Resolution and predictability: An approach to the scaling problem

摘要: We analyzed the relationship between resolution and predictability found that while increasing provides more descriptive information about patterns in data, it also increases difficulty of accurately modeling those patterns. There are limits to natural phenomenon at particular resolutions, “fractal-like” rules determine how both “data” “model” change with resolution. land use data by resampling map sets several different spatial resolutions measuring each. Spatial auto-predictability (Pa) is reduction uncertainty state a pixel scene given knowledge adjacent pixels scene, cross-predictability (Pc) corresponding other scenes. Pa measure internal pattern Pc ability some represent pattern. strong linear log (measured as number per square kilometer). This fractal-like characteristic “self-similarity” decreasing implies may be best described using unitless dimension summarizes changes While generally (because being included), falls or remains stable easier model aggregate results than fine grain ones). Thus one can define an “optimal” for problem balances benefit terms resolution, cost (Pc).