Socioeconomic Impacts of Global Warming are Systematically Overestimated

Socioeconomic Impacts of Global Warming are Systematically Overestimated

Part I: Why are Impacts Overestimated?

Indur M. Goklany

[Note to the Reader: For the sake of argument, in this post I will accept the IPCC’s estimates of global warming. I will show that even if one takes those estimates for granted, the impacts of global warming are, nevertheless, overestimated.]

Most of the scientific debate on global warming has focused on “climatological” issues that have been the province of IPCC Work Group I’s Science Assessment. However, there are even greater grounds for skepticism when it comes to estimates of the impacts of climate change, which is the monopoly of IPCC’s Work Group II, not least because these estimates are based on a chain of linked models with the uncertain output of each unvalidated model serving as the input for the next unvalidated model. [Yes, it’s that bad!].

The first link in this chain are emission models which use socioeconomic assumptions extending 100 or more years into the future to generate emission scenarios, which strains credulity. As Lorenzoni and Adger (2006: 74) noted in a paper commissioned for the Stern Review, socioeconomic scenarios “cannot be projected semi-realistically for more than 5–10 years at a time.”

The results of these emissions models are fed into coupled atmosphere-ocean general circulation models (AOGCMs) to estimate spatial and temporal changes in climatic variables (such as temperature and precipitation) which are, then, used as inputs to simplified and incomplete biophysical models that project location-specific changes in biophysical factors (e.g., available habitat, or crop or timber yields). Notably, the uncertainty of estimates of climatic changes increases as the scale at which they have to be specified becomes finer. This is particularly true for precipitation, which is a — if not the — critical determinant of natural resources (e.g., water and vegetation) that human beings and all other living species depend on either directly or indirectly. As the US Climate Change Science Program (CCSP) review, Climate Models: An Assessment of Strengths and Limitations, notes:

“Climate model simulation of precipitation has improved over time but is still problematic. Correlation between models and observations is 50 to 60% for seasonal means on scales of a few hundred kilometers.” (CCSP 2008, p. 3).

“In summary, modern AOGCMs generally simulate continental and larger-scale mean surface temperature and precipitation with considerable accuracy, but the models often are not reliable for smaller regions, particularly for precipitation.” (CCSP 2008, p. 52).  [Emphasis added.]

In colloquial English this means that the AOGCMs have not been validated for less-than-continental or less-than-regional scales because they are unable to reproduce historical temperatures and precipitation simultaneously (and, moreover, cannot endogenously reproduce major climatic features such as the ENSO, PDO, etc.).  But the real world distribution of climate-sensitive resources and climate itself is heterogeneous and varies considerably on “scales of a few hundred kilometers.” Therefore, we necessarily should be using finer scale models to estimate impacts on these resources.

No matter, depending on the human or natural system under consideration, the outputs of the biophysical models (which also are not generally validated; see Nogues-Bravo 2009) may have to be fed into additional models to calculate the social, economic, and environmental impacts on those systems.  Ideally, the outputs from this set of models should be fed back into the emissions models, thereby closing an iterative loop of models. But models have, so far, not yet incorporated this feature.

Notably, I have never seen an end-to-end analysis of the uncertainties/confidence limits associated with impacts estimates derived from the entire chain of models at relevant scales (including uncertainties associated with the basic assumptions feeding the emissions models). I have often wondered why such a step, that should be fundamental to any scientific analysis, is ignored.

In any case, this post will not deal with the level of confidence or uncertainty attached to impacts estimates but with reasons why impacts estimates are systematically overestimated.  Also, this post will not address potential “catastrophes,” i.e., low-probability but potentially high-consequence outcomes such as a shut down of the thermohaline circulation or the melting of the Greenland and Antarctica Ice Sheets. These are deemed unlikely to occur during this century, and are grist for other post(s).

The major reason for systematic overestimation is that the magnitude of future damages depends critically on society’s future “adaptive capacity” — a fancy word for “adaptability.”  But the methodologies used to estimate impacts underestimate individuals’ and society’s future capacity to make self-directed (or autonomous) adaptations to global warming.  [Adaptations should include measures to either reduce any adverse effect of global warming or take advantage of any of its positive impacts.]

Figures 1 and 2, based on cross-country data, show how two climate-sensitive indicators of human well-being — cereal yield and available food supplies per capita — have improved with wealth and time (a surrogate for technology). This makes perfect sense since wealthier societies ought to be better able to afford technologies that can enhance crop productivity (Figure 1). And if that is insufficient to meet food demand, wealthier societies also ought to be able to purchase the food supplies they need (Figure 2) via internal or external trade.  Not surprisingly, hunger and malnutrition are lower in wealthier societies.

