Scenario modeling of ecosystem service trade-offs and bundles in a semi-arid valley basin

Abstract

Amidst the rapid urbanization process, substantial transformations have emerged within ecosystem services, exerting profound ramifications on the sustainability of ecosystems. Nevertheless, an existing dearth persists in delineating the intricate interplay of trade-offs and synergies, as well as ecosystem services bundles under diverse future scenarios. This study harnesses the Convolutional neural network-Long and short-term memory-Cellular automata model to prognosticate and dissect the temporal and spatial dynamics of four distinct ecosystem services (soil retention, water yield, carbon storage, and habitat quality) across the semi-arid valley city of Lanzhou from 2000 to 2030 under multiple scenarios. The SPSS model methodically quantifies the intricate trade-offs and synergies interwoven among these services, while cluster analysis reveals ecosystem services bundles across varying scales. Notably, (1) convolutional neural network-long and short-term memory-cellular automata model demonstrates remarkable proficiency in accurately forecasting land use, boasting an elevated precision level of 0.93. (2) The trajectories of carbon storage and water yield demonstrate a diminishing pattern between 2000 and 2020, against the ascending trends observed in soil retention and habitat quality. Within the ecological priority scenario for 2030, water yield experiences a sluggish decrease, while the abatement of soil retention, carbon storage, and habitat quality degradation is maximized. (3) The trade-offs (3 pairs) and synergies (3 pairs) among ecosystem services within the study locale exhibit a state of relative equilibrium. Noteworthy among these interactions is the prominent trade-off correlation (−0.101) between soil conservation and carbon storage within the city priority scenario. (4) Four discernible ecosystem services bundles manifest at the grid scale from 2000 to2030. Conversely, three bundles at the county scale in 2000, eventually decrease to two bundles between 2010 and 2030, hinting at a diminishing trend concerning the number of ecosystem services bundles as spatial scales coarsen. Furthermore, whereby these bundles progressively aggregate spatially over time.