Multi-scale assessment of the relationship between building weight and urban land subsidence: A case study of Shanghai

• A 10 m resolution building weight estimation method was proposed. • The relationship between multiple building load indicators and land subsidence at multi-scales were revealed. • Building weight could be a better building load indicator for indicating land subsidence risk. • Building load had a greater negative impact on land subsidence at larger scales. • The loads of residential buildings are more likely to cause land subsidence than other building types. Land subsidence in coastal cities may increase the risk of being submerged, and building load is one of the important causes of urban land subsidence. Considering the rapid urbanization of China, it is increasingly urgent to examine the impact of the growing building load on urban land subsidence in coastal cities of China. At present, few studies have introduced building weight as an indicator to characterize building loads, and there is a lack of large-area and multi-scale research on the relationship between land subsidence and building loads. On such a basis, this study raised a 10 m resolution urban building weight estimation method, and proposed a multi-scale, multi-index, and multi-subsidence-level investigation framework for the relationship between urban land subsidence and building load. Specifically, XGBoost model was applied on Sentinel images and OSM data to estimate the building floors in Shanghai; Then, building material intensity factors and building type vectors were introduced and combined with the estimated building floors to calculate building weight(The total weight was 2.12×10 9 t, the weight per unit area was 1.46×10 6 t/km 2 , the average weight exceeded 5,000 t). Meanwhile, SBAS-InSAR technology and Sentinel data were employed to obtain the land subsidence situation in Shanghai from 2015 to 2022(The average subsidence rate ranged from -18.25∼12.19 mm/y, the cumulative subsidence range was -128.71∼79.10 mm). Subsequently, the relationship between urban land subsidence and four building load indicators (building weight, building floor, building volume, and building density) at multi-scale (pixel, building, block, and grid) and multi-subsidence levels (all areas, subsidence area, and severe subsidence area) were revealed. The main findings could be demonstrated as follows: (1) the correlation between building weight and land subsidence was more significant than other building load indicator(-0.22∼-0.15 versus -0.20∼-0.02 at block scale); (2) in areas where subsidence has already occurred, places with heavy building weight would be more likely to aggravate subsidence(-0.22∼-0.04 versus 0.05∼0.07 at block scale); (3) the relationship between building loads and land subsidence in various scales were significantly different, the negative impact of building loads on ground subsidence would be more pronounced on a large scale(-0.22∼-0.15 at block scale versus -0.12∼-0.01 at pixel and building scale); (4) the construction loads of residential buildings would be more likely to cause land subsidence than other building types(-0.15∼-0.09 versus -0.02∼-0.01 at building scale). Our proposed framework can conducive to a more targeted selection of research scale, indicator system, and monitoring scope for coastal urban land subsidence research in the future. It would also be of great significance to the geological risk management and sustainable development of coastal cities.