Climate change and shifting variability patterns are expected to affect the worldwide productivity of crops and hence the food supply. The case of maize in Mexico is interesting because its production is of capital importance to the country's culture and economy, being also the staple crop for most subsistence farmers. Maize crops account for 40% of the country's arable land, and 65% of the total maize production happens under rainfed regime. This dissertation addresses, via three independent studies, the uncertainties associated with current climate variability and change in Mexico and the impacts on rainfed maize production.

In Chapter 1 I defined a set of climate parameters relevant for maize production, evaluated their trends over six decades, and investigated the potential influence of large-scale climate teleconnection patterns. In Chapter 2 I assessed the impact of climate variability on state-level maize yield anomalies. Three decades of seasonal observations were used to fit multiple regression models, in which geographical differences of agro-climate conditions selectively defined the set of predictors for each state. Finally, in Chapter 3 I propose a new statistical framework to assess the impacts of climate stressors on maize crops. Using municipalities as spatial units and observations over four growing seasons, the occurrence of crop failure was modeled as a function of climate stressors specific to each stage of the growing period.

In brief, significant portions of Mexico experience trends towards climate conditions potentially hazardous for maize growth, such as increasing dryness, delayed rainy seasons, more extreme precipitation and heat stress events. However, predictability could be improved by further understanding of the driving roles played by both the Atlantic and Pacific oceans on Mexico's climate. The state-level regression models show that climate variability affects maize yields differently across the country, and the varying sensitivity could be associated not only to region-specific climates but also to the heterogeneity of production scales, adaptation practices, and reliability of yield reports. Lastly, statistical modeling of crop failure considering stage-specific climate stressors offers a potentially robust alternative to process-based crop models, as the effect of main stressors such as drought and excessive heat was successfully represented.