DEEP LEARNING-BASED STEREO MATCHING FOR HIGH-RESOLUTION SATELLITE IMAGES: A COMPARATIVE EVALUATION

We developed NZ-BeachTopo30, a full-coverage 30 m beach topography dataset for New Zealand, by integrating Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) and Sentinel-2 data with extreme gradient boosting (XGBoost). It expands valid intertidal coverage by 145.8 % and achieves a 0.94 m root mean square error against airborne light detection and ranging (airborne LiDAR) data, supporting sea-level rise and coastal erosion planning.

Central–South Asia and the Tibetan Plateau depend on fragile water systems shaped by rain, snow, ice, and heat. We used future climate model results to test how warming and sunlight-reflecting climate intervention could affect water, plants, snow, and seasonal timing. Warming greatly increases water extremes and shifts seasons earlier. Intervention reduces heat and some extremes, but cannot fully prevent water stress, and its effects vary strongly by region and design.

This work presents GGFD-POM, the first global gap-free daily 4 km dataset covering particulate organic carbon (POC), particulate organic nitrogen (PON), and their ratio (POC:PON) from 1998 to 2023. The dataset provides continuous high-resolution data, enhancing studies of particulate organic matter (POM) dynamics in relation to the carbon cycle and biogeochemical stoichiometry.

Mangroves protect coasts and store carbon, but their thick leaves hide the ground from satellites, making it hard to map the land beneath. We created the first detailed map of New Zealand’s mangrove floor using free satellite data and a smart learning model. By combining laser measurements with radar that "sees" through branches, we achieved high accuracy. This tool helps coastal communities predict flooding from rising seas and better understand how these vital forests fight climate change.