Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, Zambia
Wetlands are among the most productive natural ecosystems globally, providing crucial ecosystem services to people. Regrettably, a substantial 64 % –71 % of wetlands have been lost worldwide since 1900, mainly due to changes in land use and land cover (LULC). This issue is not unique to Zambia'...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-01-01
|
| Series: | Environmental Challenges |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2667010024000325 |
| _version_ | 1797292877944979456 |
|---|---|
| author | Misheck Lesa Chundu Kawawa Banda Chisanga Lyoba Greyfold Tembo Henry M. Sichingabula Imasiku A. Nyambe |
| author_facet | Misheck Lesa Chundu Kawawa Banda Chisanga Lyoba Greyfold Tembo Henry M. Sichingabula Imasiku A. Nyambe |
| author_sort | Misheck Lesa Chundu |
| collection | DOAJ |
| description | Wetlands are among the most productive natural ecosystems globally, providing crucial ecosystem services to people. Regrettably, a substantial 64 % –71 % of wetlands have been lost worldwide since 1900, mainly due to changes in land use and land cover (LULC). This issue is not unique to Zambia's Bangweulu Wetland System (BWS), which faces similar challenges. However, there is limited information about the LULC changes in BWS. Furthermore, finding accurate and cost-effective methods to understand LULC dynamics is complicated by the multitude of available techniques for LULC classification. Non-parametric methods like Machine Learning (ML) offer greater accuracy, but different ML models come with distinct strengths and weaknesses. Combining multiple models has the potential to create a more precise LULC classification model. Open-source software like QGIS and spatial data like Landsat also play a significant role in this endeavour. The primary objective of this study was to enhance the accuracy of modeling LULC changes in wetland areas. Six ML models: Support Vector Machine (SVM), Naive Bayes (NB), Decision Tree (DT), Artificial Neural Network (ANN), Random Forest (RF), and K-Nearest Neighbour (KNN) were used for LULC image classification of Landsat 8 (2020 image) and Landsat 5 (1990, 2000, and 2010 images) in QGIS. Four models: SVM, NB, DT, and KNN, performed better than the other models. Consequently, The Quad (4) hybrid model was created by fusing the maps from these four models with the highest performance. Results revealed that the fusion of the four classified maps of the SVM, NB, DT, and KNN (Quad hybrid model) showcased superior performance compared to the individual models with Kappa Index scores of 0.87, 0.72, 0.84 and 0.87 for the years 1990, 2000, 2010 and 2020, respectively. The analysis of the LULC changes from 1990 to 2020 showed a yearly decline of -1.17 %, -1.01 %, and -0.12 % in forest, grassland, and water body coverage, respectively. In contrast, built-up areas and cropland increased at rates of 1.70 % and 2.70 %, respectively. This study underscores the consistent growth of cropland and built-up areas from 1990 to 2020, alongside the reduction of forest cover and grassland. Although the water body experienced a gradual decrease over this period, the decline was minimal. Long-term monitoring will be essential for evaluating the success of interventions, guiding conservation efforts, mitigating negative impacts on the wetland ecosystem, and determining whether the reduction in water bodies is a sustained trend or a short-term phenomenon. |
| first_indexed | 2024-03-07T20:03:56Z |
| format | Article |
| id | doaj.art-3a48f43f616c4f169076e80a4d726aa6 |
| institution | Directory Open Access Journal |
| issn | 2667-0100 |
| language | English |
| last_indexed | 2024-03-07T20:03:56Z |
| publishDate | 2024-01-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Environmental Challenges |
| spelling | doaj.art-3a48f43f616c4f169076e80a4d726aa62024-02-28T05:14:30ZengElsevierEnvironmental Challenges2667-01002024-01-0114100866Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, ZambiaMisheck Lesa Chundu0Kawawa Banda1Chisanga Lyoba2Greyfold Tembo3Henry M. Sichingabula4Imasiku A. Nyambe5Corresponding author.; Department of Geology, The University of Zambia Integrated Water Resources Management Centre, School of Mines, P.O. Box 32379, Lusaka, ZambiaDepartment of Geology, The University of Zambia Integrated Water Resources Management Centre, School of Mines, P.O. Box 32379, Lusaka, ZambiaDepartment of Geology, The University of Zambia Integrated Water Resources Management Centre, School of Mines, P.O. Box 32379, Lusaka, ZambiaDepartment of Geology, The University of Zambia Integrated Water Resources Management Centre, School of Mines, P.O. Box 32379, Lusaka, ZambiaDepartment of Geology, The University of Zambia Integrated Water Resources Management