Browsing by Author "Mulabbi, Andrew"

Now showing 1 - 7 of 7
  • Thumbnail Image
    Item
    Assessment of landslide susceptibility and settlement exposure via geospatial techniques in Bulambuli distrcit, Eastern Uganda
    (Environmental Research Institute, 2025-10-15) Mulabbi, Andrew; Esagu, John Calvin; Gertrude, Akello; Turyahabwe, Remigio
    Landslide susceptibility is a significant concern in Elgon County, Uganda, particularly during the rainy season. This vulnerability is attributable to several factors, including steep slopes, fertile soils, and dense settlements on volcanic ridges. Landslide susceptibility maps are important in mitigating the risk particularly at the local level. The objectives of this study were 1) to model landslide susceptibility via an interpretable machine-learning model, 2) to identify the most influential factors for landslide susceptibility in the study area, and 3) to assess the exposure of settlements to landslide risk. This study employed the XGBoost model trained on nine conditioning factors via GIS data. Exposure analysis was performed through the zonal statistics and spatial overlay of the landslide susceptibility map with the settlement footprint data and classified into four risk exposure classes. The results show that the XGBoost model attained an AUC of 95.2%, indicating its precision. The results further revealed that approximately 50% of the slopes are susceptible to landslides and that 76% of the settlements in the study area are highly exposed to landslide risk. Bulugunya, Sisiyi, Lusha, and Buginyanya subcounties located on the middle slopes are the most susceptible areas in Elgon County and have relatively high settlement exposure because of the overlap of dense settlements with unstable terrain. The SHAP analysis identified slope, elevation, and the NDVI as the key influencing factors of susceptibility. This study highlights the importance of conducting detailed, local-scale landslide susceptibility and risk exposure mapping as necessary for risk and vulnerability assessment. The generation of such maps has the potential to inform land-use planning and risk-reduction strategies, thus offering significant advantages over regional models. Furthermore, by interpreting the XGBoost model, this study provides valuable insights into the decision-making processes of machine learning models, promoting their practical application in designing appropriate disaster mitigation plans.
  • Thumbnail Image
    Item
    Automatic landslide mapping with interpretable attention-based convolutional neural networks using remote sensing data
    (Association of Geoinformatics Technology, 2025-08-02) Mulabbi, Andrew; Danoedoro, Projo; Samodra, Guruh
    Landslide mapping plays a vital role in disaster management by providing essential information that can help decision making on mitigation and early warning strategies. However, existing automated methods often lack interpretability and miss crucial details, which limit their practical utility. This study addresses these limitations by introducing a novel Spatial Attention U-Net that leverages human visual attention to improve landslide detection and interpretability. Our proposed method integrates spatial attention modules throughout the U-Net's encoder and decoder paths, enabling the model to focus on critical image features for landslide identification. The model is trained and evaluated using a combination of high-resolution Pleiades RGB imagery, Brightness Index, and slope data. The model’s performance was evaluated using the F-1 score, precision, recall, and intersection over Union (IoU). The findings demonstrate that the Spatial Attention U-Net outperforms baseline models (U-Net, Squeeze-and-Excitation U-Net, and Channel-wise Attention), achieving F-1 scores of 73% and 79% on the testing and benchmark datasets, respectively. When applied to the inference/hold-out area, all the attention-based models outperformed the standard U-net, missing only three landslide events compared to five missed by the baseline model. Furthermore, the saliency maps reveal that the models focus on diverse regions of saliency, including edges, textures, tone, and brightness. The spatial attention U-net primarily highlights landslide edges (terrain discontinuities), while the baseline models use a mix of edges, texture, tone, and brightness. The results also indicate that dual-path attention does not lead to significant improvement in model accuracy. This approach offers a powerful tool for rapid and automated landslide mapping, indicating areas of saliency that can aid data annotation process by paying more attention to landslide object boundaries. The model interpretability further facilitates the creation of landslide inventories, especially in regions with limited ground truth data.
