Environmetrika supports scientifically defensible accuracy assessments of remotely sensed data with complex sample surveys.
The most common type of remotely sensed data is based on pixel-level predictions of categorical and continuous land-cover variables with multispectral data from earth-observing orbital satellites, such as Landsat. The reference data can be a multiphase/mutistage probability sample of field observations and higher resolution remotely sensed observations.
The typical product is an ""error matrix", which is a contingency table that cross-classifies remotely sensed categories with categorical reference data. Environmetrika employs robust statistical estimators for complex sampling designs.