The Economic Value of a More Accurate Climate Observing System

Bruce Wielecki, NASA, b.a.wielicki@nasa.gov

Climate change drives a wide range of current and future societal impacts that cross the spectrum of economic activities. Unfortunately, large uncertainty remains in key climate science questions that in turn drive uncertainty in cost/benefit analyses of societal mitigation and adaptation strategies. One of the largest of these factors is the uncertainty in climate sensitivity which remains a factor of 4 at 90% confidence level (IPCC, 2013). Climate sensitivity can be thought of as the volume dial on the climate system: it determines the amount of long term warming that will occur for a given level of radiative forcing from greenhouse gas increase. The amount of warming in turn drives a host of global and regional climate system changes including sea level rise, temperature and precipitation extremes, water resources, and ecosystems. Those climate system changes then drive economic impacts. While economic impacts of climate change including costs of mitigation and adaptation strategies have been studied extensively (e.g. IPCC Working Group II and III reports), little attention has been placed on the economic value of improved climate science. For science, business as usual means doing the best science for the usual societal investment in scientific research. In the U.S., federal government investment in climate science is ~ $2 Billion dollars/year, and has remained constant for the last 25 years when adjusted for inflation (see USGCRP annual reports). Yet the proper economic question to ask is “How much should society invest in climate research?” Such a question falls under the umbrella of research called “Value of Information” or VOI. We will summarize in this article the need for an improved climate observing system, as well as recently documented estimates of such an observing system’s economic value and return on investment.

There are many observations that are used by climate scientists to determine climate change over decades and even centuries. Unfortunately, very few of them were designed with climate change observations in mind. A good example is our weather observing system: with typical temperature absolute accuracy of 0.3K, compared to the desired 0.03K for decadal climate change (NRC, 2015). For many if not most observations, climate change observations would typically require a factor of 5 to 10 more accuracy than weather or process observations including high accuracy traceability to international standards (e.g. SI standards maintained by the international metrology laboratories). A second challenge is that there are roughly 50 essential variables in the climate system (WMO GCOS, 2016) compared to 5 for weather prediction. This large difference is driven by the many complex systems that interact in determining the Earth’s climate system and its impact on society. These include measures of the global atmosphere, ocean, cryosphere, biosphere (land and ocean), land use, land hydrology, chemistry, solar variability and geology including volcanism. Third, weather can be thought of as one small part of the climate system at a subset of climate time scales: those out to a few days as opposed to those including seasonal, annual, decadal, and even century time scales. As a result, climate system observations must deal with much greater complexity, at much higher accuracy, over much longer time scales than weather. The observations must maintain their accuracy and traceability to international standards over decades: times longer than the life of in-situ or even satellite based instrumentation, indeed longer than the length of a scientist’s or engineer’s career.

The challenge of such an observing system far exceeds typical scientific observing systems including those of weather, large particle physics experiments, or astronomy which are some of the largest current scientific endeavors. It is perhaps not surprising that we currently lack such a rigorous detailed and designed climate observing system. Instead we have a collage of weather, resources, and research observing systems that are cobbled together in a heroic effort to study climate change. In some cases like surface air temperature there are 7 different weather observing systems (surface sites, weather balloons, ocean buoys, ocean ships, aircraft, infrared satellite sounders, and microwave satellite sounders) allowing sufficient independence to verify, improve, and eliminate most artifacts that might confound climate change. But for most of the 50 essential climate variables there are at most 1 or 2 or none, leading to major challenges in detecting calibration drifts, changes in instrument design or sampling, or accurately crossing gaps in observations that may last several years.

There are many national and international documents that discuss the shortcomings in our current climate observations (Dowell et al. 2013; WMO GCOS, 2016; Weatherhead et al. 2017, NASEM, 2018, NRC 2015, Trenberth et al., 2013). But the bottom line remains that we lack a rigorous designed and maintained climate observing system. In the most recent U.S. National Academy Earth Science Decadal Survey (NASEM, 2018), an examination of over 30 quantified and prioritized climate science objectives shows that critical observations are missing for 80% of the “Most Important” climate science objectives, 71% of “Very Important” objectives, and 47% of “Important” objectives. Critical observations are missing for roughly 2/3 of all climate science objectives. See Chapter 9 and Appendix B and C of the report for details (NASEM, 2018).

