My study in 2012
Why Are We Still Not Good at Understanding of Climate Feedbacks? : Causes and Possible Way of Solution
Kishore B. Ragi
Department of Chemistry
,
National Institute of Technology, Rourkela, India.
Email: kishoreragi@gmail.com
Understanding of feedbacks in the climate system has become one of the toughest challenges for the climate scientific community. Although climate predictions have been started in nineteenth century, active research in this area has begun a little earlier than mid twentieth century right after the invention of computers with a bad move. In this paper, we review how the climate research is evolved, progressed and what is lost in understanding feedbacks. It is known that no original origin for the climate science, but it has been evolved from advanced research of weather forecasting. This made the climate community to get adjusted with the knowledge of weather forecasting and get go ahead with it. The interesting part of this paper will be to explain how to tackle with this problem of feedbacks of complex system. This paper presses the retrospection of climate modeling to get good at understanding of feedback systems.
Introduction:
Complexity systems are of three kinds. They are Chaotic Systems, Complex Adaptive Systems (CAS) and Non-linear Systems. Earth system is a highly non-linear, which is due to non-linear interactions among a number of physical processes within the components of climate system. Interaction between components further increases the non-linearity. Most of the times, these interactions are cyclic in nature (Alan Robock, 1984). These cyclic processes may either enhance or dampen one of the processes in the cycle. We can observe hundreds of cycles in the climate systems. These cycles technically termed in climate science as feedbacks. One of the processes in any feedback cycle may enhance or dampen the severity of the particular process. If the process is being enhanced by feedback loop, then it is called positive feedback whereas the opposite is termed as negative feedback.
Climate system is so unpredictable as it has too many feedback loops. Unpredictability is further increased by the nature and extent of feedback. To study such systems is not impossible but it is a daunting task. Scientific community has chosen a number of paths to un-entangle the system but unfortunately, from the review on global surface temperatures by R. Knutti et al,2008, not much has been done so far to feedback study. Before going into the possible solution to the feedback study, it is needed to understand where and if the problem is in the study so far.
Ellipses ----- Components of Climate System
Blue-dots ----- Physical processes in Climate System
Blue-lines ----- Interactions between physical processes within component
Red-lines ----- Interactions between physical processes between components
Fig. 1 Schematic representation of physical processes and feedbacks within and in between the components of climate system
In Fig.1, ellipses represent components in the climate system, blue dots are physical processes, and blue and red colored solid-lines indicate interaction between physical processes within the same component and among the components of climate system respectively. It is a fact that real climate system is much more complex than it is represented in fig.1. Many more physical processes and few more components will come into picture when considering the real climate system. Interaction between various processes of climate system is called climate feedback, where in the result of one process triggers changes in another process that in turn will influence either the initial one or the other interacting process. It is obvious from fig.1 that this process may be influenced by other process and this chain will continue. Hence, feedback mechanism is highly complex one. It may be imagined that climate system has too many feedback loops that can either amplify or dampen the effect of climate change. A cycle in fig.1 is one feedback loop and we can see several hundreds of cycles in figure and many more in real system. It is also evident that anyone cycle is not independent with other physical processes. Hence, it is impossible to quantify any individual feedback loop. However, it is said to be a positive feedback, if over-all climate system amplifies the effects of climate change whereas it is a negative feedback, if the system diminishes the effects of climate change.
Origin and Evolution of Climate Science:
Not until 1922, the modeling of Earth system was done. Richardson in the same year, tried hard to model the weather numerically with paper and pencil calculation. As it can be speculated, the work he had done was completely failed due to being unable to calculate accurately with hand calculation which is an impossible task. After, almost three decades, when the invention of computers in 1940’s in 1950, Phillips constructed a moderately successful numerical weather prediction model on an ANIAC computer. What he predicted was a 24 hour forecast in 24 hours using the slowest computer. It is obvious that lack of computer power made the model worthless.
However, by the inspiration of Richardson and Phillip work , and growing the power of computers, Kirk Bryan and Syukuro Manabe in 1964 ( Bryan et al, 1964) constructed an Atmospheric Global Climate Model (AGCM) in Geophysical Fluid Dynamics Laboratory (GFDL), Princeton University. After three years, Kasahara et al, 1967 a little advanced a Global General Circulation Model of atmosphere at National Center for Atmospheric Research (NCAR), USA. Anyway, these two models could perform few tasks related to atmosphere but failed to model climate because climate has strong interaction between physical processes of oceans. This idea of interaction of oceans with atmosphere led those same scientists, Syukuro Manabe and Kirk Bryan in 1969 to have done climate calculations using their own coupled Ocean-Atmosphere Model (Manabe et al, 1969 ). Manabe et al, 1969 used slab-ocean component in which only shallow oceanic processes were included, leaving the physics of deep oceanic processes. However, Kirk Bryan et al, 1975 developed the ocean circulation part of Global Ocean-Atmosphere Climate Model in which atmospheric part is improved by Manabe et al, 1975.
An excerpt from McGuffie and Henderson-Sellers, 2001: “The first atmospheric general circulation climate models were derived directly from numerical models of the atmosphere designed for short-term weather forecasting. These had been developed during the 1950s (e.g. Charney et al., 1950; Smagorinsky, 1983) and, around 1960, as advances in computer technology allowed more extensive simulations, ideas were being formulated for long enough integrations of these numerical weather prediction schemes that they might be considered as climate models. Indeed, it is rather difficult to identify the timing of the transition from weather forecasting to climate prediction in these early modelling groups. The numerical requirements of weather prediction were extended to hemispheric domains (global calculations were not introduced until later) and the extension to longer integration periods sometimes became simply a matter of availability of computer resources. Indeed, to this day, climate modelling and numerical weather forecasting groups co-exist, especially in national meteorological bureaux. However, the needs and focus of the two disciplines differ: for example GCMs have to conserve mass, energy and moisture, while many forecast simulations are over too short a period for conservation to be an issue”. It is documented by Steve Easterbrook, 2010 that HadGEM3 Global Circulation climate model is still using Hadley Centre weather forecasting atmospheric circulation model.
