It makes sense for an industrialised country, with a relatively small population, a high annual rate of carbon dioxide emissions per person, and a very large land area, to seek ways of using the land as a natural carbon store. Hence the interest in ‘carbon farming’ as a way to compensate for some of the CO2 we emit into the atmosphere. The CO2 that Australia emits mixes with that emitted by other nations to ramp up the atmospheric concentration of CO2 year on year, thus contributing to a global and inexorable change in climate. Can we at least reduce our net emissions of CO2 by storing some extra carbon in vegetation and soils?
The idea presents challenges for ecosystem science – and for the scientists at e-MAST who are developing infrastructure that will help Australians work out the best solutions for several problems about living with climate change. The director of the facility, Professor Colin Prentice, explains the team’s approach.
First, the principles involved need to be clear – in much public discussion they aren’t – and second, we need to get a reasonably good grasp of the numbers. How much carbon can be stored, where, and for how long? How much difference will it make? This information will be needed equally by stakeholders with interests at different scales: from landowners responding to incentives to include carbon farming in their portfolio of land uses, to the Commonwealth government as it assesses climate policy in the context of international negotiations.
Let’s consider the principles. These are some of the most basic. One, the physical environment determines how much carbon can be stored in vegetation and soils. Vegetation needs light, warmth, water and nutrients. Australia has plenty of the first two, but large areas are deficient in the others. The environment also determines wildfire risk – there is no way that we can completely avoid fires – and this factor also restricts the amount of carbon that can be kept sustainably in ecosystems.
Two, on the other hand, land use over the years has depleted the natural stocks of carbon in vegetation and soils. So there is a certain scope to put back some of this lost carbon by planting forests, and also by adopting farming practices that encourage soil carbon to increase. We can think of this as a ‘buffer’ that could be refilled. It’s important to realise that it’s finite (we can’t offset fossil-fuel CO2 emissions forever!) but it’s there, nonetheless.
Three, there is a trade-off between storing carbon and other uses of land. For example, converting a parcel of land from crops to forestry certainly increases its average carbon storage, but if the forest is regularly logged, the increase is modest. And there are good reasons why large areas of productive farmland are not going to be turned back into natural woodlands or forests. It makes better sense to think about carbon farming on degraded natural lands.
Now let’s look ahead, and consider the kind of tools that will be needed to assess how much carbon can be stored and where. We need a ‘calculator’ – a computer model. This model should be able to predict the current rates of exchange of carbon dioxide (CO2 being converted to biomass, and biomass decomposing and turning back to CO2, the processes called net primary production and heterotrophic respiration), and the amount of water used by the ecosystem in the process, for any given ecosystem and land use. The model should also be able to predict how much carbon the ecosystem contains, estimate how much carbon it could contain (taking into account changes over time in atmospheric CO2 concentration and climate), and thus indicate the potential for carbon storage to increase. The model would also predict how fast the carbon store would grow, its potential vulnerability (for example to a change in the fire regime), and the extent of its impact on water resources.
Ecosystem modellers have been doing this kind of thing for decades, but there are some long-standing problems. Either models tend to be made for very local applications (in which case they can’t be used with confidence outside the context for which they were developed), or they are global models, most of which are pretty approximate – they haven’t been designed to answer these kinds of questions, and when they have been tried out in the Australian context, they haven’t done too well. For example, the last time anyone asked the simple question, ‘What is the total annual net primary production of Australia?’ to a series of computer models, the answers varied by a factor of five!
Data at different scales the key
This might sound discouraging, but fortunately there is a great opportunity to do far better. The key is that now we have data at many scales, from measurements of CO2 in the atmosphere (which ‘see’ the exchange of carbon between the land, ocean and atmosphere over millions of square kilometres), to satellite measurements of topography and ‘greenness’ that have a resolution to 250 metres or better.
