in general and with forests in particular, so they had higher omission and commission errors for all the years’ classifications (Figure 8). According to [51], heterogeneous classes that occupy smaller areas can easily be confused with other LULC classes.

Table 3. Accuracy assessment for land use and land cover classification for the years 1984, 2000, and 2020.

Year

Overall accuracy(%)

Kappa

Mean omission/class

(%)

Mean commission/class (%)

1984

83.3

0.77

10.25 ± 14.58

13.24 ± 15.83

2000

86.8

0.82

8.25 ± 13.56

11.35 ± 14.35

2020

87.7

0.84

11 ± 10.24

11.41 ± 6.70

Figure 8. Commission and omission errors per class for each year’s classification: (a) 1984; (b) 2000; and (c) 2020.

5.1.2. Land Use and Land Cover Change

The Primary Forest lost up to 1350 km2 between 1984 and 2000, and 900 km2 from 2000 to 2020 (Figure 9). During the 1984-2000 period, agricultural lands gained up to 1300 km2 and more than 700 km2 from 2000 to 2020.

More than 1200 km2 of Primary Forest lost from 1984 to 2000 were converted into agricultural areas (croplands and agroforestry), and 42 km2 were converted into Bare Soils and Mines (Figure 10). The loss of Primary Forest from 2000 to 2020 was up to 913 km2, including 720 km2 converted into Agricultural lands, 80 km2 converted into Settlements, 100 km2 gained by Wetlands, and 13 km2 gained by Bare Soils and Mines (Figure 11). From 1984 to 2020, forest cover lost more than 2250 km2 that were converted into other LULC classes, with the larger area converted into agricultural lands.

(a)

(b)

Figure 9. Total area (km2) gained or lost by each LULC class from: (a) 1984 to 2000 and (b) 2000 to 2020.

(a)

(b)

(c)

(d)

Figure 10. Contribution to net change in each class: (a) Primary Forest, (b) Croplands/Agroforestry, (c) Settlements, and (d) Bare Soils/Mines from 1984 to 2000.

(a)

(b)

(c)

(d)

(e)

Figure 11. Contribution to net change in each class: (a) Primary Forest, (b) Croplands/Agroforestry, (c) Settlements, (d) Wetlands, and (e) Bare Soils/Mines from 2000 to 2020.

Similar results were observed in the Lwiro area (Sud-Kivu) [21] where the trees were cut as timber for more lucrative benefits, and some parts of the forest were converted into agricultural lands. Furthermore, it has been observed that changes that occurred in and around mining areas were very significant, leading to significant deforestation. This is in line with [11], who observed many conflicts for land use in the DRC, particularly between mining activities and forests. The observed direct impacts of mining activities on forests include deforestation at mining sites, timber consumption, and pollution.

5.1.3. Prediction of Land Use and Land Cover

The LULC prediction for 2040 indicates that forest areas will cover 9601.90 km2, while agricultural lands will cover 4213.32 km2, and there will be a total area of 59.99 km2 for Bare Soils and Mines, 218.20 km2 for Wetlands and 208.21 km2 for Settlements (Figure 12, Table 4). This will represent a total deforestation of 28% from 1984 to 2040, while a deforestation of 16.93% is observed from 1984 to 2020. This indicates a deforestation rate of 83.12 km2 per year from 1984 to 2000, 46 km2 per year from 2000 to 2020, and 73.92 km2 per year from 2020 to 2040.

Table 4. Surface area of different land use and land cover classes by year.

Class

1984

2000

2020

2040

Area (km2)

%

Area (km2)

%

Area (km2)

%

Area (km2)

%

Bare soils/Mines

10.15

0.07

66.36

0.46

50.94

0.36

59.99

0.42

Wetlands

0

0

0

0

119.96

0.84

218.2

1.53

Croplands/Agroforestry

871.95

6.1

2179.41

15.24

2885.24

20.17

4213.32

29.46

Primary Forest

13338.02

93.26

12008.15

83.96

11080.26

77.48

9601.90

67.14

Settlements

81.49

0.57

47.69

0.33

165.23

1.16

208.21

1.46

Total

14301.62

100

14301.62

100

14301.62

100

14301.62

100

Figure 12. Change detection maps: (a) 1984, (b) 2000, (c) 2020 and (d) 2040.

The Markov transition matrices present the transition trends or transition probabilities from 1984 to 2020 (Table 5). In the transition matrices, the diagonal values represent the probability that each LULC class remains persistent from time 0 to time 1, while the other values represent the probability that a given LULC class undergoes transition to another LULC class. Over the whole period from 1984 to 2020, only Primary Forest attains the highest probability of persistence, with a transition probability exceeding 0.75 [51].

Table 5. Markov chain matrix of LULC transition probabilities from 1984 to 2020.

