International Journal of Geoinformatics

Grid-Based Semantic Enrichment of Human Mobility Data for Urban Spatial Patterns Profiling: Hotspot and Service Gap Area Detection in Kota Yogyakarta, Indonesia

L. Iswari — 2026-09-29

Understanding urban spatial patterns from human mobility data requires semantic context beyond raw trajectory analysis. Existing approaches face three key limitations: grid size selection relies on empirical judgment rather than systematic evaluation, semantic enrichment assigns a single dominant label without quantifying degrees of dominance, and hotspot detection relies solely on demand-side density without considering urban facility supply. This study proposes a grid-based semantic enrichment framework for urban spatial pattern detection, comprising four components: a multi-criteria grid-size evaluation framework, confidence-score-based semantic enrichment, supply-demand synthesis for hotspot identification, and service gap identification. The framework was applied to Kota Yogyakarta, Indonesia, using over 4.6 million trajectory points from 130,395 mobile users. A grid size of 500 m was identified as optimal across ten evaluated resolutions. Supply-demand synthesis was applied to 342 grid cells. Results show that only 10.5% of cells qualify as full hotspots, concentrated in the central-eastern urban area and dominated by functionally mixed zones. Unmet demand is the primary service gap type, covering 24% of the study area, more than four times the proportion of underutilized supply (5.6%), suggesting the presence of informal activity not captured in spatial datasets. These findings demonstrate that the proposed framework provides a systematic methodology for extracting meaningful urban insights from large-scale human mobility data, with practical implications for evidence-based urban planning in mid-sized Indonesian cities.

Categorizing Toll Gates using Data of Sentinel-1, Sentinel-2, Sentinel-5P, NOAA-20 Satellite, and Objects on Google Map: A Case Study in Indonesia

V.S. Moertini — 2026-09-29

This study proposes a framework to categorize toll gates based on the development level of their surrounding areas and localized economic activity profiles. The methodology integrates multi-source satellite imagery with geospatial features extracted from the Google Maps API. The localized economic context of 419 toll gates was analysed within a 1-km circular buffer zone. Within this radius, Sentinel-2 spectral bands (B2, B3, B4 and B8) extracted from Google Earth Engine (GEE) were classified into built-up and non-built-up pixels using a Random Forest algorithm to quantify the built-up area percentage. Gates with a built-up density of at least 40% were categorized as Class-1, representing infrastructure in highly developed regions. For the remaining gates, Sentinel-1 VV polarization, Sentinel-5P tropospheric NO₂ column density, and NOAA-20 VIIRS nighttime lights data were extracted via GEE, alongside point-of-interest (POI) counts from Google Maps. Following feature engineering across five initial variables, the backscatter coefficient (VV) and POI density were identified as the optimal features for clustering. Using these features, a k-Means algorithm grouped the remaining gates into Class-2 and Class-3 clusters, whose final socioeconomic profiles were interpreted using all variables. The classification yielded 297 Class-1 gates (highly developed areas with dense human and commercial activities), 51 Class-2 gates (moderate activity zones), and 71 Class-3 gates (low-to-moderate activity zones). To assess the long-term impacts of toll roads on regional economic growth, Class-3 gates warrant continuous observation, as their surrounding economic activity remains suboptimal. This research contributes to the fields of applied remote sensing and transportation infrastructure mapping by providing a cost-effective, automated methodology to periodically monitor spatiotemporal economic shifts around critical transport infrastructure.

National-Scale Mapping of Oil Palm Plantation Ages in Indonesia Using Global Canopy Height Data on Google Earth Engine