  • Thumbnail Image
    Item
    Determinants of fruit tree adoption as a climate change adaptation strategy amongst smallholder farmers in Lake Kyoga Basin: A Case study of Budaka District, Eastern Uganda
    (Wiley, 2025-07-10) Wambede, Nabalegwa M.; Kiconco, Milliam; Ewongu, Denis; Mulabbi, Andrew; Tweheyo, Robert; Mukisa, Geoffrey
    This study investigated the socioeconomic determinants of fruit tree adoption amongst smallholder farmers in Budaka District, Eastern Uganda. Specific objectives included describing the characteristics of fruit tree gardens, mapping their spatial distribution, and analysing socioeconomic factors influencing adoption. This study is one of the first empirical studies in agroforestry to relate socioeconomic factors in Eastern Uganda to the spatial distribution of fruit trees. The study employed a combined approach incorporating GIS-based spatial mapping and socioeconomic analysis. A cross-sectional design was employed, with data collected from 276 randomly selected farmers, key informants, and focus groups. GIS was used to visualise the spatial patterns and descriptive statistics, and chi-square tests were applied to identify differences between adopters and nonadopters. Results indicated that fruit farming is predominantly undertaken by males aged 40 and above. Fruit tree distribution is concentrated in the north and northwest, grown on small holdings averaging 0.5 acres with 10–40 trees. Chi-square tests confirmed significant differences in age, labour type, farm size, and income between adopters and nonadopters, whilst there were no significant differences in gender, family size, and access to credit. Policy interventions should expand youth- and gender-inclusive extension services that support climate resilience and sustainable fruit tree farming, and address land tenure limitations to increase adoption.
  • Thumbnail Image
    Item
    Factors affecting adoption of climate change adaptation strategies by small holder farmers in mountain and lowland agro-ecological zones of Eastern Uganda
    (Universitas Muhammadiyah Surakarta, 2022-12-05) Turyahabwe, Remigio; Turybanawe, Gumisiriza; Asaba, Joyfred; Mulabbi, Andrew; Geofrey, Geoffrey
    Several challenges confront farmers in tropical rural areas, but climate change can only be overcome by adopting climate change resilience strategies. This study assessed the factors affecting the adoption of strat-egies to enhance climate change resilience in the Muyembe sub-county, Bulambuli district, Uganda. We used questionnaires, interviews, focused group discussions, and field observations to collect the required data, which was analysed using basic descriptive statistics and a logistic regression model. The results indi-cate that the dominant climate change resilience strategies adopted in the study were soil/water conservation (65%), drought-resistant crop varieties (59.4%), and irrigation (55.6). In addition, the logistic regression indicated that gender and family size were the most important factors influencing the adoption of climate change resilience strategies with coefficients -0.86 and P<0.05, and0.18 and P<0.05, respectively. On the other hand, financial constraints and adulteration of farm inputs were the dominant barriers to adoption most farmers with 93.4% and 74%, respectively. We concluded that many farmers remain locked in indigenous practices that have made them vulnerable to climate change effects characterized by low yields, crop failure, low incomes, poverty, and food insecurity. We recommended that government should support the adaptation strategies to climate change by the smallholder farmers technically by providing both ground and surface water irrigation facilities and financially by providing agricultural loans as well as focusing on promoting awareness and advancing education on climate change to farmers through knowledge and skill sharing plat-forms such as training, conferences, and seminars.
  • Thumbnail Image
    Item
    Factors affecting the adoption of soil and water conservation practices by small-holder farmers in Muyembe Sub-County, Eastern Uganda
    (University of Ghana, 2022-03-14) Turyahabwe, Remigio; Wambede, Nabalegwa Muhamud; Asaba, Joyfred; Mulabbi, Andrew; Turyabanawe, Loy Gumisiriza
    Farmers in tropical rural areas are confronted with several challenges but outstanding among these challenges is soil degradation arising from soil erosion. This study involved identifying the dominant soil and water conservation practices and assessing the factors affecting their adoption in the Muyembe sub-county, Eastern Uganda. A total of 500 respondents were used to obtain primary data. As the study adopted a crosssectional design, we used questionnaires, interviews, focus group discussions and field observations to collect the required data. Data were analyzed using descriptive statistics and the non-parametric (Chi-square) test. The results indicated that the dominant soil and water conservation practices adopted in the study area were, contour cropping (77%), mixed cropping (59% and crop rotation (51%). The remaining five practices had less than a 50% adoption rate. The chi-square test revealed that the age and gender of the farmers had a significant association with the levels of the adoption of soil and water conservation practices among farmers at P<0.001. We concluded that the adoption of soil and water conservation practices was low, which left the majority of farmers vulnerable to soil erosion effects such as low yields and crop failure. We recommend that stakeholders who work on soil and water conservation programs use model farmers in the area to educate and demonstrate the importance of soil and water conservation practices to other farmers.
