将来の環境流量に対する気候変動の影響評価

在上一财年，研究发现，与使用传统的单一Deme GP算法开发的模型相比，使用并行遗传编程(PGP)算法开发的模型显示出更好的泛化技能，并产生更少的非物理大异常值。然而，传统的GP和PGP算法都不能准确地捕捉到观测时间序列中的极值。在初步调查中发现，月度降尺度方法结合月到日分解方法具有更好的模拟极值的潜力，但模式的整体性能不如日降尺度模式。此外，众所周知，当观测中存在总计/平均值相似且月内分布明显不同的月份时，分解方法的执行情况很差。在这一年中，对改进降水、温度和径流极值的模拟进行了全面的调查，重点是分解的使用。还研究了将每日模拟与每月至每日分类模拟相结合的可能方法。用改进的方法预测了未来的径流，调查了满足环境需求的水的可用性。根据这些新的分析结果，结果被总结成手稿，并以论文形式提交给期刊。

In the previous fiscal year, it was found that models developed with Parallel Genetic Programming (PGP) algorithm (introduced in this research) show better generalisation skills and produce fewer unphysically large outliers compared to that of models developed with the traditional single deme GP. However, both conventional GP and PGP algorithms were not able to capture the extremes in the observed time series accurately. In a preliminary investigation, it was found that a monthly downscaling approach coupled with a monthly to daily disaggregation method has a better potential to simulate extremes, but the overall performance of the model is not as good as that of a daily downscaling model. Also, disaggregation methods are known to perform poorly when months with similar totals/averages with significantly different intra-monthly distributions are present in observations. In this year, a comprehensive investigation on the improvement to the simulation of extremes in precipitation, temperature and streamflows was conducted focussing on the use of disaggregation. Potential methods to combine the daily simulations with monthly to daily disaggregated simulations were also investigated. Streamflows were projected into the future with the improved methodology, availability of water to satisfy the needs of the environment was investigated. Results were summarized into manuscripts and submitted as papers to journals based on the finding of these new analyses.