Where do Modeling Requirements Come From?

February 19th, 2008 <-- by Richard Rood -->

Requirements vs Requirements of scientists

I sit in my share of meetings on models and modeling. I listen to plans about model development and impassioned statements of the importance of “the science.” There are struggles on how to make the interface to other communities, the proverbial policymaker. In a room full of scientists they always come around to the need to follow “the science.”

What does it mean to follow “the science?” Science is a process of investigation – a method. It is one of several ways that we generate and accumulate knowledge.

Following the science always generates a number of research projects, which usually fall on two paths. There is the path of inclusion – adding additional processes to models, for example, adding land-ice parameterizations, making the carbon-cycle interactive or improving the radiative treatment of aerosols. There is a path of higher resolution and increased rigor and accuracy in a component model like the atmosphere; for example, moving to algorithms that represent the non-hydrostatic processes in the atmosphere and resolve the behavior of cloud systems. This is the path of increased fidelity to the first principles of physics – or chemistry, or biology. There is always tension between increased fidelity and more inclusion. The tension is related to both limited computational and intellectual resources.

The motivations that drive the advocacy for these different paths are well grounded. In general, a group of priorities rise to the top, and they address, demonstrably, important problems. The importance is determined, often, by uncertainty. For example, the uncertainty associated with the model’s ability or inability to represent clouds. Since increased clouds provide a cooling term, determination of changes to clouds are important to knowing how fast the planet will warm.

Scientists identify the uncertainties and the source of the uncertainties. They develop strategies to address these uncertainties. The determination of priorities to address these uncertainties is an imperfect process, and the “requirements” appear as a list of important problems. Outside of the science community, the value of addressing these uncertainties might be unclear.

The publication of the 2007 IPCC report fundamentally changes the demand for climate information by society. This challenges the traditional approach of developing requirements for model development based on the uncertainties of climate predictions. The result of this challenge will be two paths of scientific investigation: a traditional path of basic research that is focused on addressing the greatest uncertainties in climate predictions, and a second path that develops science-based climate information based on requirements that come from a wide variety of applications. This second path is applied research.

To some extent, both paths currently exist, but the exploding demand for climate information will amplify the applied research path. For a given application, for example the impact of the development of a large corn-based ethanol energy capability, it is possible to analyze the impact on water resources and carbon balances. An analysis can be made from existing data bases of observations and simulation experiments. But this is a process that relies largely on using information that was developed to perform basic research of the climate system. The details of the application extend the information in these data bases beyond the purposes for which the information was developed. Spatial and temporal resolution are, perhaps, too coarse; the impact of urban environments and water engineering not adequately considered. It would be possible to design numerical experiments with existing models and computational resources that would provide science-based investigation with, potentially, much more robust information for the application at hand.

The process of developing models to address the uncertainties in the climate system, the requirements of scientists, are often of little relevance to addressing the applications that are important for adaptation to climate change. Improvement of the representation of marine stratus clouds, a priority in improving our understanding of the climate system, will not be consequential to the water resource manager in the western United States. ( blog on role of uncertainty )

The time scale for the development of policy is, now, a small number of years. The life time for energy infrastructure is a small number of decades, and decisions on the nature of expenditures in energy infrastructure are needed today. The development and implementation of strategies to manage water in an environment where less water is stored in ice and snow are already under consideration. ( California water and climate change ) The design and funding of adaptation plans for societal infrastructure near sea level is imperative. ( Impacts of Climate Variability and Change on Transportation Systems and Infrastructure — Gulf Coast Study ) The information needed for these decisions will be demanded on time scales that are much shorter than the development cycles of climate models that are pursuing traditional scientific development paths.

It is imperative that the climate science community develop the capabilities to provide the best science-based answers to these externally posed applications at any given time. A natural response to this demand for information is the development of a climate service, perhaps in the spirit of the National Weather Service. In the U.S. this approach to providing environmental information has often led to the dichotomy of research versus operations. In this dichotomy the flow from research to operations is inefficient. We need a more modern approach.

This is a blog, not an essay … stopping here – looking for reactions.


Figure 1: From the European Sea Level Service , a focused collaborative organization.

2 Responses to “Where do Modeling Requirements Come From?”

  1. David B. Benson Says:

    By following


    for some time now, and more recently also


    it casually appears that, for better or worse, tens of millions of dollars are being invested every day in various bioenergy projects around the world. So some investors have decided they have enough information right now to sign agreements and start building toward production.

    On the other side of these isssues, various climate modelers are dissatisfied with the regional forecasts the GCM climate models produce and would like to do better, although they complain that progress is rather slow.

  2. Søren Kjær Vestergaard Says:

    Thank you for an insightful article. I have seen though that any climate model seems to come up with good results when there alghoritms are based on thourogh thinking. I have speculated about the commen sence in the arguemnt that good theory always generate more good theories. In the end of creating theories one may ask: “Is this a good Unvierse” (A. Einstein). I see that we get the results we are asking for and this is the great truth of the Universe that can serve of well when we want to change to world to the better.

Leave a Reply