The role of carbon emissions resulting from land use change in the compilation of national greenhouse gas emission inventories is of paramount significance. This study is centered on the Mianyang Science and Technology City New Area located in Sichuan Province, China. We used the CLUE-S model and Sentinel-2A remote sensing data from 2017 to simulate and validate land use changes in 2022. Based on this validation, we established three simulation scenarios: a baseline scenario, an agricultural development scenario, and a construction development scenario. Using remote sensing data from 2022, we projected the land use for 2030. We also used CO₂ concentration data collected in 2022 and 2023, processed the data using ArcGIS and Python, and conducted a quantitative analysis of carbon emissions under each scenario. Ultimately, the accuracy of both measured and predicted CO₂ values for 2023 was juxtaposed and authenticated, thus concluding the investigative cycle of this study. Key findings include: (1) The accuracy of the CLUE-S model in the study area was assessed using overall accuracy, quantity disagreement and allocation disagreement indexes. In 2022, the overall accuracy is 98.19%, the quantity disagreement is 1.7%, and the allocation disagreement is 2.2%. (2) Distinct land resource utilization characteristics in scenarios, highlighting potential impacts on economic development and pollution. (3) Increased carbon emissions across scenarios, with construction development showing the highest rise (4.170%) and agricultural development the lowest (0.766%). (4) The predictive accuracy of the validation group's CO₂ concentration values can reach 99.5%. This study proposes precise CO₂ prediction at the county level, thus laying the groundwork for future research endeavors. Such findings are indispensable for informing carbon policy formulation and promoting low-carbon development strategies.

Terrestrial net primary productivity (NPP) integrates vegetation responses to climate, terrain, and human activities, yet their combined effects in mountainous–basin regions remain unclear. Focusing on the Chengdu–Chongqing Economic Circle (CCEC) in southwest China, we build a framework that couples spatial diagnosis, interaction-aware attribution, and scenario-based projection. Using 500 m MODIS NPP (2000–2020) with climatic, topographic, land-use, and socio-economic data, we quantify NPP trends, use optimal-parameter GeoDetector and partial correlations to separate driver contributions and interactions, and train a random forest (RF)–SHAP model driven by CMIP6–SSP climate projections to 2050. The CCEC shows strong greening: 85.17% of the area exhibits increasing NPP and 68.56% shows extremely significant increases, with productivity peaking at mid-elevations (~1950 m) and intermediate slopes. Elevation, NDVI, and temperature dominate, while precipitation, slope, and soil moisture are secondary, and enhancement-type interactions, especially between elevation and precipitation, prevail. Land-use statistics and NPP transfer matrices highlight cropland-to-forest/grassland conversion as the main greening source. CMIP6-based simulations indicate stable or modestly higher NPP through 2050, with western mountain forests remaining key carbon sinks and basin lowlands constrained by warming and land-use pressure.

Post-event building damage assessment from bi-temporal satellite imagery is challenging because damaged pixels are sparse, structural changes can resemble radiometric variation, and sensor changes can dominate direct feature subtraction. This study evaluates SiameseCTDM with a Sensor-Adaptive Cross-Temporal Difference Module (SA-CTDM). At each encoder stage, SA-CTDM combines absolute and magnitude-normalized relative discrepancies through a learned per-channel gate, followed by channel recalibration. The gate provides adaptive interpolation between two discrepancy definitions; it is not treated as an independently superior accuracy component. A source-preserving class-conditional objective is also evaluated for optical-to-SAR adaptation by combining supervised BRIGHT learning, continued xBD supervision, and multi-stage alignment of intact and damaged prototypes. The adaptation terms are training-only and add no inference-time parameters. On the xBD hold set, complete SA-CTDM obtains 66.68 ± 0.54% F1, compared with 66.16 ± 1.12% for original CTDM, representing a modest 0.52 percentage-point mean increase with lower variance. A normalized-only ablation is the strongest single run, so an independent gating advantage is not claimed. A unified batch-one benchmark at $512\times 512$ input resolution reports measured latency and throughput in addition to parameter counts and GFLOPs. On a fixed seed-42 sample-level BRIGHT split, the complete configuration using 611 labeled pairs reaches 34.33% F1, whereas the aligned 305-pair configuration does not outperform vanilla fine-tuning. The results therefore characterize split-specific and label-budget-dependent behavior rather than uniform few-shot superiority.