Primary Forest

Croplands/Agroforestry

Settlements

Wetlands

Bare Soils/Mines

Primary Forest

0.86

0.12

0.01

0.01

0.00

Croplands/Agroforestry

0.28

0.70

0.02

0.00

0.00

Settlements

0.14

0.37

0.45

0.01

0.03

Wetlands

0.25

0.25

0.25

0.00

0.25

Bare Soils/Mines

0.25

0.42

0.14

0.06

0.13

The Primary Forest in the Pangi Territory that covered 13338.02 km2 (93.26% of the total area) in 1984 was reduced to only 11080.26 km2 (77.48%) in 2020 (Table 4), representing a loss of 2256.8 km2 (15.78% of the total area). Croplands/Agroforestry, which covered 871.95 km2 (6.10% of the total area) in 1984, accounted for 2885.24 km2 (20.17% of the total area) in 2020, representing an increase of 2012.24 km2 (14.07% of the total area). This means that 89% of the total loss of the forest area was transformed into Croplands/Agroforestry. Agriculture was identified by [52]-[55] as being by far the main cause of deforestation in the tropical world. The expansion of infrastructure, the development of the mining sector, and the extraction of timber represent the other observed causes.

The probability of transition from Primary Forest to Croplands/Agroforestry was the greatest change probability (0.28) compared with the transition probability from Primary Forest to Bare Soils/Mines (0.25), which is similar to that from Primary Forest to Wetlands (Table 5). This effectively proves that mining activities are among the most important causes of forest cover decrease in the study area.

5.2. Relationship between Land Use and Land Cover Change and Mining Activities in the Kampene Area

5.2.1. Land Use and Land Cover Mapping

The extension of some LULC classes such as Bare Soils/Mines, Wetlands, Croplands/Agroforestry and Settlements, and their impacts on the destruction of Primary Forest cover, which decreased significantly from 1984 to 2020 in the Kampene mining area and surroundings, are presented in Figure 13. Furthermore, it is predicted that, in 2040, a significant loss will be observed in Primary Forest (up to 349.45 km2), which will mainly be converted into Croplands/Agroforestry. The increase in Bare Soils/Mines ranges from 0.36 km2 in 1984 to 5.28 km2 in 2020 (Table 6), representing up to 1466.66%. This exceptional increase in mining activities is one of the drivers of deforestation and several other changes observed in the local landscapes. In the eastern DRC, the environmental consequences of artisanal mining exploitation are all highly negative, including chemical pollution of water tables, deforestation, diversion of rivers, levelling of hills and disappearance of arable land as well as intensive poaching (to feed the miners) [15].

Table 6. Surface of different land use and land cover classes in the Kampene mining area and its surroundings.

Class

1984

2000

2020

2040

Area (km2)

%

Area (km2)

%

Area (km2)

%

Area (km2)

%

Bare soils/Mines

0.36

0.02

4.45

0.21

5.28

0.25

13.39

0.64

Wetlands

0

0

0

0

7.43

0.36

13.43

0.65

Croplands/Agroforestry

146.38

7.04

302.79

14.56

384.69

18.49

714.82

34.37

Primary Forest

1926.42

92.61

1769.99

85.09

1670.6

80.32

1321.15

63.52

Settlements

6.87

0.33

2.8

0.13

12.04

0.58

17.25

0.83

Total

2080.03

100

2080.03

100

2080.03

100

2080.03

100

Figure 13. LULC distribution of the Pangi Territory at three different times: (a) 1984, (b) 2000, and (c) 2020; and Markov model prediction for 2040 (d).

5.2.2. Land Use and Land Cover Change

Primary Forest lost up to 22.4 km2 from 1984 to 2000, gained by Agricultural lands and Bare Soils/Mines, with the majority transformed into Agricultural lands (21 km2, Figure 14). The loss was greater between 2000 and 2020 (33 km2) and mainly gained by Agricultural lands and Settlements.

Settlements gained 1.30 km2 from Agricultural lands. The Primary Forest lost 28 km2 to Agricultural lands between 2000 and 2020, while Agricultural lands lost 3.50 km2, which were converted into Settlements (Figure 15). Settlements also gained 3.10 km2 from Primary Forest and 0.50 km2 from Bare Soils/Mines. Wetlands gained 1.85 km2 from Primary Forest, 0.20 km2 from Agricultural lands, and 0.15 km2 from Bare Soils/Mines (Figure 16). Figure 17 presents the change detection map from 1984 to 2000 and the map of LULC change detection from 2000 to 2020.

(a)

(b)

Figure 14. Land use and land cover distribution in the Kampene mining area and the surroundings at: (a) 1984; (b) 2000.

