The information diffusion model is widely used in natural disaster risk assessment because of its operationalization, low data requirements and clear significance of the evaluation results. However, the commonly used normal information diffusion model is based on an average distance assumption, and the diffusion widow with h is directly affected by sample size and extreme values, especially when the sample size is more than 11, which may lead to unstable results of normal diffusion, with typical uncertainty. Considering the influence of the whole samples on the diffusion points, incorporating the connection number method, a connection number structure-based information diffusion model was proposed, and also a relative connection number method is proposed to calculate the connection number, which was applied to the risk assessment of drought disasters in Jianghuai watershed area. The results show that the connection number structure-based information diffusion model reduces the direct influence of sample extremes and sample size, eliminates the constraint of the average distance assumption in the traditional model, and performs more consistently with increasing sample size for risk assessment. And the relative connection number method can contain the traditional connection number expression and offers a wider application. A study of drought risk in the Jianghuai watershed shows that the indices of agriculture drought-suffered, drought-affected and drought-damaged fluctuated greatly, but the trends were generally consistent with each other from 1990 to 2020. It is more likely to suffer from drought in the area, but the disaster rate after drought tends to decrease over time. With the increase in drought intensity, the disaster risk will decrease significantly compared to the drought risk. The connection number structure-based information diffusion model has no two-point proximity principle and the average distance assumption, with better stability in the disaster risk assessment for different samples size compared with the traditional model, which will have more extensive application in risk assessment.
Read full abstract