Traditional landslide monitoring techniques and instruments are limited by spatial constraints. This study applies remote sensing-based Interferometric Synthetic Aperture Radar (InSAR) technology to enable large-scale surface displacement monitoring in mountainous areas. Two multi-temporal InSAR approaches-Persistent Scatterer InSAR (PS-InSAR) and Small Baseline Subset-InSAR (SBAS-InSAR) -were adopted. Ren'ai Township in Nantou County, Taiwan, was selected as the demonstration area. A total of 30 ascending Sentinel-1A radar images from 2017 were used to compare the applicability of these two methods for surface displacement monitoring, and correlation analysis was conducted with Global Navigation Satellite Systems (GNSS) data. Both PS-InSAR and SBAS-InSAR exhibited statistically significant positive correlations at the LSAN GNSS station, with correlation coefficients of 0.486 and 0.399, and root mean square errors of 5.004 mm and 7.685 mm, respectively. The SBAS-InSAR results effectively captured the spatial distribution of actual landslides and surface deformation in the mountainous area, indicating that this technique may offer greater advantages for landslide monitoring in such complex terrains.

Traditional path planning algorithms for Unmanned Aerial Vehicles (UAVs) primarily optimize for geometric metrics such as path length and energy efficiency. However, in GPS-denied environments, where external positioning is unreliable, the quality of visual localization is paramount for mission success. This study introduces a novel Deep Reinforcement Learning (DRL) framework designed to co-optimize the UAV path for both geometric efficiency and visual localization robustness. Specifically, our method integrates the density of matched image feature points, extracted from post-processed aerial imagery, directly into the planning process, ensuring the generated trajectory passes through visually rich areas that enhance navigation accuracy. To tackle the path planning challenge and address issues related to sparse rewards and unstable training, we employ an advanced DRL architecture: Noisy Dueling Double DQN with Prioritized Experience Replay (Noisy D3QN with PER). This integration leverages Double DQN to refine value estimation, Dueling DQN to improve generalization, PER to enhance sample efficiency, and Noisy Networks to promote robust and efficient exploration. The proposed framework is implemented within a simulated 2.5D environment with a customized reward function that considers both UAV state parameters and terrain features. Experimental results demonstrate that the method generates efficient, visually coherent, and dynamically smooth trajectories. Crucially, it enables path inference for multiple independent missions from various starting points after a single training session, achieving superior computational efficiency compared to traditional geometric planners. This highlights the potential of integrating visual features into a reinforcement learning-based UAV path planning to significantly enhance visual localization performance in complex environments.

This study takes Miaoli County as a demonstration area to develop a high-resolution NO2 estimation model by integrating multiple data sources, including meteorological data, ground monitoring, satellite remote sensing, land use, and data from the Tongxiao Power Plant. An hourly 50 m × 50 m resolution model was constructed using the XGBoost algorithm combined with a SHAP-based incremental feature selection mechanism, with lag1 (1-hour lag) identified as the optimal time delay. The model achieved an R2 of 0.80 and an RMSE of 1.78 ppb. SHAP analysis revealed that NOx and road density were the most influential predictors, while emissions from the power plant also exhibited regional impact. Multilevel validation confirmed the model’s robustness across temporal, spatial, and high-pollution scenarios. The results demonstrate the potential of Geo-AI in air quality estimation for small and medium-sized counties and provide a quantitative basis for risk warning and pollution control.

The primary contribution of this study lies in the application of UAV-derived digital surface models (DSMs) to refine cross-sectional data for hydrological analysis and flood inundation simulations in the Han River, Taichung City, Taiwan. Based on 104 years of rainfall records from the Taichung Weather Station, the one-day maximum rainfall was estimated using five probability distribution models. Goodness-of-fit tests and error analyses identified the Log-Pearson Type III distribution as the most appropriate. Flood hydrographs were then generated using the triangular unit hydrograph method, which provides a more accurate representation of the hydrological characteristics of the Han River. Flood condition was simulated in HEC-RAS under seven recurrence intervals, with the results indicating overtopping at Section Four of the Han River under the 50- and 100-year recurrence intervals. By integrating UAV-derived DSMs to improve cross-sectional accuracy, this study enhances the precision and reliability of flood simulation outcomes. The findings provide valuable references for river management and flood mitigation planning by water resource authorities.