(a)

(b)

(c)

(d)

Figure 15. Contribution to net change in each land use and land cover class: (a) Primary Forest; (b) Croplands/Agroforestry; (c) Settlements; (d) Bare Soils/Mines) in the Kampene mining areas from 1984 to 2000.

(a)

(b)

(c)

(d)

(e)

Figure 16. Contribution to net change in each land use and land cover class (a) Primary Forest; (b) Croplands/Agroforestry; (c) Settlements; (d) Bare Soils/Mines; (e) Wetlands in the Kampene mining areas from 2000 to 2020.

Figure 17. Change detection maps (a) from 1984 to 2000 and (b) from 2000 to 2020.

5.2.3. Forest Cover, Forest Loss, and Mining Activities

Several authors have identified agriculture as being by far the main cause of deforestation in the tropical world (e.g., [51] [53]-[55]). The Congo basin is not an exception to the rule as agriculture is the most important direct cause of deforestation. This same trend is observed in the Kampene mining area (Table 6), where the Primary Forest was reduced by 255.82 km2 (12.3% of the total area) from 1984 to 2020. During the same period, the area covered by the Croplands/Agroforestry class increased by 238.31 km2 (11.46% of the total area). This means that 93.16% of the lost forest area has been turned into Croplands/Agroforestry. An increase of 4.92 km2 (0.24% of the total area and 1.92% of the lost forest area) for Bare Soils/Mines, 7.43 km2 (0.36%) for the Wetlands and 5.17 km2 (0.25%) for Settlements was observed during this period.

The same trend was observed in the littoral region of Cameroon, where the area of high-value forest landscapes decreased by 12% from 1975 to 2017, but conversely, disturbed vegetation, cleared areas, and urban areas all expanded by 32%, 5.6%, and 6.6%, respectively [56]. In the Sangha Trinational Park region of Cameroon, the Central African Republic, and the Republic of Congo, the amount of deforestation induced by small-scale mining is lower compared with the impact of other activities such as logging, subsistence farming, or charcoal production [57] [58].

This pattern shows that the direct impacts of mining on deforestation are modest compared with their indirect impacts in Kampene. Only a small portion of forest cover was converted to bare land/mines (1.8 km2), corresponding to the land cleared for pits and infrastructure, whereas a much larger area was converted to agricultural lands and settlements (33 km2 and 0.8 km2, respectively), both supporting mining activity. These constitute indirect impacts, essential for daily life but outside the immediate extraction footprint. Moreover, despite relatively low direct conversion to mines and settlements, the creation of new infrastructure (e.g., roads) further exacerbates LULC change by improving access, a key driver of landscape conversion. Areas closest to roads are more affected by land-cover change [59].

In the copperbelt region (SE DRC, Haut Katanga and Lualaba provinces), another mining region of DRC, slash and burn agriculture represents the main cause of deforestation, followed by brick making, firewood collection, charcoal production, logging, and mining [60]. This indicates that mining, although representing one of the causes of deforestation, is not the principal one.

The results of [56] [57] [60] are consistent with [61]’s research on deforestation due to mining activities in the DRC. This study shows that mining could have a smaller footprint than the effects of other major commodity production processes such as agriculture or charcoal production. Figure 15 illustrates the replacement of forests by settlements and tree cutting for the installation of mines.

6. Environmental Impact of Mining Activities

The Pangi Territory is generally well known as a forest-covered area, but mining activities have induced the cutting of trees. Whether it is from artisanal miners or industrial companies, the extraction of minerals leads ipso facto to the conversion of natural habitats, with many ecological consequences both at the local and landscape scales. In a recent assessment of global primary forest landscapes, [62] highlighted the scale of land cover change worldwide. In the DRC, while the general trend reflects a relative stability in the forested areas of the central part of the Congo basin, the eastern regions of the country (e.g., Sud Kivu, Nord Kivu, and Maniema) are known to be alarmingly losing their forests [63] [64]. Following an analysis of Landsat images, from the data published by CARPE during the period between 2000 and 2010, [63] indicated that a significant proportion of Congolese primary forest cover was lost, mainly in the Kivu provinces.

If the loss of vegetation cover (deforestation) is a phenomenon directly associated with mining and other extractive activities, expansion of agriculture and other anthropogenic pressures; natural forests also lose their original quality through degradation, resulting in intense ecosystem imbalances. Such landscape conversion undeniably affects the migration of a large number of animal species (mammals, birds, etc.) and contributes to increasing landscape fragmentation with immense ecological consequences. Similar to habitat destruction, fragmentation is also a cause of declining biodiversity. It is known to severely impact animal and plant communities by altering natural processes in ecosystems [6] [65]-[71].