There are many kinds of data that can determine where models are doing well or badly, and help to improve them. When the current generation of ecosystem models was developed, most of these data didn’t exist. Now TERN is gathering together some of the main datasets needed to make models work well for Australia. They include the many satellite-derived products providing data on land cover and greenness, being put together by AusCover; flux-tower measurements of the land-atmosphere exchanges of CO2 and water in different types of ecosystems, being collected by OzFlux; and recording of ecosystem composition and properties on the ground, done by the Multi-Scale Plot Network.
E-MAST is responsible for integrating these and other datasets into a software framework that will make it easy for ecosystem modellers to test and improve their models. It’s the lack of this infrastructure, whether nationally or globally, that has hindered modellers from doing so, and stopped them from making good use of the wealth of observations that now exist. Beyond TERN data, e-MAST is bringing in extremely precise data on regional CO2 concentration, and streamflow data – these too are relevant to ecosystem models. They are already being collected, for other purposes, by various branches of CSIRO and the Bureau of Meteorology. The latter also provides the primary weather data that are needed to drive the models. E-MAST is all about creating a common framework for these different data sources, and a simple way that ecosystem models can be ‘plugged in’ to the framework.
There is no shortage of ecosystem modelling expertise in Australia, and e-MAST’s role is to provide the infrastructure needed for models to become more accurate, and more useful, at scales ranging from the individual hill slope up to the whole continent.
Water, biodiversity and land use important, too
E-MAST isn’t only about carbon: to give reliable enough results to be useful it must also consider water, biodiversity and the effect of changes in land use on climate. Carbon and water go together because for every molecule of carbon dioxide taken up into a plant leaf, a certain number of molecules of water are evaporated from that same leaf. So ecosystem models have to predict the exchanges of water as well as CO2 – and they can be used to look at how a change in land use, or climate, will affect river flows as well as carbon stores.
How climate change is likely to affect biodiversity is a huge topic that scientists and policy makers still don’t know enough about. We need to know more, if we are to make effective conservation policies recognising that the climate in 20 or 50 years’ time simply will not be the same as it is now. E-MAST can help. For example, we model the bioclimatic constraints on the growth and survival of different types of plants, and how changes in atmospheric CO2 can have additional effects – such as helping to mitigate the effects of drought. There’s a lot more to be done in this area. It would bring together the biodiversity conservation community with the ecosystem research and science community, and build on other TERN activities, such as ÆKOS.
One area that isn’t usually considered by ecosystem scientists is the effect of changes in the land surface on the climate. This is important, because, for example, planting forests doesn’t just ‘use’ more water – it also means that more water is being evaporated into the atmosphere and some of this may re-appear as increased rainfall. We really don’t know the balance of these effects. But Australia does have a world-class team, mainly at CSIRO and BOM, working on the CABLE land-surface model – the kind of model that works at the interface of ecosystems and climate modelling. E-MAST has close links to the CABLE team and hopes that the progress made in e-MAST will help this area of science as well.
The e-MAST team brings to bear skills from several disciplines to build the sophisticated infrastructure needed for accurate modelling. There are four chief investigators. Colin, who is based at Macquarie University, is an expert on the global carbon cycle who has been in the forefront of the international development of large-scale land ecosystem models for over 20 years. Professor Mike Hutchinson, at the Australian National University, devised the methods used worldwide to convert weather station data to topographic grids, and is working in e-MAST to provide the best possible climate data at 1 km resolution based on the data provided by BOM. Dr Helen Cleugh, the Deputy Leader of CSIRO Marine and Atmospheric Research, is an expert in land–atmosphere interactions, and heads the laboratory that hosts both AusCover and OzFlux. Professor Damian Barrett, at the University of Queensland and CSIRO Land and Water, is developing advanced data-assimilation techniques that use different satellite products together with modelling to quantify important physical properties of ecosystems. Also involved are Ben Evans of the National Computational Infrastructure, and Dr Gab Abramowitz, University of New South Wales, whose ground-breaking PALS (Protocol for the Analysis of Land Surface models) software forms the initial basis of the e-MAST system to bring models and data together.
Published in TERN e-News July 2012
