Research often fails to account for the specific pathways by which climatic factors can cause social unrest. One challenge lies in understanding the distinct effects of food insecurity and water insecurity – which we term ‘staple insecurities’ – while accounting for their interrelated nature, especially at high-resolution spatio-temporal scales. To unpack these dynamics, we leverage geolocated Twitter data across urban areas in Kenya and deploy a supervised machine learning approach to separately identify geolocated tweets concerning food and water insecurity, in both English and Swahili. The data are then aggregated to create daily measures of food and water insecurity for standardized grid-cells to examine how perceived food insecurity moderates and/or reinforces perceived water insecurity’s impacts on social unrest, and vice versa. Our findings suggest that food and water insecurities’ respective effects should be interpreted as mutually reinforcing – in compelling citizens to take to the streets – rather than as independent. Those concerned with climate change’s impact on conflict should hence endeavor to jointly account for both forms of insecurity, and their interactive effects.

研究通常无法说明气候因素可能导致社会动荡的具体途径。挑战之一是要理解粮食不安全和水不安全的独特影响（我们称之为“主要不安全感”），同时还要考虑到它们相互关联的性质，特别是在高分辨率时空尺度上。为了解开这些动态，我们利用肯尼亚城市地区的地理定位Twitter数据，并部署了有监督的机器学习方法，以英语和斯瓦希里语分别识别与粮食和水不安全有关的地理定位推文。然后将数据汇总起来，为标准化网格单元创建粮食和水不安全状况的每日度量，以检查感知到的粮食不安全状况如何缓解和/或增强感知到的水不安全状况对社会动荡的影响，反之亦然。我们的研究结果表明，粮食和水不安全的各自影响应被解释为是相互促进的（强迫公民上街）而不是独立的。因此，那些与气候变化对冲突的影响有关的人应努力共同解决两种形式的不安全及其互动影响。