Due to poor conservation measures and the absence of updated inventory data, the risk of large-scale biological erosion is high, especially in the absence of research studies and a thorough environmental impact assessment of extractive activities. At a time when the whole world is focusing on the preservation of forest ecosystems in the process of fighting climate change and conserving biodiversity, an analytical study of the consequences of anthropogenic pressures on natural ecosystems in eastern DRC becomes indispensable. Moreover, it is common knowledge that the loss or reduction of vegetation cover inevitably affects the hydrological cycle of the concerned region, inducing the disruption of biogeochemical cycles and the physiological phenomena specific to plant organisms. All of this can, in turn, affect the supply systems of local people and their well-being in general.

Land conversion and intense deforestation in the watersheds significantly increase the risk of erosion, resulting in a positive feedback loop. Erosion accelerates the sedimentation of aquatic ecosystems. Sedimentation and its subsequent turbidity (lack of clarity in water) are known to reduce the biological productivity of aquatic systems through the decrease of ecosystem primary productivity, thus disrupting the whole trophic levels in water bodies. The resulting species reduction affects the overall biodiversity and impacts human communities that depend on fishing for their livelihood.

Pollution is another threat to the water bodies in eastern DRC because of soil erosion, due to mining and other extractive activities. The immediate consequence for aquatic ecosystems is eutrophication through mineral and nutrient enrichment, all this resulting in decreased biodiversity and reduced water quality. Pollution of aquatic ecosystems may also result in serious health issues for human populations, among which, in the case of eastern DRC, are water-borne diseases such as cholera, enteritis, dysentery, typhoid fever, etc. that plague vast regions in the developing countries, due to difficult access to clean drinking water. The various problems affecting natural ecosystems in this part of the globe, if not addressed properly, can result in unprecedented consequences for global biodiversity and the lives of millions of people who depend on the various services offered by those natural ecosystems.

Furthermore, the use of chemicals in ASM results in a massive release of liquid elemental mercury into the environment, with serious environmental consequences. The tailings produced from mining activity generally have high levels of mercury and uranium, and they are considered or classified as hazardous waste. Small-scale miners dump them directly into rivers and lakes. In these tailings, the concentrations of toxic heavy metals such as cadmium, zinc, and lead are generally 2 to 10 times higher than international standards. For example, in the Amazon basin, approximately 63% of measured mercury concentrations in the atmosphere were attributable to gold mining [72].

Finally, from a single biodiversity point of view, it appears very important to assess the affected areas and determine conservation priorities following international criteria, such as the IUCN Red List of species. Figure 18 and Figure 19 illustrate the degradation of the environment due to mining activities in the Kampene mining area. The degradation mainly affects soil, water, and forest cover.

(a) (b)

(c) (d)

Figure 18. Illustration of deforestation in the Kampene area by ((a) and (b)) settlements and ((c) and (d)) mining activities.

(a) (b)

(c) (d) (e)

Figure 19. Environmental degradation due to mining activities in the Kampene area in active artisanal mines ((a)-(d)) and abandoned mines becoming a lake (e).

7. Study Limitations

A potential limitation of this study is its reliance on medium-resolution (30 m) Landsat imagery for the land use and land cover analysis. While the Landsat archive is invaluable for its long-term temporal coverage, enabling a multi-decadal analysis, its spatial resolution may not fully capture the extent of certain fine-scale land cover changes. For instance, small-scale artisanal mining operations, which can be characterized by small, isolated pits and localized clearings, may be smaller than a single pixel. This can lead to them being missed or misclassified, an issue reflected in the higher omission and commission errors observed for the “Bare Soils and Mines” class, which covers smaller and heterogeneous areas. Similarly, highly fragmented agricultural plots, which are common in subsistence farming systems, may be challenging to distinguish accurately from the surrounding forest matrix, potentially leading to an underestimation of their total footprint. Future research could benefit from integrating higher-resolution satellite data or targeted analyses with unmanned aerial vehicles (UAVs) to conduct more granular assessments and validate the changes driven by these small-scale activities.

8. Conclusions

The Maniema Province is characterized by its forestry and mining activities, which have represented for decades one of the main activities in the Kivu region. The pressure on natural resources is high because of the extreme poverty of the population living around the mines and the lack of employment. People are obliged to look for ways of obtaining or increasing their financial income.

In this study, an evaluation of the artisanal mining impact on the land cover and land use change dynamics in the Pangi Territory in general and in the Kampene mining area in particular was conducted using the Support Vector Machine algorithm for LULC classification and post-classification change detection, and the Markov Chain Analysis for LULC prediction.

The results of this study indicate that:

1) The Primary Forest is the predominant LULC class in Pangi Territory.

2) There is a consistent decrease in forest cover from 1984 to 2020, with a deforestation rate of 83.12 km2 per year from 1984 to 2000 and 46 km2 per year from 2000 to 2020.

3) A deforestation of 73.92 km2 per year is predicted for the period from 2020 to 2040.

