Arcgis：Tolerance & Resolution

Tolerance：

A spatial referencealso includes tolerance values. X-, y-, z-, and m-coordinates all haveassociated tolerance values that reflect the accuracy of the coordinate data.The tolerance value is the minimum distance between coordinates. If onecoordinate is within the tolerance value of another, they are interpreted asbeing at the same location. This value is used in relational and topologicaloperations when determining whether two points are close enough to be given thesame coordinate value or they are far enough apart to each have their owncoordinate value.

The default toleranceis set to 0.001 meters or its equivalent in map units. This is 10 times thedefault resolution value and is recommended in most cases. You can set a customx,y tolerance value, but it should never approach the data capture resolution,and the minimum allowable x,y tolerance value is twice the x,y resolution.

Resolution：

The resolution represents the detail in which a feature classrecords the location and shape of geographic features. It is the minimumdistance, in map units, that separates unique x-values and unique y-values inthe feature coordinates. For example, if a spatial reference has an x,yresolution of 0.01, then x-coordinates 1.22 and 1.23 can be stored as separatecoordinate values, but x-coordinates 1.222 and 1.223 are both stored as 1.22.This is illustrated in the graphic below. The final digit on the latter pair ofx-coordinates is truncated because the change in value is less than the x,yresolution. The same would apply for y-coordinates. 

Software engineers like to think about integer coordinatesrepresented on a Cartesian grid with a very fine mesh. The spacing of the meshis defined by the coordinate resolution. All feature coordinates aregeoreferenced according to the chosen coordinate system and snapped to thecoordinate grid. This grid, known as the coordinate resolution grid, is definedby the resolution which determines the precision (the number of significantdigits) of your coordinate values. The resolution establishes the fineness ofthe coordinate resolution grid that covers the extent of your feature class orfeature dataset. All coordinates snap to this grid, and the resolution defineshow far apart the individual lines of the grid are.

Resolutionvalues are in the same units as the associated coordinate system. For example,if a spatial reference is using a projected coordinate system with units ofmeters, the resolution value is defined in meters. The default resolution valueis 0.0001 meters (1/10 millimeters) or its equivalent in map units. Forexample, if a feature class is stored in state plane feet, the defaultprecision will be 0.0003281 feet (0.003937 inches). If coordinates are inlatitude-longitude, the default resolution is 0.000000001 degrees.

Featurecoordinates with smaller (finer) x,y resolutions can have more digits ofprecision. However, x,y resolutions that are extremely small may impactperformance in terms of excessive disk use and increased I/O. As the x,yresolution value is increased (becomes more coarse), the precision associatedwith feature coordinates diminishes; feature boundaries will be smoothed,simplified, or not shown at all.

In thefollowing graphic, the grid with the large x,y resolution will not be able tostore the polygon feature with as much precision. Conversely, the grid with thesmall x,y resolution will store the polygon feature using greater precision,better preserving its shape. 

I strongly recommend using the default x,yresolution in most cases because it has proved to perform quite well and canstore adequate coordinate precision for most situations.

Note:

Differenttolerance values can produce different answers for relational and topologicaloperations. For example, two geometries might be classified as disjoint (nopoints in common) with the very small tolerance, but a larger tolerance mightcause them to be classified as coincident and thus assigned the same coordinatelocations.

图层运算的时候要根据要求设置限差，即使图层中有限差信息。有时使用Load功能把数据导入可以保证限差不变，把数据copy进图层可能会出现问题。尽量在限差一致的环境内进行操作。  比如都在限差为0.0001的库图层之间操作。通过Environment可以设置toolbox运算时的T&R。