This post adds an important perspective to the climate debate rarely discussed.

A key point often overlooked in discussions about climate change is the statistical significance of human contributions. If human impact is negligible, climate change is simply natural variations in our climate system. However, if human activities significantly contribute, it can be claimed we are facing serious issues.

Agencies like the MET, BoM, and the IPCC use a non-integrative , first-order autoregressive model (AR(1)) to assess the statistical significance of human contributions. This model claims to show that human activity is the primary cause of climate change. But does it really?

The AR(1) model is under-parameter ized and has a poor Goodness of Fit (GOF) to the temperature data. It is a single-order autoregressive model that assumes no relationship between today's temperature and that of two or more years ago, despite known cyclical trends spanning many years. This choice of model makes no sense. Additionally, because it is a stationary model, it drifts when applied to the non-stationary temperature datasets, becoming less reliable over time. In short, the AR(1) model cannot determine the statistical significance of human contributions to climate change.

Dr. Doug Keenan proposed an integrative model using third-order autoregression, the ARIMA(3,1,0) model, which I validated in my paper presented to the UK Parliament as testimony. This model has an excellent GOF and impressive predictive skills. So, why don't MET, BoM, the IPCC, and others use this model? Because it shows that human contribution to climate change is not statistically significant.

In a surprising admission, MET acknowledged their AR(1) model is flawed, yet claimed they don't need models to determine statistical significance—th ey can make this determination by simply eyeballing the data. Despite acknowledging the AR(1) model's inadequacies, they continue to use it, maintaining it shows statistical significance of man's contribution to climate change -- a claim not at all supported in their choice of model.

The attached image is a test I made of the AR(1) model and it's predictive skills. It has no predictive skills whatsoever. I might go into why this is important. But for brevity, I'll table that discussion for now.