4) A total of 1920 km2 of forest cover was converted into agricultural land, which is equivalent to 82.7% of the total forest cover lost from 1984 to 2020.

Despite the considerable share of artisanal mining in forest loss, the study shows that the conversion of forests into agricultural areas is the primary driver of forest loss in Pangi Territory.

The techniques of sustainable agriculture should be advertised and enforced, and artisanal miners should work together in cooperatives to allow proper supervision and training on sustainable mining techniques. This will enable a change in this trend in the next 20 years; otherwise, the environment will suffer a loss of 1478 km2 as predicted for the Primary Forest.

Given the strong influence of proximity to roads on LULC change, land-use controls should prioritize road corridors. Establishing zoning buffers along new and upgraded roads could improve environmental protection and reduce the probability of LULC transitions.

Acknowledgements

The authors thank Professor Nindel Reinhardt of the Glauchau State Academy for proposing this study in the Maniema Province. The authors also thank the anonymous reviewers of the manuscript and the editor for their comments that improved the quality of the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest regarding the publication of this paper.

References

[1] Lee, D.R., Neves, B., Wiebe, K., Lipper, L. and Zurek, M. (2009) Rural Poverty and Natural Resources: Improving Access and Sustainable Management. ESA Working Paper, 09-03.
https://openknowledge.fao.org/server/api/core/bitstreams/0e8e24ea-89a6-4749-bd99-168db158b8b2/content
[2] Zakari, S., Imorou, I.T., Thomas, O.A.B., Djaouga, M. and Arouna, O. (2018) Application de la télédétection et du sig au suivi des formations végétales de la forêt classée des trois rivières au nord-est du bénin. European Scientific Journal, ESJ, 14, 450-469.[CrossRef
[3] Laurance, W.F. (1999) Reflections on the Tropical Deforestation Crisis. Biological Conservation, 91, 109-117.[CrossRef
[4] Laurance, W.F., Albernaz, A.K.M. and Costa, C.D. (2001) Is Deforestation Accelerating in the Brazilian Amazon? Environmental Conservation, 28, 305-311.[CrossRef
[5] Sodhi, N.S., Koh, L.P., Brook, B.W. and Ng, P.K.L. (2004) Southeast Asian Biodiversity: An Impending Disaster. Trends in Ecology & Evolution, 19, 654-660.[CrossRef] [PubMed]
[6] Sodhi, N.S. and Ehrlich, P.R. (2010) Conservation Biology for All. Oxford University Press, 399 p.
[7] Redmond, I. (2001) Coltan Boom, Gorilla Bust: The Impact of Coltan Mining on Gorillas and other Wildlife in Eastern DRC. A Report for the Dian Fossey Gorilla Fund Europe and the Born Free Foundation.
https://www.academia.edu/43511872/Coltan_Boom_Gorilla_Bust_The_Impact_of_Coltan_Mining_on_Gorillas_and_other_Wildlife_in_Eastern
[8] IUCN/PACO (International Union for the Conservation of the Nature and Natural Resources/Program on African Protected Area and Conservation) (2010) Parks and Reserves of Ghana: Management Effectiveness Assessment of Protected Areas.
[9] Nellemann, C., Redmond, I. and Refish, J. (2010) The Last Stand of the Gorilla—Environmental Crime and Conflict in the Congo Basin. A Rapid Assessment. United Nations Environment Programme, GRID-Arendal.
https://www.grida.no
[10] Mensah, A.K., Mahiri, I.O., Owusu, O., Mireku, O.D., Wireko, I. and Kissi, E.A. (2015) Environmental Impacts of Mining: A Study of Mining Communities in Ghana. Applied Ecology and Environmental Sciences, 3, 81-94.
[11] Mwitwa, J., German, L., Muimba-Kankolongo, A. and Puntodewo, A. (2012) Governance and Sustainability Challenges in Landscapes Shaped by Mining: Mining-Forestry Linkages and Impacts in the Copper Belt of Zambia and the DR Congo. Forest Policy and Economics, 25, 19-30.[CrossRef
[12] Chuhan-Pole, P., Dabalen, A.L. and Land, B.C. (2017). Mining in Africa: Are Local Communities Better Off? The World Bank. [Google Scholar] [CrossRef
[13] Ericsson, M. and Löf, O. (2019) Mining’s Contribution to National Economies between 1996 and 2016. Mineral Economics, 32, 223-250.[CrossRef
