Whether you want to acknowledge it or not the field is inherently uncertain because repeatable real world experiments cannot be done. Any claimed "consensus" is manufactured and only evidence of group think.
Experimental manipulation of variables in controlled (eg double blind) settings can mimic and elucidate very specific effects related to climate change. But their validity can't be generalized beyond those controlled settings. To understand the larger picture in the REAL world requires the collection over time of data on a large number and variety of factors about climate itself, and about factors that may impact on climate, all in situ - in the natural settings where they occur - with multiple factors and interactions of factors intact.
Computerization since the 70's has given us the capacity to collect and analyse these huge datasets.
Experimental manipulations of factors - like double blind studies - can't tell you how the factors would act and interact in the real world. But they can generate hypotheses that can be applied to and evaluated through analyses of large real world datasets.
Those analyses may reveal that factors tested in laboratory studies may have different effects, depending on interactions with other, possibly unanticipated, factors in the natural environment.
(Eg, forested vs deforested areas, etc). Those additional factors can then be included in future experimental manipulations to assess their effects more closely, generating more specific hypotheses that can then be applied to the large datasets .... etc.
Double blind studies are not the be all and end all of science (as believed in the computerless 1960's). They are necessary for generating specific hypotheses, but they cannot stand alone as they can only produce findings relevant to specifically controlled factors in unnatural settings. Unlike the 60's, we now have the capacity to test those hypotheses on real world data.
We'd be silly fools to cherry pick preferred methods and fight about them, when the reality is so complex that it requires all methods interacting appropriately to elucidate the complexities of climate.