How Locate Outliers (3D Analyst) works
Locate Outliers takes an input TIN or terrain dataset and finds points outside a given z-range and/or that are significantly different in height than their neighbors. The intention is to locate measurement points that are suspicious and potentially in error. The locations are then written as points to an output feature class so an analyst can take a closer look and verify, on a point-by-point basis, whether the points are indeed in error. Those points verified as blunders can be used as input to the Delete Terrain Points tool.
When the Apply Hard Limit option is turned on, measurements with z-values outside the specified minimum and maximum are located.
When the Apply Comparison Filter option is turned on, the tool locates measurements that exceed a height difference relative to their neighbors and create local surface patches with slopes that exceed a given threshold. All mass points are examined. For each, the height difference and slope tests are made between the point and its natural neighbors. If the ratio of violations for a test point exceeds the tolerance ratio, the point is flagged as an outlier.
When the number of outliers located exceeds the Outlier Cap value, the tool stops. This is a safety switch used to prevent too many points from being written when the tolerance parameters have been assigned values that are overly sensitive for the data.
When both hard and comparison filters are turned on, the hard filter is applied first.