With the accelerated urbanization process, cities are suffering from extremely heavy rain and urban storm water logging disasters in recent years. To provide reliable and effective information for urban management and emergency decision-making, the accuracy of basic data must be guaranteed in any urban rainwater model. This paper presents a novel MKFCM-MRF (Multiple Kernel Fuzzy C Means-Markov Random Field) model to segment high-resolution Unmanned Aerial Vehicle (UAV) images. The core ideas of MKFCM-MRF model are as follows. Firstly, in order to increase the correlation information between pixels, multiple-kernel functions are introduced into Fuzzy C Means (FCM) clustering algorithm, which automatically filters out the optimal weight combination among kernel functions according to the distribution characteristics of pixels in feature space. Secondly, in order to better segment the texture and edge of the image, the clustering results of multiple-kernel FCM clustering algorithm are introduced into Markov Random Field (MRF) model, a novel spatial energy function integrating fuzzy local information is constructed. Finally, based on the total of data and spatial energies, the raw clustering results are regularized by a global minimization of the energy function using its iterated conditional modes (ICM). The effectiveness of MKFCM-MRF is performed by high-resolution UAV images data. The experimental results indicate MKFCM-MRF can refine the classification map in homogeneous areas, while reducing most of the edge blurring artifact, and improving the classification accuracy compared with FCM clustering algorithm. In addition, the validation of the urban storm flood model shows that the trend of the two clustering results is the same, but the runoff producing time and the peak time of FCM clustering results are advanced, the peak flow and the total runoff are larger; the simulation results corresponding to MKFCM-MRF clustering results are more realistic.

Frequent seasonal phase transitions in cold and arid lakes require different remote sensing indices for frozen and open-water periods, complicating the use of traditional empirical indices for automated monitoring. To address this challenge, this study proposes an intelligent indexing framework integrating the heuristic reasoning of Large Language Models (LLMs) with Random Forest (RF) feature selection. Leveraging the Google Earth Engine (GEE) and Landsat 8 data from Ulansuhai Lake, five LLMs such as Gemini and ERNIE were employed to generate candidate spectral indices based on typical sample spectra. Optimal band combinations were identified via RF importance, and Land Surface Temperature (LST) was incorporated as a physical constraint for unified cross-seasonal classification and determine the optimal threshold. Results show that the LLM-derived ERNIE-WISI and Gemini-WISI exhibit high robustness. During the freezing period, ERNIE-WISI significantly outperformed other indices, achieving an Overall Accuracy (OA) of 89% and a Kappa of 0.86. Spatially, it yielded snow and ice mapping with clear textures and low commission errors. During the non-freezing period, ERNIE-WISI achieved an OA of 95% with a Kappa of 0.84. While Gemini-WISI achieved an OA of 94% with a Kappa of 0.80, performing comparably to MNDWI. Notably, ERNIE-WISI effectively suppressed background interference in complex landscapes like narrow channels and aquaculture areas, maintaining high geometric fidelity and spatial continuity. A key advantage of ERNIE-WISI is its consistent performance without seasonal threshold adjustments. Aligned with the AI for Science paradigm, this methodology bridges AI-driven heuristic discovery and physical remote sensing, offering a robust, transferable solution for long-term dynamic lake monitoring in extreme environments, thereby facilitating sustainable water management.

The combined impacts of urban pluvial flooding and traffic congestion can severely delay emergency response. However, existing studies often focus on isolated scenarios, failing to systematically quantify the reduction in overall service capability and specific route disruptions to critical functional nodes under compound hazards. To address this problem, this study proposes a three-tier analytical framework to systematically evaluate the resilience of emergency services under compound hazards. The framework first utilizes spatial network analysis to simulate the overall spatial evolution of service capabilities for Emergency Medical Service (EMS) and Fire and Rescue Service (FRS) across various return periods and traffic conditions. It then delves into the emergency response coverage for vulnerable population places. Finally, the fastest-route analysis is employed to identify variations in rescue routing. The study reveals several critical insights. (1) As rainfall intensity and traffic congestion intensify, the coverage areas of EMS and FRS exhibit significant contraction and boundary erosion. Notably, the service areas of FRS show a distinct fragmentation pattern. (2) The protection levels for vulnerable population places (e.g., kindergartens, primary and secondary schools, and nursing homes) show a pronounced stepwise decline. Under extreme rainfall and the heaviest congestion, the 5 min coverage for these sites drops from 89.9% to 23.6% for EMS, and from 72.4% to only 15.1% for FRS, revealing a severe risk exposure for vulnerable groups. (3) Road inundation leads to a substantial extension of rescue routes and even results in the complete isolation of 141 primary and secondary schools. Overall, the framework provides actionable decision support to enhance urban emergency response under compound hazards.