[14] Butsic, V., Baumann, M., Shortland, A., Walker, S. and Kuemmerle, T. (2015) Conservation and Conflict in the Democratic Republic of Congo: The Impacts of Warfare, Mining, and Protected Areas on Deforestation. Biological Conservation, 191, 266-273.[CrossRef
[15] International Alert (2010) The Role of the Exploitation of Natural Resources in Fuel-ling and Prolonging Crises in the Eastern DRC. 90 p.
[16] Dorleku, M.K., Nukpezah, D. and Carboo, D. (2018) Effects of Small-Scale Gold Mining on Heavy Metal Levels in Groundwater in the Lower Pra Basin of Ghana. Applied Water Science, 8, Article No. 126.[CrossRef
[17] Journal Officiel RDC (2018) Loi n°18/001 du 09 mars 2018 modifiant et complétant la Loi n° 007/2002 du 11 juillet 2002 portant Code minier. Présidence de la Ré-publique, Kinshasa, DR Congo, 60 p.
[18] Journal Officiel RDC (2018) Décret N° 038/2003 du 26 Mars 2003 portant Règlement Minier tel que modifié et completé par le Décret N°18/024 du 08 Juin 2018. Primature, Gouvernement de la République Démocratique du Congo, Kinshasa, DR Congo, 584 p.
[19] Malaviya, S., Munsi, M., Oinam, G. and Joshi, P.K. (2009) Landscape Approach for Quantifying Land Use Land Cover Change (1972-2006) and Habitat Diversity in a Mining Area in Central India (Bokaro, Jharkhand). Environmental Monitoring and Assessment, 170, 215-229. [Google Scholar] [CrossRef] [PubMed]
[20] Chandra, P.G. (2012) Remote Sensing of Land Use and Land Cover: Principles and Applications, CRC Press, 477 p.
[21] Bagalwa, M., Majaliwa, J., Kansiime, F., Bashwira, S., Tenywa, M., Karume, K., et al. (2016) Land Use and Land Cover Change Detection in Rural Areas of River Lwiro Micro-Catchment, Lake Kivu, Democratic Republic of Congo. Journal of Scientific Research and Reports, 9, 1-10.[CrossRef
[22] Liping, C., Yujun, S. and Saeed, S. (2018) Monitoring and Predicting Land Use and Land Cover Changes Using Remote Sensing and GIS Techniques—A Case Study of a Hilly Area, Jiangle, China. PLOS ONE, 13, e0200493.[CrossRef] [PubMed]
[23] Areendran, G., Rao, P., Raj, K., Mazumdar, S. and Puri, K. (2013) Land Use/Land Cover Change Dynamics Analysis in Mining Areas of Singrauli District in Madhya Pradesh, India. Tropical Ecology, 54, 239-250.
[24] Basommi, P.L., Guan, Q. and Cheng, D. (2015) Exploring Land Use and Land Cover Change in Themining Areas of Wa East District, Ghana Usingsatellite Imagery. Open Geosciences, 7, 618-626.[CrossRef
[25] LaJeunesse Connette, K., Connette, G., Bernd, A., Phyo, P., Aung, K., Tun, Y., et al. (2016) Assessment of Mining Extent and Expansion in Myanmar Based on Freely-Available Satellite Imagery. Remote Sensing, 8, Article 912.[CrossRef
[26] Javed, A. and Khan, I. (2012) Land Use/Land Cover Change Due to Mining Activities in Singrauli Industrial Belt, Madhya Pradesh Using Remote Sensing and GIS. Journal of Environmental Research and Development, 6, 834-843.
[27] Lu, D., Mausel, P., Brondízio, E. and Moran, E. (2004) Change Detection Techniques. International Journal of Remote Sensing, 25, 2365-2401.[CrossRef
[28] Xiaolu, S. and Bo, C. (2011) Change Detection Using Change Vector Analysis from Landsat TM Images in Wuhan. Procedia Environmental Sciences, 11, 238-244.[CrossRef
[29] Myint, S.W. and Wang, L. (2006) Multicriteria Decision Approach for Land Use Land Cover Change Using Markov Chain Analysis and a Cellular Automata Approach. Canadian Journal of Remote Sensing, 32, 390-404.[CrossRef
[30] Huang, W., Liu, H., Luan, Q., Jiang, Q., Liu, J. and Liu, H. (2008) Detection and Pre-diction of Land Use Change in Beijing Based on REMOTE sensing and GIS. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 37, 75-81.
[31] Wulder, M.A., Masek, J.G., Cohen, W.B., Loveland, T.R. and Woodcock, C.E. (2012) Opening the Archive: How Free Data Has Enabled the Science and Monitoring Promise of Landsat. Remote Sensing of Environment, 122, 2-10.[CrossRef