How would one design a rigorous international climate observing system? Discussion of this topic can be found in recent Academy of Science reports: the “Continuity Report” (NRC, 2015), and the Earth Science Decadal Survey (NASEM, 2018). An overview of the topic is also discussed in a recent journal article in AGU Earth's Future (Weatherhead et al. 2017). We give a summary of key points in the list below.

• Use of quantified climate science objectives based on major national and international reviews and reports such as the IPCC and USGCRP reports. Examples would be to narrow the uncertainty in long term climate sensitivity or aerosol radiative forcing by a factor of 2. Or to reach a specific level of accuracy in the rate of global and regional sea level rise. See a wide range of examples in the 2018 Decadal Survey (chapter 9 and Appendix B of NASEM, 2018).
• Rigorous quantitative requirements for instrument accuracy, sampling accuracy, and remote sensing retrieval accuracy sufficient to eliminate large delays in quantifying climate change trends. Observing system lack of accuracy increases trend uncertainty beyond the minimum caused by climate system internal natural variability. This increase typically extends the time to detect climate change trends by decades. (NAS, 2015, Leroy et al. 2008, Wielicki et al. 2013, Trenberth et al 2013).
• Improved use of Observation System Simulation Experiments (OSSEs) to quantify the utility of a given observation to reduce scientific uncertainty in past and future climate change (NRC 2012, NASEM 2018, Weatherhead et al. 2017).
• Traceability of instrument observations to international (SI) standards to enable removal of calibration drifts and the ability to rigorously deal with data gaps. This is especially critical for space based observations which provide many of the global climate change data sets. (NRC 2007, NASEM, 2018)
• Provision of a much more complete set of climate system observations based on quantified climate science objectives, which currently suggest that critical observations are missing for 2/3 of all climate science objectives in the recent 2018 Decadal Survey report. GCOS implementation plans provide definition of the 50 essential climate variables (WMO GCOS, 2016).
• Follow existing GCOS observing principles (WMO GCOS, 2016)
• Provide independent observations of all essential climate variables (instruments, techniques, systems) to allow verification of climate system surprises after they occur.
• Provide independent analysis (methods, research groups) of all essential climate variables. Almost all computer code has errors, but independent development of analysis systems will have different errors, thereby allowing comparisons to discover and correct issues.

The above list indicates that major improvements are needed in climate system observations: both long term climate change and climate process observations. Some of the advances would simply require more rigorous processes than currently employed (independent analysis) while others would require improved global sampling, or design of instrumentation with more accurate traceability to international standards. In many cases more complete observations would require application of new technologies such as space based advanced lidar (wind profiles, aerosols, clouds, ocean phytoplankton), radar (rain, snowfall, convective vertical velocities) and radio occultation temperature profiles. New technologies for in-situ observations would also be key: such as adding chemistry measurements to deep ocean floats, and increasing the depths the floats reach.

How much would such an observing system cost? Adding independent observations, independent analysis, higher accuracy, and more complete observations might triple the cost of current global investments in climate research (observations, analysis, modeling, data storage/archive/distribution). Building a global climate observing system will also require increased investment in the data analysis, climate modeling, and data stewardship needed to benefit from such a system. The total of global climate research investments are currently estimated at 4 billion/yr, so that an additional 8 billion/yr might be required. The investment would be required for many decades (at least 30 years) because of the intrinsic long term nature of climate change itself. Once built, however, efficiencies of reproduction and scale might decrease costs over time for the basic instrumentation which is one of the largest costs.

Given that 8 billion/yr is a significant global investment, how could we estimate what the return on that investment might be? Four recent research papers (Cooke et al. 2014, 2016, 2019; Hope 2015) have estimated that economic value and concluded that through 2100 it ranges from 5 to 20 Trillion U.S. dollars. The cost of tripling the global investment in climate research (including development of the more rigorous climate observing system above) was estimated to provide a return on investment of roughly 50 per dollar invested (Cooke et al. 2014). All economic values are given in net present value using a discount rate of 3% (The nominal value from the U.S. Social Cost of Carbon Memo, 2010, hereafter SCCM2010).

