Urban forest carbon storage and sequestration on the Qinghai-Tibetan plateau: Machine learning analysis and management implications for Xining, China

Urban green space (UGS) in cold regions may have distinct carbon storage (CS) and sequestration (CSE) mechanisms due to low temperatures, low oxygen, and high radiation. This study focuses on Xining, the largest city on the Qinghai-Tibet Plateau. Based on 215 field samples and quantified using i-Tree Eco, CS and CSE densities are 4.08 kg C m −2 and 0.69 kg C m −2 y −1 , respectively. A correlation analysis and the XGBoost-SHAP model were employed to analyse three aspects: plant natural environmental factors, plant community structure factors, and biodiversity factors. The findings revealed that, while CS and CSE in Xining's UGS are influenced by tree specifications such as Diameter at breast height (DBH)and Hight (H), unique phenomena still exist. Tree specification diversity (DBHS) and tree height diversity (HS) have a greater impact than biodiversity indices like Shannon-Weaver index (SW) and the Pearson diversity index (PU). When Crown width (CW) exceeds 3.5, the growth of CS and CSE in trees and plant communities decreases. Shrubs respond more to factors like Elevation (DEM) and Rainfall (P), whereas trees show a less significant response. Herbaceous plants are more sensitive to species diversity changes but are less significant in CS and CSE. Strategies for optimizing CS and CSE in UGS are proposed, including protecting plant resources, refining designs, and using adaptive native species like Styphnolobium japonicum (L.) Schott, Ulmus pumila L., and Picea crassifolia Kom. These findings highlight the unique carbon storage and sequestration mechanisms in cold cities, offering scientific evidence for their optimization and management.