R. Hernawati — 2026-09-29

The oil palm (Elaeis guineensis) is one of the most strategic plantation commodities in Indonesia. However, spatially explicit information on plantation age remains limited, which hampers evidence-based decision making for productivity assessment, replanting prioritization, and sustainable plantation management. This study is to investigate the spatial distribution of oil palm age across Indonesia based on 10 m Global Canopy Height (GCH) product within the Google Earth Engine (GEE) platform. The methodology included canopy-height extraction and masking, age estimation using a height–age regression model, multi-site validation, and national-scale mapping. Canopy height obtained from GCH was compared with field measurements at the Asahan calibration site were strongly correlated (R² = 0.85). The age model was validated using planting-year records from five plantations in Sumatra and Kalimantan. At the Asahan calibration site, the model reproduced recorded ages with the suitable is R² = 0.88 and RMSE = 2.8 years. However, model reliability may vary depending on topography, weather conditions, plantation management, and oil palm variety. The model was applied to 11.62 million hectares of mapped oil palm area. The mapped plantations were dominated by mature stands (5–25 years, 43.7%), followed by young stands (0–5 years, 30.3%) and old stands (>25 years, 26.1%). Regional mapping shows that Sumatra contains a high density of old stands targeted for replanting, particularly in Jambi, where old stands account for 30.4% of the mapped area. In contrast, Central Kalimantan is dominated by young stands, which account for 40.0%. At the national level, the resulting age map provides a spatially consistent baseline for identifying replanting priorities, supporting productivity monitoring, improving plantation inventory, and informing sustainable oil palm governance across Indonesia.

Urban Growth Dynamics of Special Region of Yogyakarta, Indonesia (1994–2024): Population Trends and Expansion Prediction

I.N. Hidayati — 2026-09-29

The urban area of Yogyakarta (Indonesia) is observed to have experienced quite rapid development due to the changes in built-up land over the past three decades. The growth is motivated by population growth as well as increased educational activities and equitable access to infrastructure. Therefore, this study aims to understand the complex changes through two main objectives which include analyzing patterns of built-up land change from 1994 to 2025 and predicting the direction of urban development using Cellular Automata-Markov Chain (CA-Markov) method. Random Forest (RF) analysis was used to analyze the variations in land use and land cover (LULC) based on Landsat 5 TM, 7 ETM+, 8 OLI, and 9 OLI-2. These imageries were adopted to observe changes in LULC and the direction of urban development every 5 years. The results showed that NE direction experienced the most rapid development from 986 ha in 1994 to 7,648 ha in 2024. Logistic regression analysis further identified population as the main factor of land conversion (β = 1.673). Meanwhile, educational facilities had a large contribution to the pressure of vegetation change (β = −118.898), which may reflect the increasing demand for educational infrastructure accompanying population growth and urban expansion. The transportation network also showed that the city center and areas with high connectivity had reached a saturation point and the trend led to the tendency of new development shifting to peri-urban areas. Shannon entropy value reflected a change in urban development patterns with the growth observed to be concentrated at the start identified to have spread and become increasingly fragmented after 2014. Moreover, the simulations conducted using CA–Markov showed the possibility of future urban development continuing to move toward peri-urban areas in a northeast–northwest (NE-NW) direction. The combination of multitemporal Landsat, RF classification, regression analysis, and CA–Markov models produced a more comprehensive picture of urban development dynamics in Yogyakarta and also provides a scientific basis for policymakers regarding spatial utilization.

Fuzzy Tsukamoto-GIS Framework with Pixel-Level SAR Validation for Flood Susceptibility

J. Handoyo — 2026-09-29

Physical flood susceptibility in Cepu Subdistrict is shaped by low relief, proximity to the Bengawan Solo River, land-cover conditions, soil permeability, river density, and seasonal rainfall. This study applies a documented Fuzzy Tsukamoto-GIS workflow to estimate physical flood susceptibility across 17 villages using six parameters: slope, elevation, soil permeability, monthly rainfall, land use, and river density. Susceptibility scores range from 0.440 in Mulyorejo to 0.765 in Kapuan. Validation uses two approaches: pixel-level Sentinel-1 SAR imagery for the February 2022 flood event, and village-level rank comparison with BPBD flood records (2018-2023). Pixel-level validation shows strong performance with overall accuracy of 93.85%, ROC-AUC of 0.876, Spatial Kappa of 0.708, CSI of 59.08%, and FAR of 32.42%. Village-level Spearman correlation (ρ = 0.990, p < 0.001) confirms consistency with reported flood frequencies. Sensitivity analysis identifies monthly rainfall (SI = 0.28) and land-use susceptibility (SI = 0.24) as the most influential parameters. Cross-referencing BPBD records with media reports revealed potential reporting bias in Mulyorejo, highlighting the value of SAR-based validation. The pixel-level susceptibility map can support mitigation prioritization and spatial planning while providing a transparent basis for further calibration. This study assesses physical susceptibility rather than comprehensive vulnerability, as socioeconomic and coping capacity dimensions are not included.

