The literature on climate change and crop yields recognizes the need to allow for highly non-linear marginal effects. This study combines these two areas of the literature by using Flexible Fourier Transforms (FFT’s) to ensure flexibility for both the time trend and the weather effects. This study also illustrates how FFT’s can be combined with quantile regression (QR) to provide both robustness to outliers and information on the scale effects of time and weather variables. For U.S. county level data on corn, soybeans, and winter wheat, we estimate the relationship between yield and temperature and precipitation using a traditional parametric expected-yield estimator, our quantile-FFT regression evaluated at the median, and our QR-FFT regression that incorporates information on the tails of the distribution. We find that quadratic terms are not sufficient for capturing nonlinearities in the relationship between yield and the explanatory variables.