Is it even possible to reliably measure this? Sure, the statistical test says 'significant', but the number of potential confounding geographic variables must be enormous. How can this study have possibly ruled out other pollutants, economic conditions, or suchlike correlating with how weapons test sites are chosen?
It seems very plausible that nuclear weapons testing caused pollution, but it also seems plausible that the weapons were being tested in areas that were wilting. I don't know, but I am very cautious about using this data to conclude what specific effect radiation pollution has. The 'Robustness' section looks fairly cursory.