If you’ve been following this blog in any way, you would probably have noticed that climate data is one of our main concerns regarding the assessment of impacts of climate change on agriculture and ecosystems. We need to know what will happen, and when, in order to focus adaptation and research investments. Governments must be properly informed, and uncertainties must be taken into account.
Despite all those needs, we still have several gaps in climate science, mostly accounting to our still very limited knowledge on the climate system. CIAT’s DAPA is not a meteorological center, so we’re not supposed to develop any global climate model, or to apply it to any area, but only to use outputs of climate models in our impact assessments. Within our work, however, there lies the need of providing high resolution future climate surfaces to analyze future patterns at local scales. Not an easy task, though.
The problem: Providing future climate data at fine scales
The source of future climate data: Global Climate Models (GCMs) are complex algorithms that simulate the earth processes on a global scale, based on a series of driving forces. Each output variable of a Global Circulation Model (GCM), run under a certain emission scenario, has a spatial resolution of 3 to 5 degrees (with two exceptions from Japan and Germany), which means very rough predictions when trying to analyze local impacts of climate change, therefore leading to many questions in terms of the usability and impact of the different GCMs at regional scales regarding policy making processes.
The solution: RCMs
To surpass this bottleneck, climate and meteorological centres that initially developed GCMs have developed the so-called “Regional Climate Models” or RCMs. An RCM is mostly a representation of a particular GCM pattern but in a lower scale. RCMs are, in most cases, as complicated as a GCM, but the difference lies in that an RCM is run under a smaller region, with increased spatial resolution and for specific boundary or driving conditions. This “regionalization” of a GCM involves directly a decrease of about 90% of the spatial resolution. A 50km spatial resolution allows performing analyses in most level-1 administrative divisions but the level-2 administrative divisions remain to be undetermined or in the best case “roughly” determined. Is it, or is it not a solution then?
The problem, again: Reliability, access, and visibility
Several pros and cons can be formulated against RCMs, but, no matter how those arguments stand up, the key issue is: we need access to climate predictions. RCMs have been only run for particular areas, mostly in developed countries (i.e. European countries, Canada and United States). And even for these, the data are not freely available. Only few RCM approaches have been performed in Latin America, and even fewer in Africa and Asia. Most of these approaches make use only of the UK Meteorological Office (UKMO) PRECIS model. Regional predictions of future climate are thus fairly limited for countries for which this kind of information is critical, as are expected (according to the IPCC 4th assessment report) to receive most of the negative impacts. Even supposing that these countries’ meteorological departments and/or offices have the necessary background, training, and processing capacity to run an RCM, they would only be able to run PRECIS.
Empirical or statistical downscaling then turns into a useful tool within the toolkit of impact modellers in developing countries. The low required processing capacity, the quick results that can be obtained, and the applicability of such results at fine scales have made statistical downscaling techniques a widely used approach when creating regional surfaces of future climates. No justification, however, exists, from the meteorological science perspective, to ignore the complex circulation patterns at fine scales (given by RCMs), but from a practical point of view, statistical downscaling provides high resolution and credible surfaces with which complete uncertainty analyses can be performed. One would desire, however, to perform both approaches, but that is simply not possible.
So, does it turn into a fight?
There is no need of a fight between empirical downscaling and Regional Climate Modelling. For lots of researchers in developing regions, and even in developed regions (which may not have access to detailed climatic databases derived from RCM outputs), the usage of empirically downscaled outputs is the only option, and statistically downscaled future surfaces turn into a critical input to impact assessment models and thus for decision making. However, RCM science is still to grow and forecasts will improve as well as accessibility and proper training on the usage or RCM software and data are expected to improve, thus improving impact model results, adaptation measures, policy making and finally, the whole decision making processes in all regions of the world.
PS Ex-DAPA researcher Christian Seiler recently completed a most impressive evaluation of PRECIS for Bolivia. If only this quality level of model development and testing were available for all countries….