As scientists, how do we understand such large economic value and return on investment estimates? We first need to consider some basic economic concepts. We begin by scaling the magnitude of global gross domestic product or “global economy” as roughly 85 Trillion U.S. dollars. Second, “business as usual” carbon dioxide emissions are predicted to cause climate damages in 2050 to 2100 that range from 0.5% to 5% of GDP annually (SCCM2010). Such damages would range from 400 billion to 4 Trillion per year. The large range is to first order because the uncertainty in climate sensitivity remains a factor of 4 at 90% confidence level (IPCC, 2013, SCCM2010). Climate sensitivity measures the amount of global temperature change per unit change in atmospheric carbon dioxide. A range of economic impact studies conclude that impacts rise roughly as the square of the amount of global temperature change (SCCM2010). The economic value of narrowing the uncertainty in critical issues like climate sensitivity as a result are very large (Cooke et al. 2014, 2016, 2019; Hope 2015).

Relating the economic value of benefits that return in the future to alternative investments that could be made requires the use of a concept called Discount Rate. All future benefits are discounted X% per year to account for the fact that most people would prefer to have money now vs the future, and to allow comparison of how the same funds could be invested in alternative investments, including those with short term goals. The nominal discount rate used for long term climate change is 3% (SCCM2010) but arguments have been made for both lower values at 1.5% (Stern, 2008), or higher values at 5%. Using the nominal 3% discount rate, an investment that pays back in 10 years is discounted by 1.03^10 or a factor of 1.3, 25 years by a factor of 2.1, 50 years by a factor of 4.4, and 100 years by a factor of 21. This makes it obvious that discount rate is very important to such calculations, and that paybacks 100 years in the future are negligible. For climate change returns on investment, discount rate is then used to derive the Net Present value by discounting any return by the number of years into the future that it will be realized. There is another way to think about discount rate and why 3% might be a reasonable value for global issues such as climate change. The growth rate of global GDP averages about 3% and has so for a long period of time. Therefore discounting at 3% per year also provides a reference to returns that are above those expected for global average GDP increase.

Now that we have a few basic concepts in mind, Figure 1 provides a schematic for the economic value of information (VOI) estimates in the Cooke et al. papers (2014, 2016, 2019). This figure shows the methodology for converting improved climate science knowledge into economic value. The blue boxes at left gives the baseline condition with Business as Usual greenhouse gas emissions (e.g. SCCM2010), which through climate sensitivity lead to the baseline amount of climate change, which in turn leads to the baseline amount of economic impacts. This is the state with no or modest societal action on climate change. Meanwhile society (and scientists) are looking through 3 fuzzy lenses at climate change: the first fuzzy lens is that of natural variability of the climate system such as swings between warm and cold phases of the ENSO cycle or the Arctic Oscillation or the Pacific Decadal Oscillation. All these are examples of internal variability of the climate system itself and represent noise that we must detect human climate signals against. Even a perfect observing system cannot eliminate this fuzzy lens. The second fuzzy lens is the fact that our climate observations are themselves inaccurate whether through calibration, sampling, or through weak relationships to the climate variable desired (e.g. indirect proxy observations). When added to the fuzzy lens of natural variability, these observing system uncertainties can delay the time to detect climate trends by 5 to 50 years (Leroy et al. 2008, NRC, 2015, Wielicki et al. 2013).  This large information time delay is the factor in societal decisions that improved accuracy in our climate observations can directly impact.  The third fuzzy lens is that of climate model uncertainty.  Climate models are used to predict the change that will occur under a range of proposed emissions scenarios (e.g. weak, moderate, or strong greenhouse gas emissions policies).  But those models are imperfect and currently show a range of a factor of 4 uncertainty in climate sensitivity (IPCC, 2013).  Improving the models requires both improved climate process observations for driving model uncertainties (e.g. aerosol forcings, cloud feedbacks, glacier melt) as well as improved long term decadal observations of climate change to verify model performance and uncertainties.  As a result, this fuzzy lens can also be improved through a more rigorous climate observing system.