Wet and dry: Models on climate extremes questioned

It seems to me that the researchers' claim of discrepancies between climate models and proxies during the most recent century is based on an “apple-to-oranges-comparison”. They use proxies of water availability (that is affected by both precipitation and evaporation) but compare against modelled precipitation only. The model simulations show increased precipitation over the past 100 years, consistent with an intensification of the hydrological cycle as temperatures rise related to increasing greenhouse gases in the atmosphere. With increasing temperatures, however, also the amount of evaporated water may increase, and this may compensate the increases in precipitation. This means that the water availability proxies and modelled precipitation are not reasonably comparable over the last century (which is the only period where their comparison seems to point to discrepancies between proxies and models).

In a recent study (Donat et al., 2016, Nature Climate Change, doi:10.1038/nclimate2941) we found that, when aggregating over the dry and wet regions of the world, precipitation changes are consistent between models and observations over the past 60 years. Nevertheless, it is true that modelling and analysis of precipitation changes are still related to a number of uncertainties, especially when it comes to regional changes in precipitation. This is partly related to the large temporal variability in local precipitation time series, but also shortcomings in the models with simulating processes related to precipitation.

This paper provides substantial new proxy evidence of Northern Hemisphere hydrological changes over the past 1200 years, ideal and very welcome for comparing with climate model simulations.

There seems to be broad agreement on trends in time and space, but at smaller scales and over shorter periods, the models and the proxy observations are bound to disagree. This is because of inherent uncertainties in the proxy climate reconstructions, and uncertainties in models. Plus, the large random component to how the climate varies regionally, especially for rainfall.

I am not too surprised that there is disagreement for the 20th century as there is a strong component of random variability evident in the observational record. The picture of the “wet getting wetter and the dry getting drier” is one that is very likely to emerge over the course of this century but has not been evident, or expected, during the 20th century.

However, we do know that human-induced climate change is already affecting the hydrological cycle. For instance, the recent Syrian drought is estimated to be the worst in 900 years and has a clear climate-change fingerprint. Extreme rainfall and flood events over recent years in many parts of the world have been shown to be much more likely to occur now than in the past, because of the climate changes we have already seen.

I look forward to more work in this important area, as a better picture of the past can only help clarify our outlook for the future.

They did a lot of work and its a nice study and a valuable proxy dataset, but it has many complexities in the way it was done. They are heavily smoothing the data so as to look only at centennial-scale shifts, not what we usually think of as droughts or rainfall extremes (which would be scales of days to at most a decade or two).

Previous studies, based on models, have shown that warming-induced trends in regional precipitation have not yet emerged from natural variability (“noise”). This seems inconsistent with the paper’s claim that the changes predicted by these same models are unrealistic, since it should not yet be possible to tell even according to the models themselves.

They normalised all data series to unit variance over the 1200 years. If I understand correctly, that means if there is too little natural variability in the models (which has been reported in other studies) or noise in the proxy observations (which is inevitable), either of these would help create an apparent discrepancy such as they’ve noted. Thus, other interpretations of their results may be more consistent with past studies.

In models, they looked at precipitation, but the observational proxies they have are probably measuring soil moisture or something similar.  Many people think these two quantities are equivalent, but they are not, because as soon as you warm the whole planet they decouple (precipitation generally goes up while soil moisture may stay the same or go down). This will cause discrepancies during the last century, due to the global warming not seen in earlier periods.  Future studies should compare these hydroclimate proxy results to a more appropriate model variable such as soil moisture.

If this paper’s conclusion about model overprediction holds up to further scrutiny it will be extremely interesting; my own work focuses on model-data discrepancies so I am particularly interested.  But due to the above aspects of the study, I am not convinced that this particular conclusion will hold up.  We shall see, as I am sure this result will attract lots of attention.