Brainfart. I missed that. But they do say in their methods:
"Future studies should explore spatial and temporal multi-scale effects [50]; we do not consider potential physical interaction between the explanatory variables at different scales in different geographical units. We would also like to caveat against interpreting our regression-based analysis as causality. Needless to say, detailed studies using technically more complex statistical models, including causal models, and more extensive model diagnostics, is needed. Ignoring spatial autocorrelation does not necessarily bias the final results, however inclusion of such effects may improve the precision of the estimates. On the other hand, spatio-temporal statistical models can be computationally and mathematically very challenging. Future publications should try to address spatio-temporal dependency, diagnostics associated with statistical modeling and other issues mentioned above in global-scale studies."
Several parts of the methodology section cautioned that very simple regressioon models were used and many assumptions made. In a nutshell they concluded that there likely is a correlation between changes in food production and localized climates (claiming to study a few global crops is actually a series of localized studies).
And I'll add the first posted article was written by a politics writer so take it worth a grain of salt. She likely can't parse a scientific paper anyway. But it's good research with a VERY misleading title. I'm surprised it was accepted as titled. Without the causality (which they caution us not to glean) it's really an effective study on how a localized climate could effect crops which is very valuable to the farming community.