Variance in Climate Models and Observations

Studying the Spectrum of Climate Variability

In my last year of graduate school, I examined the dominant timescales of variability (interannual, decadal, centennial, etc.) in climate model temperature and precipitation data. For a given time series, such as the surface temperature (red) and precipitation (green) over the North Atlantic (below), I used "spectral estimation" to compare the relative importance of different timescales of variability in the time series.

For example, I estimated the variance spectrum (variability as a function of frequency) for the red time series in the panel above.

In this case, the spectrum of variability follows a 'power law' relationship, meaning it shows growing variance as a function of time period. I can describe this relationship by calculating the slope of the line (below) in log-log space. I define this slope as β.

I then calculated the slope of this line for each time series at each geographic location from a climate model simulation and mapped these β values (maps to right) for each geographic location around the globe. These maps can give us a sense where decadal-centennial scale variability (red) or interannual variability (blue) dominates around the globe.

Variability in Climate Model Data

For example, I shaded a map below of β (the spectral slope for each surface temperature time series). If the area on the map is red, this means there is more long-period variability than short-period variability in the time series. If the area on the map is blue, this means that there is more short-period than long-period variability in this geographic location.

I then compared these maps in different climate model simulations to determine where models agree that there is "spectral scaling" (more long-period variability than short-period variability) in the climate system.

Variability in Model vs Paleoclimate Data

Finally, I compared climate model data to our recent observations of climate or to reconstructions of past climates (e.g., coral paleoclimate records, shaded dots in tropical Pacific Ocean below).

Red, White, or Blue: Why does it matter?

The joint impacts of internal variability and forced change will control how climate change plays out in the 21st century. If natural decadal-multidecadal variations are weak and forced changes are strong (e.g., climate is 'white' or 'blue'), then global temperatures will increase relatively smoothly. By contrast, if internal decadal-multidecadal climate variability is relatively strong, ‘hiatuses’ will follow rapid periods of global warming. Similarly for precipitation, if natural, low-frequency variations are weak and forced changes are strong, then many regions will experience a relatively steady trend towards drier (or wetter) conditions as the globe warms. On the other hand, if internal decadal-multidecadal precipitation variability is relatively strong, this descent into drier (or ascent into wetter) conditions will be punctuated by extended droughts (or pluvials).