Still, climate models are important tools utilized to advance our understanding of current and past climate. They also provide both qualitative and quantitative information about potential future climate. But in spite of their sophistication, they remain just that-models. They are nothing more than simulations of the real world, constrained in their ability to correctly capture and portray each of the important processes that operate across multiple spatial and temporal domains to affect climate. By their very nature, climate models deal in the hypothetical. Their output amounts to nothing more than projections of future possibilities; and as such, model output can never substitute for real-world observations, especially when attempting to discern a CO2-induced influence on extreme weather.

It is also worth pointing out another weakness of climate models. The average person has little to no knowledge concerning the inner workings and limitations that exist in present-day state-of-the-art climate models. Few people are aware that although the models are quite sophisticated, they are also replete with numerous inadequacies and biases. And although such shortcomings are frequently documented in the peer-reviewed scientific literature, these imperfections rarely find their way into public discourse.

A partial assessment of model inadequacies was recently conducted and published in a major report of the Nongovernmental International Panel on Climate Change (NIPCC). Notwithstanding their admirable complexities, the NIPCC scientists found the models to be deficient in many aspects of their portrayal of climate, leading them to strongly question their ability to provide reliable simulations of the future (Idso et al., 2013).

One example of such deficiencies was presented earlier in Figure 1, where simulated global temperatures from 73 models are plotted against mid-tropospheric observed temperatures over the past three and a half decades. The universal failure of the models to correctly project global temperature over this time period is shocking, especially since global temperature is the single most important variable examined in all the models because of its expectation to rise as the air's CO2 content increases. No other variable receives as much attention. Yet, the models failed to correctly project global temperature over the past three decades. So how in the world can they be expected to produce reliable simulations of extreme weather events decades to centuries into the future? Simply put, they cannot. It is intrinsically much more difficult to simulate extreme weather events-which operate within much smaller spatial and temporal domains-than it is to simulate average global temperature.

Confidence in a model is based on the careful evaluation of its performance against actual observations. Because models fail to accurately simulate what is arguably supposed to be the simplest of all climatic variables-global temperature-confidence in their ability to simulate more complex events, such as is required with extreme weather, must be greatly tempered.

Recognizing that climate model output is no substitute for real-world data, scientists must turn to observations in their efforts to prove or disprove any CO2-induced influence on extreme weather events. And that requires datasets that have been in existence for long periods of time, datasets which are of sufficient length to adequately discern whether or not recent changes in extreme weather parameters have stretched beyond their known realm of natural variability. And this leads to the second principle presented in Figure 2: The natural variability of the parameter must be studied and known.