Ozone concentration across California

Ordinary kriging from sparse monitoring stations to a continuous surface, against the health standard

Academic coursework  /  2021-2022

Discipline
Geostatistics
Tools
ArcGIS
Methods
ordinary kriging, semivariogram modelling, interpolation from point samples, threshold exceedance mapping
Data
ESRI

Notes on method

Kriging turns a few dozen measurements into a value for every location in the state. That is useful and slightly dangerous: the output looks equally confident everywhere, and it is not.

Keeping the monitoring stations visible on top of the interpolated surface is the cheapest honest thing a map like this can do. Where the dots are dense, the surface is interpolation. Where they are sparse, it is extrapolation wearing the same colours, and a reader can see which is which without being told.

Ordinary kriging was chosen over inverse distance weighting because the semivariogram showed structure worth modelling rather than a simple distance decay.

Interpolated ozone surface across California graded from blue on the coast to orange and red inland, with the monitoring stations the surface was interpolated from drawn as points on top.
Kriged surface with the source stations drawn over it. Every value between the dots is an estimate, and showing the dots is how a reader can judge where the estimate is thin.Base data: ESRI. Scale 1:4,000,000Open larger (2432px)