How reliable are Climate Models? I'll be the first to state that no model, climate or otherwise is perfect, but then again, they are not meant be perfect. Climate models are not just predictions, they are for the most part "Scenarios"; that is what could happen if...
With respect to climate modeling, there are two questions of significant importance. (1) How accurately can they predict the past and (2), how accurately can they predict the future?
Good climate models are extremely complicated. In other words, they must contain a lot of data and variables. In constructing a climate model, one does not just input data and expect to have a reliable prediction. The main thing that goes into climate models is the parameters known to affect climate. These in short are known as "Forcings" and "Feedbacks". A forcing is something like the Sun, volcanic activity, or greenhouse gases. They are long term and directly affect climate and or temperature. Feedbacks on the other hand are short term and are a result of "Forcings". Some feedbacks are clouds, water vapor, and aerosols. Feedbacks are short lived in the atmosphere and constantly change. Feedbacks affect weather, not climate.
By understand forcings and feedbacks, in not only how they affect climate, but to what degree they affect it, is how models work. This understanding comes from actual observations. By understanding this, the test of a models accuracy can be determined by running both scenarios of the past and the future, and compare the results with the actual observed data. So, how good are they? Look at the three graphs below (source: IPCC).
In all three graphs the red is actual observed data and they gray is modeled predictions based on known sensitivity of climate forcings and feedbacks.
Notice in (a) only "natural" forcings are used in the scenario. That is the Sun and volcanic activity. It is assumed in (a) that those natural forcings did not change. Thus, the predicted outcome did not follow the actual recorded data. There have been a couple of major volcanic eruption since the 1880's affecting short term climate and the Sun did increase somewhat int total solar irradiance (TSI) up to 1940, but virtually none since then, in fact it decreased slightly. What is learned from scenario (a) is that something other than the Sun and volcanic activity has affect the actual observed climate.
Now, look at graph (b). Scenario(s) (b) are done with only anthropogenic (human influenced) forcings and no natural forcings. The anthropogenic forcings used were greenhouse gases and sulfate aerosols. Just a note, sulfate aerosols have a cooling effect unlike greenhouse gases like carbon dioxide and methane. Both carbon dioxide and sulfate aerosols come from the burning of fossil fuels. Note that in graph (b) it is fairly close to actual observed data. The big difference is the period between 1940 and 1975.
Now look at graph (c). This scenario includes both "natural" and "anthropogenic" forcings. Note that this scenario most closely correlates with what actually happened. Yet it is not completely accurate, but recall, I said no model is perfect, they are not meant to be perfect. They are only meant to give the best approximations to what could happen with what is known. Could graph (c) be better? Yes, as not all known forcings are considered in any of those models.
With respect to climate modeling, there are two questions of significant importance. (1) How accurately can they predict the past and (2), how accurately can they predict the future?
Good climate models are extremely complicated. In other words, they must contain a lot of data and variables. In constructing a climate model, one does not just input data and expect to have a reliable prediction. The main thing that goes into climate models is the parameters known to affect climate. These in short are known as "Forcings" and "Feedbacks". A forcing is something like the Sun, volcanic activity, or greenhouse gases. They are long term and directly affect climate and or temperature. Feedbacks on the other hand are short term and are a result of "Forcings". Some feedbacks are clouds, water vapor, and aerosols. Feedbacks are short lived in the atmosphere and constantly change. Feedbacks affect weather, not climate.
By understand forcings and feedbacks, in not only how they affect climate, but to what degree they affect it, is how models work. This understanding comes from actual observations. By understanding this, the test of a models accuracy can be determined by running both scenarios of the past and the future, and compare the results with the actual observed data. So, how good are they? Look at the three graphs below (source: IPCC).
In all three graphs the red is actual observed data and they gray is modeled predictions based on known sensitivity of climate forcings and feedbacks.
Notice in (a) only "natural" forcings are used in the scenario. That is the Sun and volcanic activity. It is assumed in (a) that those natural forcings did not change. Thus, the predicted outcome did not follow the actual recorded data. There have been a couple of major volcanic eruption since the 1880's affecting short term climate and the Sun did increase somewhat int total solar irradiance (TSI) up to 1940, but virtually none since then, in fact it decreased slightly. What is learned from scenario (a) is that something other than the Sun and volcanic activity has affect the actual observed climate.
Now, look at graph (b). Scenario(s) (b) are done with only anthropogenic (human influenced) forcings and no natural forcings. The anthropogenic forcings used were greenhouse gases and sulfate aerosols. Just a note, sulfate aerosols have a cooling effect unlike greenhouse gases like carbon dioxide and methane. Both carbon dioxide and sulfate aerosols come from the burning of fossil fuels. Note that in graph (b) it is fairly close to actual observed data. The big difference is the period between 1940 and 1975.
Now look at graph (c). This scenario includes both "natural" and "anthropogenic" forcings. Note that this scenario most closely correlates with what actually happened. Yet it is not completely accurate, but recall, I said no model is perfect, they are not meant to be perfect. They are only meant to give the best approximations to what could happen with what is known. Could graph (c) be better? Yes, as not all known forcings are considered in any of those models.
