Multi-scenario simulation of carbon budget balance in arid and semi-arid regions

The carbon budget has emerged as a central focus in global carbon cycle research. The limited understanding of carbon budget balance dynamics has led to an increasing imbalance between ecological and socio-economic benefits. Building upon a comprehensive analysis of carbon storage and emission in Lanzhou from 2000 to 2020, this study develops a novel deep learning model (CNN-LSTM) to simulate carbon budget under various scenarios from 2030 to 2050. Additionally, scientifically grounded recommendations for carbon compensation are provided. The results demonstrate several key findings: (1) The deep learning model exhibits outstanding performance, with an average overall accuracy exceeding 0.93. The coupled model outperforms individual models, underscoring the significance and necessity of incorporating both temporal and spatial features in land use simulation. (2) Under the ecological protection redline scenario from 2030 to 2050, a noteworthy augmentation in carbon storage and a proficient constraint on carbon emissions are observed. This substantiates the effectiveness of ecological protection interventions. (3) Carbon compensation payment areas are predominantly concentrated in built-up land, with the extent of these areas expanding over time. (4) The disparities in carbon balance effects of forest were more conspicuous than that of built-up land across diverse temporal and scenarios.

Automated summary

The carbon budget is an important topic in global carbon cycle research. This study develops a novel deep learning model to simulate carbon budget in Lanzhou from 2030 to 2050 and provides recommendations for carbon compensation. The results show the outstanding performance of the deep learning model, the effectiveness of ecological protection interventions, and the concentration of carbon compensation payment areas in built-up land.