Figures 1 and 2 also show that the crop yield and food supply curves shift upward with time, that is,  for any given level of GDP per capita, crop yield and food supplies increase as time marches on. This can be attributed to the secular change in technology which accretes over time, and is defined broadly to include both hardware (e.g., catalytic convertors and carbon adsorption systems) and software technologies (e.g., knowledge, policies and institutions that govern or modulate human actions and behavior, culture, management techniques, computer programs to track or model environmental quality, trading).

The patterns indicated in Figures 1 and 2 hold for virtually all objective determinants of human well-being — hunger, malnutrition, mortality rates, life expectancy, the level of education, greater access to safe water and sanitation.  See here and here.  All improve along with the level of economic development and time/technology, at least until they approach “saturation” (which helps accounts for the shape of Figure 2).  Similarly, spending on health care and research and development tends to go up with GDP per capita. Notably each of these determinants helps boost economic and technological development, and human and social capital (see, e.g., here and here), which themselves boost adaptive capacity. Therefore, in the future, as time marches on — and if societies become wealthier — as is assumed under all IPCC emission and climate scenarios, their adaptive capacity ought to be higher, and the net damages from global warming should be correspondingly lower.

Adaptive Capacity in Global Impacts Assessments in the IPCC’s Latest Assessment

To date, however, no impact study has fully accounted for both increasing wealth and secular technological change, as will be discussed in greater detail below. Consequently, they all overestimate the negative impacts and underestimate the positive impacts. Consider, for example, the suite of studies of the global impacts of climate change sponsored by the UK Department of Environment, Food and Rural Affairs (Defra) and undertaken by a host of authors intimately involved in the IPCC’s various assessment reports (Parry 2004;  Global Environmental Change, Volume 14, Issue 1, pp. 1-99; IPCC WGII, AR$, Ch. 2). These studies were state-of-the-art at the time the IPCC’s Fourth Assessment Report was compiled. However:

• The water resources study (Arnell 2004) totally ignores adaptation, despite the fact that many adaptations to water related problems, e.g., building dams, reservoirs, and canals, are among mankind’s oldest adaptations, and do not depend on the development of any new technologies (see here, pp. 1034–35). While, arguably, this may be acceptable methodology for an academic paper, it is simply inadequate to use “as is” to inform policymakers.
• The study of agricultural productivity and hunger (Parry et al. 2004) allows for increases in crop yield with economic growth due to greater usage of fertilizer and irrigation in richer countries, decreases in hunger due to economic growth, some secular (time-dependent) increase in agricultural productivity, as well as some farm level adaptations to deal with climate change. But these adaptations are based on currently available technologies, rather than technologies that would be available in the future or any technologies developed to specifically cope with the negative impacts of global warming or take advantage of any positive outcomes (Parry et al., 2004, p. 57; and here, pp. 1032–33).  However, the potential for future technologies to cope with climate change is large, especially if one considers bioengineered crops and precision agriculture (see here, chapter 9; and here, pp. 292-93).
• Nicholls (2004) study on coastal flooding from sea level rise takes some pains to incorporate improvements in adaptive capacity due to increasing wealth. But it makes some questionable assumptions. First, it allows societies to implement measures to reduce the risk of coastal flooding in response to 1990 surge conditions, but not to subsequent sea level rise (Nicholls, 2004, p. 74). But this is illogical. One should expect that any measures that are implemented would consider the latest available data and information on the surge situation at the time the measures are initiated. That is, if the measure is initiated in, say, 2050, the measure’s design would at least consider sea level and sea level trends as of 2050, rather than merely the 1990 level. By that time, we should know the rate of sea level rise with much greater confidence. Second, Nicholls (2004) also allows for a constant lag time between initiating protection and sea level rise. But one should expect that if sea level continues to rise, the lag between upgrading protection standards and higher GDP per capita will be reduced over time, and may even turn negative, if that seems warranted.  That is, adaptations would be anticipatory rather than reactive, particularly, for a richer society. Fourth, Nicholls (2004) does not allow for any deceleration in the preferential migration of the population to coastal areas, as might be likely if coastal flooding becomes more frequent and costly (see here, pp. 1036–37). [FWIW, New Orleans population continues to be below pre-Katrina levels, and Florida has been losing population in recent years – of course the risk of floods and hurricanes are hardly the only determinants of migration.]
• The analysis for malaria undertaken by van Lieshout et al. (2004) includes adaptive capacity as it existed in 1990, but does not adjust it to account for any subsequent advances in economic and technological development. There is simply no justification for such an assumption, particularly since there were older papers in the open literature that had pointed that adaptive capacity was a critical element in determining impacts (see here, here, here). Yet this study passed peer review!!!

In my next post, I will look at what can be said about future adaptive capacity, and show that it has been grossly underestimated in impacts studies.