High-Resolution Landslide Susceptibility Assessment along a Mountain Road Corridor in Carbonate Terrains Using Handheld SLAM LiDAR and GIS-Based AHP: Northeastern Iraq

R.S.M. Ali — 2026-09-29

Corridor landslides in road networks as well as rock slope instabilities on mountains roads represent hazards for transport systems in many areas, especially in karst terrains where slope behavior depends on discontinuities, bedding structure, and stress changes due to the removal of material during excavation. Slope instabilities on roads cuts occur often in northeast Iraq; nevertheless, their corridor vulnerability assessment is still limited due to the absence of topographic datasets for steep slopes. The present research has developed a methodology for evaluation of landslide risk along the Dokan-Topzawa road route (1.8 km) employing portable handheld Simultaneous Localization and Mapping (SLAM) Light Detection and Ranging (LiDAR) system coupled with Analytical Hierarchy Process (AHP) analysis utilizing GIS. LiDAR data (~ 400 million point cloud) have been processed to construct the Digital Elevation Model (DEM) from which nine conditioning factors related to geomorphology, geology, and environment have been extracted. The weight values of the factors have been obtained using AHP technique with an acceptable consistency ratio (CR = 0.02), which further have been combined into Landslide Susceptibility Index (LSI) with a weighted linear combination technique at 10 meter resolution. The resultant susceptibility map categorized the area into four susceptibility categories namely very low, low, moderate and high susceptibility zones. High susceptibility zones comprised 37.4% of the study area whereas zones having moderate and high susceptibility constituted 51.9% of the area. The landslide site investigations revealed that observed sites of landslides were situated in zones of moderate to high susceptibility. Profile interpretation through LiDAR and site investigations confirmed that factors influencing slope instability were mainly fractured limestone, unfavorable beddings towards the road, steepness of slope and seepage through discontinuities.

Flood Susceptibility Mapping Based on Supervised Machine Learning Algorithms with Integrated Uncertainty and Explainability Analysis: A Case Study in Quang Nam, Vietnam

T.T.H. Pham — 2026-09-29

Flood is among the natural hazards that have occurred with increasing frequency and have changed in complex ways, with serious impacts on people's lives. A flood susceptibility map the basis for zoning areas at risk of flooding became a necessary tool not only for managers but also for residents to adapt to the broader context of climate change. This study aims to present the use of a supervised machine learning model to create a flood susceptibility map of Quang Nam province (now Da Nang city, Vietnam). Four algorithms, including Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) were tested. Based on the evaluation metrics and uncertainty analysis, RF not only achieved the best predictive performance (AUC of 0.9418) but also demonstrated higher stability and reliability. An explainability analysis using Shapley additive explanation (SHAP) and Leave-One-Covariate-Out (LOCO) was conducted to assess the contribution of conditioning factors and interpret the RF model’s predictions. The result is a flood susceptibility map based on the RF model, with levels ranging from very low to very high. Low-lying plains in the lower reaches of the Vu Gia - Thu Bon River frequently flood and are the most heavily affected areas in Quang Nam. This map was validated with 2025 flood marks data, which served as an independent data source not involved in the model construction process. The results show that the majority of actual flooding points are concentrated in moderate-, high-, and very high-sensitivity areas, accounting for 98.2%, while points in low-sensitivity areas are negligible. Generally, the results will help disaster management agencies develop effective adaptation and risk mitigation policies at the commune level in flood-prone areas. Future studies should integrate flood susceptibility with exposure and vulnerability indicators to support a more comprehensive flood risk assessment

Optimizing Machine Learning Models for Landslide Susceptibility Mapping in Yen Bai Province, Vietnam

