摘要
Abstract
In order to scientifically optimize the irrigation regime for rice under drought stress,this study employed a projection pursuit clustering(PPC)model based on a real-coded accelerated genetic algorithm(RAGA)to evaluate the impact of drought stress on rice growth using experimentally measured data.The RAGA-PPC model was more suitable for non-normal and non-linear high-dimensional data processing through data-driven projection dimension reduction technology,which can objectively reflect the difference in the contribution of each growth index.In this paper,four key growth indicators,plant height,leaf area index(LAI),tiller number,and water consumption,were selected to represent the physiological and morphological responses of rice to water deficit.These indicators were normalized and processed through the RAGA-PPC model to determine their respective weights via an optimal projection direction.The model was optimized using a population size of 1 000,crossover probability of 0.7,mutation probability of 0.1,and 100 acceleration cycles,yielding a maximum projection index value of 140.902 9 and an optimal projection direction vector of a*=(0.787 6,0.458 9,0.397 9,0.104 0).The results revealed that drought stress applied during the tillering stage(at saturated water content of 40%～70%)most significantly inhibited plant height and led to abnormal dynamics in tiller number.Drought during the jointing stage reduced LAI,while stress during the heading and milk-ripening stages had relatively minor effects on growth indicators.Among all indicators,plant height contributed the most to the comprehensive evaluation(weight=0.787 6),followed by tiller number(0.458 9),LAI(0.397 9),and water consumption(0.104 0).This indicates that plant height is the most sensitive and reliable indicator for assessing long-term water stress effects.Based on the projection values and growth responses,the optimal irrigation strategy under pot conditions for the rice variety Huanghuazhan was identified as maintaining 60%～70%saturated water content during the tillering stage,≥70%during the jointing stage,and 50%～60%during the milk-ripening stage.This strategy not only mitigated the negative effects of drought stress but also resulted in a significant difference in yield compared to CK(P<0.05).The study confirms the effectiveness of the RAGA-PPC model in handling multidimensional agricultural data and providing actionable insights for water-saving irrigation practices.The results offer a scientific basis for optimizing water management in rice cultivation,particularly in drought-prone regions,thereby supporting sustainable agricultural production and enhanced water use efficiency.Furthermore,the comprehensive evaluation framework established in this study can be extended to drought response research in other crops or ecological regions,offering a generalized methodology for precise water management in smart agriculture.关键词
干旱胁迫/综合评价模型/实数编码加速遗传算法/投影寻踪模型Key words
drought stress/comprehensive evaluation model/real-coded accelerated genetic algorithm/projection pursuit model分类
建筑与水利
