食品生产消费的全生命周期环境影响突出，其中居民消费具有支配作用。由于消费主体异质性特征和消费方式显著变化，带来了认识刻画个体消费行为和量化系统性环境影响的重要挑战。论文综合运用环境社会学理论、生命周期分析和环境系统建模方法，深入研究家庭食品购买行为决策机制与环境认知影响，系统评估食品消费生命周期环境影响及关键因子，定量模拟若干社会经济发展场景下家庭和社区食品消费行为变化及潜在环境影响，研究成果可为食品可持续消费、低碳社区管理等提供研究支撑和决策依据。论文首先提出了扩展的ABC理论模型（ABC-PT），结合机器学习和结构方程方法，识别了环境意识、风险感知等关键因素影响下食品购买行为的决策机制。其次，基于全生命周期分析方法，建立了以食品购买为中心节点，连接食品供给与消费的全生命周期碳-水足迹核算模型（LCA-CWF），对五种食品类型、四种购买渠道下食品碳-水足迹进行了核算。第三，基于自底向上建模方法，开发了食品购买行为决策及碳-水足迹模拟模型（FOODCM），定量模拟了新冠疫情管控、供给设施配套、家庭人口结构演变等社会场景下居民行为决策变化响应，以及对社区尺度碳-水足迹的影响。论文以北京市海淀区三个社区为案例开展实证研究。研究结果表明，ABC-PT理论模型可以有效解释居民家庭在食品购买行为中的决策机制。其中，社区服务设施、疫情管控政策等因素在食品购买决策中依次占据支配地位，且通过感知环境风险、主观规范以及环境意识等态度因素间接影响购买行为。以消费为中心的生命周期评价结果表明，通过生鲜电商购买食品的平均碳足迹比传统购买渠道提高4%、水足迹降低11%。包装环节、冷链运输分别贡献了21%和23%碳足迹，食品损耗占全生命周期水足迹70%以上。案例社区家庭食品购买平均碳-水足迹分别为10.7±3.4 kg CO2/人/周和10.2±3.3 m3/人/周，呈正偏态分布，且具有明显边际递减效应。情景模拟结果表明，未来电商发展将进一步提高人均食品消费碳足迹13%，但水足迹将降低16%。疫情管控场景下人均碳-水足迹则分别增长了42%和22%，但可通过提高居民的环境意识和感知行为控制水平、降低风险感知和主观规范实现碳-水足迹减少52%-79%。社区碳-水足迹变化幅度相比家庭单元平均降低11%和15%，表明家庭食品消费环境影响在社区层面上具有明显尺缩效应。

The life-cycle environmental impacts of food production and consumption are prominent, with household consumption playing a dominant role. Due to the heterogeneous characteristics of consumer subjects and significant changes in consumption patterns, challenges have arisen in understanding and characterizing individual consumption behavior and quantifying systemic environmental impacts. This paper comprehensively employs theories from environmental sociology, life cycle assessment, and environmental systems modeling to investigate the decision-making mechanisms of household food shopping behaviors and the influences of environmental cognition. It systematically evaluates the life-cycle environmental impacts of food consumption and identifies key factors. It quantitatively simulates changes in household and community food shopping behaviors under various socio-economic development scenarios, aiming to provide research support and decision-making basis for sustainable food consumption and low-carbon community management.The paper firstly proposes an extended ABC theory model (ABC-PT), identifies the decision-making mechanisms of food shopping behaviors under the influence of key factors such as environmental awareness and risk perception, by combining machine learning and structural equation methods. Secondly, based on the life cycle assessment, a carbon-water footprint accounting model (LCA-CWF) has been established with food shopping as the central node, connecting food supply with consumption throughout the entire life cycle. This model calculates the carbon-water footprints of five food types across four retail channels. Third, based on the bottom-up modeling method, a food shopping behavior decision-making and carbon-water footprint simulation model (FOODCM) was developed. This model quantitatively simulates changes in household behavior decisions in response to societal scenarios such as COVID-19 control, supply facility development, and household demographic shifts. It also assesses the impact of these changes on the community-scale carbon-water footprint. This paper conducts empirical research using three communities in Haidian District, Beijing as cases. The research results indicate that the ABC-PT theoretical model can effectively explain the decision-making mechanisms of household food shopping behaviors. Among them, factors such as community service facilities and epidemic control policies sequentially dominate food shopping decisions, and shopping behaviors are indirectly influenced through perceived environmental risks, subjective norms, and environmental awareness attitudes. The consumption-centered life cycle assessment results show that the average carbon footprint of food shopping through food e-commerce is 4% higher than that of traditional shopping channels, and the water footprint is 11% lower. The packaging process and cold-chain transportation contribute 21% and 23% of the carbon footprint respectively, and food loss accounts for more than 70% of the entire life cycle water footprint. The average carbon-water footprint of household food shopping in the case community is 10.7±3.4 kg CO2/person/week and 10.2±3.3 m3/person/week respectively, showing a positively skewed distribution and an obvious diminishing marginal effect. Scenario simulation results indicate that future e-commerce development will increase per capita food consumption carbon footprint by 13%, while the water footprint will decrease by 16%. Under the COVID-19 control scenario, the per capita carbon-water footprint increased by 42% and 22% respectively. Increasing environmental awareness and perception of behavioral control while reducing risk perception and subjective norms can decrease carbon-water footprints by 52%-79%. The environmental impact of household food consumption has a significant scale-down effect at the community level, with the community‘s carbon-water footprint changes averaging 11% and 15% lower than those of the household unit.