Spatiotemporal variation analysis of regional flood disaster resilience capability using an improved projection pursuit model based on the wind-driven optimization algorithm

Abstract

Abstract Due to the weak methods available for evaluation of the resilience of regional flood disaster systems and the lack of research on the driving mechanism of resilience, by exploring the principles of regional flood disaster resilience and constructing a suitable evaluation index system, the wind driven optimization (WDO) algorithm was introduced, and an improved projection pursuit (PP) evaluation model of flood disaster resilience was proposed. Twelve farms under the Heilongjiang Agricultural Reclamation Hongxinglong Administration Bureau were included in the research area. A total of 43 primary indicators were selected from four criteria to describe the natural environment, culture, society, economic development and flood control technologies. The R clustering factor analysis method was used to determine 15 optimal indexes. The improved PP model based on the WDO algorithm (WDO-PP) was used to evaluate the flood disaster resilience of 12 farms. The results showed that the number of farms with a level IV rating on flood resilience decreased from 25% to 8.3% from 2002 to 2009. In 2009–2016, with the exception of the Bawuer and Shuguang farms, the flood disaster resilience index decreased, and that of the remaining farms increased. In 2002–2016, the Wujiuqi, Shuangyashan, Shuguang and Hongqiling farms in the central region of the Hongxinglong Administration Bureau were less resilient to disasters, and the farms that responded better to flood disasters were mainly located in the eastern or western Hongxinglong Administration Bureau near a river. Further analysis shows that the forest coverage rate, paddy field coverage ratio, shelter forest area ratio, proportion of primary industry, agricultural water use efficiency, and irrigation and drainage capacity were the key drivers of the flood disaster resilience in the Hongxinglong Management Bureau. Based on the Rastrigin and Schaffer functions, the results show that the success rate of the WDO algorithm is 100% over 10 iterations of the optimization calculation of the test function, while the success rate of the other two algorithms is relatively inadequate; however, in terms of value and standard deviation, both are better than adaptive particle swarm optimization (APSO) and adaptive genetic algorithm (AGA) algorithms. Moreover, in the convergence curve, the WDO algorithm converges fast, the number of iterations can achieve the optimal effect on average 3–5 times, and the AGA and APSO algorithms need more than 40 iterations to achieve the best-seeking effect. Taking the index of agricultural water use efficiency in Bawuer farm as an example, the index weight is greater than 60% and the utilization rate of agricultural water is more than 98%, which is closer to reality. Therefore, the evaluation results of the flood disaster resilience evaluation model proposed in this study are more accurate: WDO-PP>(adaptive genetic algorithm) AGA-PP>(adaptive particle swarm optimization algorithm)APSO-PP. In conclusion, the WDO-PP model has certain reference value for flood disaster recovery, monitoring and early warning.