Land-Atmosphere Coupling Metrics
Land-atmosphere coupling metrics can be divided into two types
Statistical or climatological metrics, calculated via some co-variability among terms over a large sample. Two-legged metrics and soil moisture memory are examples. These are useful for describing the hydro-climatology (observed or model) and can be used as skill metrics that may inform model improvements.
Physical metrics relate to instantaneous states of the land-atmosphere system, or in the case of mixing diagrams, as an evolution over several hours within the diurnal cycle. These are very closely connected with the underlying physical processes that link land and atmosphere.
Land-atmosphere coupling metrics can be applied at various scales
Historically, these metrics have been calculated at climate model grid scales (10³-10⁶ km²) . This conforms to many global gridded observational data sets as well as climate models, and are a good scale for model evaluation. However, the actual land-atmosphere interactions in nature occur at much smaller scales, concurrent with the size of land surface features (patches of forest, agricultural fields, hills and valleys, urban areas) and scales of atmospheric turbulence.
Many land-atmosphere coupling metrics can be calculated at the patch scale over individual land cover types (data from flux tower measurements or from high-resolution models like large-eddy simulations) as well as area-averaged fields representative of a weather or climate model grid. Comparison can reveal whether climate model grid cell coupling metrics are consistent with what the hi-resolution models and data suggest, i.e. getting “the right answer for the right reason”.
If there are important nonlinear processes in the system, we should not expect the metrics calculated on area-means to match the area-means of the metrics. This is a powerful diagnostic tool to understand the level of complexity in the land-atmosphere system at any location, and to indicate if something is going wrong with model approaches to representing sub-grid heterogeneity.
Land-atmosphere coupling metrics have different uses
The physical, process-based (instantaneous or diurnal-cycle) coupling metrics have two main uses. One is to better understand the natural land-atmosphere system, when applied to observations from the field. The other use to take that understanding and apply these metrics as diagnostic tools for weather and climate model development and implementation, to help insure the land-atmosphere interaction is simulated well.
The statistical metrics are also useful for understanding the nature of hydro-climate on longer time scales. They are also applicable as skill metrics for model behavior after a number of simulations have been run. Regional and global patterns of statistical metrics have appeared to be quite robust across many different studies.
Future directions for land-atmosphere coupling metrics
Going forward, there is currently work to try to make purely observational (from satellite data) global gridded estimates of many land-atmosphere coupling metrics . There are a lot of issues with satellite data that the research community is having to address to make this viable (e.g., signal-to-noise problems in the data, dealing with coverage gaps in space and time, cross-calibration between different instruments and platforms...).
Another new direction is use of the techniques from Information Theory, which are more general and not restricted by the assumptions that are inherent in classical statistical approaches (e.g., that data are normally distributed and that all relationships are purely linear). Many of the newer metrics being added to this list are in this category, and many statistics originally based on linear statistics like correlations or standard deviations are being modernized using more general Information Theory approaches.
Below is a table of land-atmosphere coupling metrics - each is linked to a 1-page PDF describing the metric, its calculation, uses, caveats, and relevant literature for further exploration.
A set of "cheat sheets" for the community
Cheat Sheets (click name to access PDF)
Is the metric primarily statistical (physical linkages implied) or based directly in physical processes?
Does the method require terrestrial (Land) or atmospheric (Atmos) state variables, or surface flux data?
Is it limited to local in space (no horizontal relationships, only in the vertical) or in time (no lagged relationships)?
Can it be applied only to model output (Obs'ble=N)?