Assessing changes and predictability of crop yields and failure risk in the United States: The Impact from Large-scale Climate Circulations

Abstract

The weather of the growing season influences crop production and yield. These changes in crop yields can result in economic loss and increases in global food insecurity drastically especially when high production areas are threatened. Seasonal influences on crop yields may be in part due to climate oscillations, which have been linked to floods and droughts. This study will focus on analyzing the impact of climate oscillations on summer (maize) and winter (winter wheat) crop yields from 1960 to 2016 in the rainfed United States, a region affected by several climate oscillations, including Atlantic Multidecadal Oscillation (AMO), North Atlantic Oscillation (NAO), El-Niño Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), and Pacific-North American (PNA).

The first chapter of this thesis is to explore and assess the linkage of crop yields variability and climate oscillations. Principal Component Analysis (PCA) shows the first five rotated principal components explain over 70% of the spatial and temporal variability of crop anomalies. AMO is strongly associated with the first rotated principal component. Linear regressions support previous findings that the reproductive period is the most sensitive period for yield forecasting. Categorical yields (low yields below the 30th percentile and high yields above the 70th percentile) are well predicted by climate oscillations using Random Forest, with AMO as the leading predictor in nearly half of maize and a third of winter wheat climate divisions.

The second chapter assesses changes of crop failure, defined as the lower quartile of yield anomalies, influenced by climate oscillations. A Bayesian approach is used to assess crop failure risk. The results show that positive AMO and negative PNA phases greatly increase maize crop failure. For winter wheat positive NAO increases frequencies of crop failure. Combinations of climate oscillation phases show a positive AMO and negative PDO increase maize crop failure for the majority of the study area, while a negative AMO and any phases of PDO combination increases winter wheat crop failure. The second combination, ENSO and PDO, show that when the oscillations are out-of-phase, the largest changes to crop failure frequencies are experienced.

The findings from this work have implications for improving seasonal forecasting of yields, risk management, and seasonal decision making for various stakeholders. To expand on these findings, future work in this area can include different crops and include other sources of data. By including other data sources, event case studies can be conducted to validate crop loss causes. Combinations of climate indices can be analyzed in more detail. These additional measures would further contribute to the understanding and improvement of seasonal forecasting in the rainfed United States.