The cornerstone of climate science must be the enhanced greenhouse effect caused by rising CO2 levels. Therefore understanding how CO2 may increase in the future is of fundamental importance, and you would assume that any Earth Systems Model(ESM) should as a minimum be able to describe past increases in CO2. I had naively assumed that the BERN model was tuned so as to match emissions to CO2 increases, but it turns out this is not true.
Here is the result for CO2 levels if you integrate the AR4 Bern model using the historic emission data (CDIAC) and then compare it to the actual CO2 measurements from Dome-C and Moana Loa. (see last post for details)
There are two clear problems. Firstly the CO2 measurements rise well above cumulative emissions from 1800 to 1950, which is seemingly impossible. The second problem is that even after 1950, the AR4 Bern model only begins to agree with ML after 2000 and even then cannot reproduce the time trend. So what is going wrong or did I perhaps miss something? Well I have not included Land Use ’emissions’ which has been estimated from historical records of deforestation. Deforestation reduces net carbon uptake by living biota leading to an annual ’emission’. To remedy this I have simply used the AR5 Land Use data from this R.J.Houghton’s study (CDIAC).
Here is the new result after adding in the ‘land use’ data to the emissions data, where I now also use the latest BERN multi-model result (see Joos et al. 2013) .
This gives a better agreement with the Dome-C data up until 1950, which now also lies correctly below the cumulative emissions. However the BERN model significantly overestimates CO2 levels thereafter. This demonstrates that merely adding in land use emissions still does not reproduce past CO2 levels. Nor am I alone in discovering this discrepancy.
A recent discussion paper by Millar et al. which uses a fit to ESMs to derive a ‘Finite Amplitude Impulse Response (FAIR) function, also fails to reproduce the actual measured CO2 levels. The AR5-IR model as described in the AR5 report does even worse.
Figure 4a) shows a reconstruction their ‘FAIR’ model and an AR5-IR model integrated using emissions data. The AR5-IR model fails miserably while their FAIR model also overshoots emissions. The actual emissions data used are those shown in Figure 4b) while the blue curve is what hypothetical emissions would have been needed to reproduce the actual CO2 Keeling curve. Their result is essentially the same as mine. A boost in emissions is needed before 1930 to explain the dome-C measurements and the AR5 model overestimates CO2 levels.
A NASA paper recently reported on the CO2 greening effect. So did somehow the global biota switch from being a source of emissions (deforestation) to being a net sink (greening) as CO2 levels increased? Suddenly everyone thinks this was obvious all along.
Prof Corinne Le Quéré, director of the Tyndall Centre at the University of East Anglia said: “Natural vegetation is a fantastic help in slowing down climate change by absorbing about a quarter of our carbon emissions from burning fossil fuels.
Yet it seems clear to me that ESMs did not take sufficient account of any fertilisation effect. Nor did AR5 report any global uptake from CO2 fertilised growth in plants either. It merely refers to possible regrowth from past clearances in specific regions.
The second major source of anthropogenic CO2 emissions to the atmosphere is caused by changes in land use (mainly deforestation), which causes globally a net reduction in land carbon storage, although recovery from past land use change can cause a net gain in land carbon storage in some regions.
Global net CO2 emissions from land use change are estimated at 1.4, 1.5 and 1.1 PgC yr–1 for the 1980s, 1990s and 2000s, respectively, by the bookkeeping method of Houghton et al. (2012)
This makes it clear that any fertilisation effect was also ignored by ESMs. This likely means that IPCC projections of future CO2 levels based on the Bern model are too high.