However, all climate models from simple to most complex GCMs simulate past climates reasonably well and hence will do well for the present and future. It is obvious that all models are to be validated to simulate to climate observations, in spite of having a lot of defects inside the theatrical model. Most of climate research is going on validation of climate models but not the verification of the same.
Anatomy of Climate Models
Global Climate Models (GCMs) are the best tools we have gotten now to study our climate system. It is common practice to add new processes, modules or even components to existing climate models to better predict climate. But, unfortunately, with lack of considerable amount of expertise in computer science knowledge for climate scientists, it is not of practice to go back into the climate model code and analyze to check if it is working properly ie the validation of models. We press the readability of code in this paper to analyze code to draw a diagram like the Fig. 1. Readers are encouraged to see Kaitlin Alaxander and Steve Easterbrook, 2010 for anatomy up-to the level of components of few of the climate models used in IPCC fourth assessment report.
Retrospective Method:
By now, climate models have grown very complex. Still a lot of work is under-going to make those models much more complex. As a result, it became a tough challenge to look into the model code to see if what is happening inside the code. Our interest in this paper is to propose a method to understand feedback processes in the climate system. As we mentioned in the previous section, it is impossible to quantify the feedback processes as they are inter-related with several processes in climate system. This proposed retrospective method will try to make the sense how each and every variable is behaving in the climate system. Retrospection is in the sense of complexity of climate models. This method analyses the code to see how the climate variables and hence the processes are interacted with one another. Analysis of code will give a picture like figure-1. Now, the climate sensitivity, which is a metric used to characterize the response of global climate system to the forcing is estimated by running the model with doubling of CO2 (Cubasch et al, 2001, Roe and Baker, 2007, R. Knutti et al, 2008, Baker and Roe, 2009).
It is cumbersome task to understand whole source code of such big models like GCMs, so we suggest starting with stand-alone components, keeping the issues of decoupling. In this method, it is tried to make the component system simple a bit by bit up to relatively much simpler one. For example, when atmospheric component is considered to study, we analyze the code using many source code analyzing tools out there such as understand for Fortran, Source-Navigator, Code Visual to Flow Chat, etc. This reverse engineering is the process analyzing a subject system to create the representation of the system at the higher level of abstraction (Gerardo Canfora and Massimiliano Di Penta, 2007), will make one feel comfortable handling with the code and hence can be modified according to the requirement. Here, it is required to analyze code to check how many variables are there and how they are interconnected to one another.
Same is done with the other components in the system. It is now, the time to look into the coupler code. It is such an important code because it couples/ makes the interactions among variables of different components of climate system. In real climate system and hence GCMs, time steps of each of the components have lot of differences because of rate of change of physical processes. Consideration of all this procedure is otherwise called “validation” of climate model.
It is obvious that no feedback loop is independent either with other feedbacks or with other variables in the same component or in the other components of the system. Hence, it is concluded that feedbacks can be quantified. But, by our advance technology, we can figure out how each variable is contributing to climate feedback mechanism. Here, what we are suggesting is the retrospective method of understanding of climate system. This method may be anti-analogous to a complex computer video game. A player new to the game struggles a lot at the beginning of the game because he/she has to get good at understanding of the positive points, negative points, traps, bullets, turns, jumps, etc. As the level increases, number of variables in the game increases and hence, more practice is needed. With considerable effort and perseverance, one can be comfortable playing at the game at the ending.
Climate models are no more than video games but, they are dubiously programmed scientific research findings. Unfortunately, climate research can’t be done on real climate system but luckily, unlike real system, we can go back and forth on climate models like GCMs that are thought of as virtual climate systems. Retrospective method of understanding the climate system is double-edged one. One is validation of climate models and other is the understanding of interactions and hence, feedback mechanisms. Like a video game, it is checked how each variable is influencing and interacting in the climate system. As is discussed earlier this chapter, code is analyzed to see the interactions. Once, it is done, the real part is to reduce the complexity of the system by eliminating one variable from one component and check how it is different from previous run towards total system and a mix of components. Conclusions of extent of interactions may be drawn from the results. Same is repeated for every climate variable until system is gotten understood. This kind of experiments and results will give detailed understanding of whole climate system.
What is needed to succeed?
Retrospection of climate modeling is done through reverse engineering of software of climate models, though this process is time-consuming and very expensive. The process of taking the model code apart and revealing the way in which it works is often an effective way to learn how the technology is built and to make improvements to it. Inspection of inner working patterns of system is called “white-box” reverse engineering.
This method demands not only expertise in computer science related to climate modeling software but also deeper knowledge in climate science. S.M. Easterbrook and T.C. Johns, 2009, says that climate scientists built large and complex climate models with no or a little software engineering training and they do not readily adopt the latest software engineering tools. Moreover, it is common practice for big climate research organizations to hire software engineers who don’t have climate science background, to help building climate model software.
In these circumstances, climate modeling can go forward without thinking much about what is happening inside that software but, converse may not be true until getting expertise both in climate science and software engineering.
References: