Exploring the economic loss characteristics of meteorological disasters in China based on CGE model improved loss function

The extreme events caused by global warming have had profound impacts on natural ecosystems and socio-economic structures. The Integrated Assessment Model (IAM) typically quantifies the economic losses resulting from climate change by constructing a loss function. However, the construction of loss functions in IAM is relatively simplified, and there is no unified standard for the function form and quantitative indicators of extreme events. We aim to introduce the impacts of climate change into Computable General Equilibrium (CGE) model in the form of loss functions. To more accurately assess the impact of extreme events on economic losses, we selected the extreme precipitation and temperature index and the Standardized Precipitation Evapotranspiration Index (SPEI), to explore their nonlinear relationships with direct economic losses from different disasters using MLP neural networks and three ensemble learning algorithms: Random Forest (RF), Extreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM). The results show that the LightGBM algorithm performs the best, with R ˆ 2 over 92 % and MAPE dropping below 10 %, and the level of economic development is the dominant factor in regional disaster losses. In the last four years, China has not experienced fluctuation in economic losses caused by serious extreme events, the disaster prevention and reduction work has achieved great results. Direct economic losses caused by heavy rainfall and flood disasters have significantly decreased, while low-temperature freezing and drought disasters have increased to varying degrees. The affected areas tend to be concentrated as a whole, with certain spatial heterogeneity.