T.L. Tran — 2026-09-29

Landslide hazard assessment is crucial for managing and mitigating landslide risks. Landslide Susceptibility Mapping (LSM) provides a practical and cost-effective tool for zoning areas prone to landslides. LSM expresses in the form of a probability of landslide risk in each pixel. This study implements Machine Learning (ML) models to generate Susceptibility Maps in Van Yen (VY), making them applicable to diverse topographic regions, particularly in areas significantly affected by human activities. Landslides were mapped after landslide events occurred. The datasets were created by combining balanced numbers of landslide and non-landslide points with 17 contributing factors, including topographic, geological, hydrological, anthropogenic, and vegetation factors. To ensure all factors have the same range of values, landslide areas standardized using the Frequency Ratio (FR). The effectiveness of the method will be evaluated on the same dataset before and after applying FR. Four ML models Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), and Extreme Gradient Boosting (XGBoost) were trained in VY area, then validated on the dataset in Mu Cang Chai (MCC). Evaluation process was conducted using Accuracy Evaluation and Efficient Global Optimization (EGO). The results indicate that RF and XGBoost achieved the highest and most consistent performance, with the highest learning capacity. However, according to the statistical evaluation, Geologic factors have low influence on classification, suggesting the need for including other critical geologic factors. The study can act as future framework for susceptibility analysis on python environment.

Fusing Sentinel-2 and Sentinel-1 Data in Google Earth Engine for Road Infrastructure Mapping in Data-Scarce and Conflict Environments

S.P.A. Mali — 2026-09-29

Reliable road network information is a prerequisite for infrastructure planning, humanitarian operations, and disaster response; however, such information often remains incomplete or outdated in conflict-affected and data-scarce regions. In Juba County, South Sudan, decades of conflict, institutional limitations, and environmental challenges have resulted in substantial gaps in road network data. This study introduces a low-cost, cloud-based approach that fuses radar and optical satellite imagery within Google Earth Engine to enhance road visibility and mapping under these conditions. By integrating multi-sensor satellite data with local knowledge and publicly available reference datasets, the approach enables the identification of previously unmapped roads and the correction of outdated road surface classifications. The results show that more than 71 km of undocumented roads were detected, increasing the known road network by approximately 1%, while achieving a connectivity rate of 98.8% with the existing network. These findings demonstrate that multi-sensor data fusion, combined with informed local interpretation, provides a scalable foundation for improving road inventories and generating GeoAI-ready baseline data in fragile and data-scarce environments.

Time-Series InSAR Analysis Using PS-InSAR, SBAS-InSAR, and Merged Approach: A Case Study of Khun Dan Prakarn Chon Dam, Nakhon Nayok, Thailand

N. Khumpongpan — 2026-09-29

This study compares Persistent Scatterer InSAR (PS-InSAR), Small Baseline Subset InSAR (SBAS-InSAR), and the StaMPS merged processing mode for line-of-sight (LOS) ground-motion screening around Khun Dan Prakarn Chon Dam, Thailand, using 43 ascending Sentinel-1 Interferometric Wide Swath scenes acquired from January 2015 to July 2025. EZ-InSAR, ISCE, and StaMPS were used to generate mean LOS-velocity products and evaluate point density, standard deviation, normalized median absolute deviation (NMAD), mapped interferometric coherence, and agreement on spatially matched points. PS-InSAR produced the smallest velocity dispersion (STD = 0.601 mm/yr; NMAD = 0.513 mm/yr), whereas SBAS-InSAR showed greater dispersion (STD = 1.417 mm/yr; NMAD = 0.882 mm/yr). The merged result provided the greatest spatial sampling (35,143 points; 65.08 points/km2), with intermediate dispersion (STD = 1.002 mm/yr; NMAD = 0.647 mm/yr), and was therefore interpreted as a coverage-precision trade-off rather than a more accurate product. A common-support comparison matched 28,257 merged-PS points within 30 m and yielded an NMAD of velocity differences of 0.334 mm/yr, but weak pointwise association (R² = 0.258), showing that small robust differences do not imply strong equivalence. Most velocities were near zero; however, these relative single-orbit LOS observations cannot independently resolve vertical and horizontal motion or verify structural safety. Residual atmospheric effects, vegetation decorrelation, and viewing geometry limit physical interpretation. Without collocated field observations, the results therefore provide an internal comparison and remote-sensing screening indicators rather than external validation of dam performance.