[32] Fernandez-Alonso, M., Kampata, D., Mupande, J.F., Dewaele, S., Laghmouch, M., Baudet, D., Lahogue, P., Badosa, T., Kalenga, H., Onya, F., Mawaya, P., Mwanza, B., Mashagiro, H., Kanda-Nkula, V., Luamba, M., Mpoyi, J., Decrée, S. and Lambert, A. (2015) Carte Géologique de la République Démocratique du Congo au 1/2.500.000, Notice explicative—Ministère des Mines, République Démocratique du Congo, 106 p.
[33] Xie, Y., Sha, Z. and Yu, M. (2008) Remote Sensing Imagery in Vegetation Mapping: A Review. Journal of Plant Ecology, 1, 9-23.[CrossRef
[34] Bernstein, L.S., Jin, X., Gregor, B. and Adler-Golden, S. (2012) Quick Atmospheric Correction Code: Algorithm Description and Recent Upgrades. Optical Engineering, 51, Article ID: 111719.[CrossRef
[35] Welch, R. and Ehlers, W. (1987) Merging Multiresolution SPOT HRV and Landsat TM Data. Photogrammetric Engineering & Remote Sensing, 53, 301-303.
[36] Bey, A., Diaz, A.S.P., Pekkarinen, A., Patriarca, C., Maniatis, D., Weil, D., Mollicone, D., Marchi, G., Niskala, J., Rezende, M. and Ricci, S. (2015) Collect Earth User Manual: A Guide to Monitoring Land Use Change and Deforestation with Free and Open-Source Software. FAO, 24 p.
[37] SPIAF (Service Permanent d’Inventaire et d’Aménagement Forestier) (2007) Normes de stratification forestière République Démocratique du Congo. Ministère de l’environnement et développement durable.
[38] Bahari, N.I.S., Ahmad, A. and Aboobaider, B.M. (2014) Application of Support Vector Machine for Classification of Multispectral Data. IOP Conference Series: Earth and Environmental Science, 20, Article ID: 012038.[CrossRef
[39] Candade, N. and Dixon, B. (2004) Multispectral Classification of Landsat Images: A Comparison of Support Vector Machine and Neural Network Classifiers. 2004 ASPRS Annual Meeting Proceedings, Denver, 28 May 2004, 1-11.
[40] Mountrakis, G., Im, J. and Ogole, C. (2011) Support Vector Machines in Remote Sensing: A Review. ISPRS Journal of Photogrammetry and Remote Sensing, 66, 247-259.[CrossRef
[41] Wu, T.F., Lin, C.J. and Weng, R.C. (2004) Probability Estimates for Multi-Class Classification by Pairwise Coupling. Journal of Machine Learning Research, 5, 975-1005.
[42] Chang, C. and Lin, C. (2011) LIBSVM: A Library for Support Vector Machines. ACM Transactions on Intelligent Systems and Technology, 2, 1-27.[CrossRef
[43] Hsu, C.W., Chang, C.C. and Lin, C.J. (2010) A Practical Guide to Support Vector Classification.
http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf
[44] Vapnik, V. (1995) The Nature of Statistical Learning Theory. Springer Verlag.
[45] Eastman, J.R. (2016) TerrSet Manual. Clark Labs.
[46] Karl, J. and Axel, A. (2013) Change Detection. Landscape Toolbox Wiki.
https://landscapetoolbox.org/remote-sensing-methods/change-detection/
[47] Amoah, M.K.M. and Gorsevski, P.V. (2025) Predicting Land Use Land Cover Changes and Impact on Urban Wetlands Using Cellular Automata and Artificial Neural Networks Approach, a Case Study in Greater Accra, Ghana. Scientific African, 28, e02767.[CrossRef
[48] Simon, O., Lyimo, J. and Yamungu, N. (2023) Land Use and Cover Change in Dar Es Salaam Metropolitan City: Satellite Data and Ca-Markov Chain Analysis. GeoJournal, 88, 6119-6136.[CrossRef
[49] Grzegorz, S., Dong, B. and Christine, W. (2009) Les images satellites Spot multi-dates et la métrique spatiale dans l’étude du changement urbain et sur urbain, le cas de la basse vallée de la Bruche (Bas Rhin, France). Cybergeo: European Journal of Geography, Systèmes, Modélisation, Géostatistiques, 22 p.
http://www.cybergeo.eu/index21995.html
[50] Mustapha, M.R., Lim, H.S. and MatJafri, M.Z. (2012) Analysis of Land Cover Classification in Arid Environment: A Comparison Performance of Four Classifiers. In: Chemin, Y., Ed., Remote Sensing of Planet Earth, InTech, 117-142.[CrossRef
[51] OCHA (2016) Democratic Republic of Congo: Internally Displaced People and Returnees.
[52] Halmy, M.W.A., Gessler, P.E., Hicke, J.A. and Salem, B.B. (2015) Land Use/Land Cover Change Detection and Prediction in the North-Western Coastal Desert of Egypt Using Markov-CA. Applied Geography, 63, 101-112.[CrossRef