Centre, School of Mines, P.O. Box 32379, Lusaka, ZambiaDepartment of Geology, The University of Zambia Integrated Water Resources Management Centre, School of Mines, P.O. Box 32379, Lusaka, ZambiaWetlands are among the most productive natural ecosystems globally, providing crucial ecosystem services to people. Regrettably, a substantial 64 % –71 % of wetlands have been lost worldwide since 1900, mainly due to changes in land use and land cover (LULC). This issue is not unique to Zambia's Bangweulu Wetland System (BWS), which faces similar challenges. However, there is limited information about the LULC changes in BWS. Furthermore, finding accurate and cost-effective methods to understand LULC dynamics is complicated by the multitude of available techniques for LULC classification. Non-parametric methods like Machine Learning (ML) offer greater accuracy, but different ML models come with distinct strengths and weaknesses. Combining multiple models has the potential to create a more precise LULC classification model. Open-source software like QGIS and spatial data like Landsat also play a significant role in this endeavour. The primary objective of this study was to enhance the accuracy of modeling LULC changes in wetland areas. Six ML models: Support Vector Machine (SVM), Naive Bayes (NB), Decision Tree (DT), Artificial Neural Network (ANN), Random Forest (RF), and K-Nearest Neighbour (KNN) were used for LULC image classification of Landsat 8 (2020 image) and Landsat 5 (1990, 2000, and 2010 images) in QGIS. Four models: SVM, NB, DT, and KNN, performed better than the other models. Consequently, The Quad (4) hybrid model was created by fusing the maps from these four models with the highest performance. Results revealed that the fusion of the four classified maps of the SVM, NB, DT, and KNN (Quad hybrid model) showcased superior performance compared to the individual models with Kappa Index scores of 0.87, 0.72, 0.84 and 0.87 for the years 1990, 2000, 2010 and 2020, respectively. The analysis of the LULC changes from 1990 to 2020 showed a yearly decline of -1.17 %, -1.01 %, and -0.12 % in forest, grassland, and water body coverage, respectively. In contrast, built-up areas and cropland increased at rates of 1.70 % and 2.70 %, respectively. This study underscores the consistent growth of cropland and built-up areas from 1990 to 2020, alongside the reduction of forest cover and grassland. Although the water body experienced a gradual decrease over this period, the decline was minimal. Long-term monitoring will be essential for evaluating the success of interventions, guiding conservation efforts, mitigating negative impacts on the wetland ecosystem, and determining whether the reduction in water bodies is a sustained trend or a short-term phenomenon.http://www.sciencedirect.com/science/article/pii/S2667010024000325BangweuluLand use/land cover changeLandsatMachine learningRemote sensingWetland |
| spellingShingle | Misheck Lesa Chundu Kawawa Banda Chisanga Lyoba Greyfold Tembo Henry M. Sichingabula Imasiku A. Nyambe Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, Zambia Environmental Challenges Bangweulu Land use/land cover change Landsat Machine learning Remote sensing Wetland |
| title | Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, Zambia |
| title_full | Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, Zambia |
| title_fullStr | Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, Zambia |
| title_full_unstemmed | Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, Zambia |
| title_short | Modeling land use/land cover changes using quad hybrid machine learning model in Bangweulu wetland and surrounding areas, Zambia |
| title_sort | modeling land use land cover changes using quad hybrid machine learning model in bangweulu wetland and surrounding areas zambia |
| topic | Bangweulu Land use/land cover change Landsat Machine learning Remote sensing Wetland |
| url | http://www.sciencedirect.com/science/article/pii/S2667010024000325 |
| work_keys_str_mv | AT mishecklesachundu modelinglanduselandcoverchangesusingquadhybridmachinelearningmodelinbangweuluwetlandandsurroundingareaszambia AT kawawabanda modelinglanduselandcoverchangesusingquadhybridmachinelearningmodelinbangweuluwetlandandsurroundingareaszambia AT chisangalyoba modelinglanduselandcoverchangesusingquadhybridmachinelearningmodelinbangweuluwetlandandsurroundingareaszambia AT greyfoldtembo modelinglanduselandcoverchangesusingquadhybridmachinelearningmodelinbangweuluwetlandandsurroundingareaszambia AT henrymsichingabula modelinglanduselandcoverchangesusingquadhybridmachinelearningmodelinbangweuluwetlandandsurroundingareaszambia AT imasikuanyambe modelinglanduselandcoverchangesusingquadhybridmachinelearningmodelinbangweuluwetlandandsurroundingareaszambia |