  • Thumbnail Image
    Item
    Flood inundation and damage assessment of the degraded Semliki River plains using SAR data, Google Earth Engine, and GIS techniques.
    (Universitas Brawijaya, 2025-06-21) Mulabbi, Andrew; Esagu, John Calvin; Gertrude, Akello; Remigio, Turyahabwe
    The Semliki River valley in Ntoroko district has experienced devastating annual floods since 2019. Recurrent floods in Ntoroko District have displaced thousands and devastated pasturelands, disrupting livelihoods. Therefore, rapid assessment of flooded areas is crucial for developing effective mitigation strategies, disaster preparedness plans, and proactive policies to enhance resilience and mitigate the impact of future flood events. This study introduced a combined approach using Synthetic Aperture Radar (SAR) imagery and a digital elevation model (DEM) to map flood extent, depth, and building exposure in the Semliki Valley. Using Sentinel-1 SAR images taken both before and during the flood, combined with the ALOS PALSAR DEM, inundated areas and flood depths were determined, based on thresholding the SAR backscatter of the VH polarisation images. The flood extent maps were generated using Google Earth Engine and GIS techniques to create depth maps by subtracting the surface elevation from the height/surface of the flood waters. Building exposure and impact analysis for two flood events was ascertained through spatial join and overlay. The results showed that the 2023 flood event inundated approximately 1,968 hectares, including 1,553 hectares of pastureland and 74 buildings, while the 2024 event covered 1,139 hectares, equally inundating 1,050 hectares of pastureland and 54 buildings. Further analysis revealed that despite the smaller extent, the 2024 flood event caused a severe impact on the buildings compared to the 2023 flood disaster.
  • Thumbnail Image
    Item
    Integrating interpretable artificial neural networks, geomorphic plausibility and transfer learning for landslide susceptibility mapping: a case study of Pacitan, East Java
    (Springer Nature, 2025-10-27) Mulabbi, Andrew; Danoedoro, Projo; Samodra, Guruh
    Landslides are recurring natural hazards that continue to cause widespread damage to agricultural fields, vital infrastructures, and human lives globally. The impact is more severe in data-scarce regions where insufficient landslide inventory data limit accurate and timely landslide susceptibility assessment. This study integrates artificial neural networks, transfer learning, interpretable machine learning, and geomorphic plausibility to address the challenge of accurate landslide susceptibility mapping in data-scarce regions. Using the Pacitan regency as a case study, we assessed landslide susceptibility via artificial neural networks and transfer learning models. A neural network was first trained from a source area with a large dataset and then transferred through fine-tuning to another area with limited inventory data for mapping landslide susceptibility. A baseline model trained only on the limited target area data achieved an AUC test score of 0.83, the source-trained model had 0.88, while the transfer learning-based model achieved an AUC score of 0.97, confirming the improved performance and strong potential for transferability. Model interpretation using SHAP values, permutation importance, interaction strength, and partial dependence plots revealed slope, elevation, aspect, and distance to the stream as the most influential, while curvature and terrain indices were less impactful directly but contributed through interactions with other features. A qualitative assessment of geomorphic plausibility showed that the susceptibility patterns generally matched the behaviour of the terrain. Zones of high susceptibility were concentrated in the mid-slopes and at concave-convex transitions. Low susceptibility is observed in the steep rocky slope (> 40 °). These findings highlight the potential of interpretable machine learning, transfer learning, and geomorphic plausibility to produce physically meaningful susceptibility maps. This approach may be applied globally to produce rapid, interpretable, and generalisable landslide susceptibility maps in data-constrained regions, thereby contributing to more effective disaster mitigation and spatial planning.