[53] Geist, H.J. and Lambin, E.F. (2002) Proximate Causes and Underlying Driving Forces of Tropical Deforestation. BioScience, 52, 143-150.[CrossRef
[54] Rudel, T.K., Defries, R., Asner, G.P. and Laurance, W.F. (2009) Changing Drivers of Deforestation and New Opportunities for Conservation. Conservation Biology, 23, 1396-1405.[CrossRef] [PubMed]
[55] Kissinger, G., Herold, M. and De Sy, V. (2012) Drivers of Deforestation and Forest Degradation—A Synthesis Report for REDD+ Policymakers. Lexeme Consulting.
[56] Gillet, P., Vermeulen, C., Feintrenie, L., Dessard, H. and Garcia, C. (2016) Quelles sont les causes de la déforestation dans le bassin du Congo? Synthèse bibliographique et études de cas. BASE, 20, 183-194.[CrossRef
[57] Mahmoud, M.I., Campbell, M.J., Sloan, S., Alamgir, M. and Laurance, W.F. (2019) Land-Cover Change Threatens Tropical Forests and Biodiversity in the Littoral Region, Cameroon. Oryx, 54, 882-891.[CrossRef
[58] Tieguhong, J.C. (2009) Impacts of Artisanal Gold and Diamond Mining on Livelihoods and the Environment in the Sangha Tri-National Park Landscape. Centre pour la recherche forestière internationale.
[59] Yuh, Y.G., Dongmo, Z.N., N’Goran, P.K., Ekodeck, H., Mengamenya, A., Kuehl, H., et al. (2019) Effects of Land Cover Change on Great Apes Distribution at the Lobéké National Park and Its Surrounding Forest Management Units, South-East Cameroon. A 13 Year Time Series Analysis. Scientific Reports, 9, Article No. 1445.[CrossRef] [PubMed]
[60] Forrest, J.L., Sanderson, E.W., Wallace, R., Lazzo, T.M.S., Cerveró, L.H.G. and Coppolillo, P. (2008) Patterns of Land Cover Change in and around Madidi National Park, Bolivia. Biotropica, 40, 285-294.[CrossRef
[61] MECNT (Ministère de l’Environnement, Conservation de la Nature et Tourisme) (2012) Synthèse des études sur les causes de la déforestation et de la dégradation des forêts en République Démocratique du Congo
[62] Center for International Forestry Research (CIFOR) (2012) The Congo Basin: The State of the Forrest, Kinshasa, D.R. Congo.
[63] Potapov, P., Hansen, M.C., Laestadius, L., Turubanova, S., Yaroshenko, A., Thies, C., et al. (2017) The Last Frontiers of Wilderness: Tracking Loss of Intact Forest Landscapes from 2000 to 2013. Science Advances, 3, e1600821.[CrossRef] [PubMed]
[64] Zhuravleva, I., Turubanova, S., Potapov, P., Hansen, M., Tyukavina, A., Minnemeyer, S., et al. (2013) Satellite-Based Primary Forest Degradation Assessment in the Democratic Republic of the Congo, 2000-2010. Environmental Research Letters, 8, Article ID: 024034.[CrossRef
[65] Nackoney, J., Molinario, G., Potapov, P., Turubanova, S., Hansen, M.C. and Furuichi, T. (2014) Impacts of Civil Conflict on Primary Forest Habitat in Northern Democratic Republic of the Congo, 1990-2010. Biological Conservation, 170, 321-328.[CrossRef
[66] Hobbs, R.J. (1993) Effects of Landscape Fragmentation on Ecosystem Processes in the Western Australian Wheatbelt. Biological Conservation, 64, 193-201.[CrossRef
[67] Bennett, A.F. (1999) Linkages in the Landscape. The Role of Corridors and Connectivity in Wildlife Conservation. IUCN-The World Conservation Union.
[68] McGarigal, K. and Cushman, S.A. (2002) Comparative Evaluation of Experimental Approaches to the Study of Habitat Fragmentation Effects. Ecological Applications, 12, 335-345.[CrossRef
[69] Fahrig, L. (2003) Effects of Habitat Fragmentation on Biodiversity. Annual Review of Ecology, Evolution, and Systematics, 34, 487-515.[CrossRef
[70] Hobbs, R.J. and Yates, C.J. (2003) Impacts of Ecosystem Fragmentation on Plant Populations: Generalising the Idiosyncratic. Australian Journal of Botany, 51, 471-488.[CrossRef
[71] Falk, D.A., Palmer, M.A. and Zedler, J.B. (2006) Foundations of Restoration Ecology. Island Press.
[72] Campbell, N. and Reece, J. (2007) Biologie (7ème edition). Editions du Renouveau Pédagogique Inc., 1334 p.

Copyright © 2025 by authors and Scientific Research Publishing Inc.

Creative Commons License

This work and the related PDF file are licensed under a Creative Commons Attribution 4.0 International License.