Accuracy of calculated results depends on a number of discrete processes, which can all compound inaccuracy in the final dataset.
The importance of Metadata in this situation, is it can be used to explain error and even identify steps error is introduced.
The most important thing is to understand that ground truthing your results, if performed correctly will always return a more accurate answer.
The aspects of GIS that introduce error into a GIS dataset are:
- Scale - the scale of a data set is so often overlooked when computing and comparing results that its embarrassing.
The scale at which the dataset was Captured or Digitised should affect your interpolation of results when comparing them to ground truthing. Scale of a data set is an important attribute in Metadata
- Interpolated/Calculated Data set - this is directly related to scale. If your data set is interpolated, calculated, classified, extrapolated etc. with data sets of varying scale, your end result will be an general indication.
As GIS data is an expensive commodity, I understand and perform operations on data sets that have been captured at varying scales.
At best your result is equivalent to the data set of the lowest resolution (highest scale - i.e 1:1000 is better than 1:1,000,000). This is important to keep in mind when using this data set. Therefore, it is important to define the processes and interpolation methods in Metadata for this reason
- Compound error - Related to Point 1 and 2, but also concerns how much processing a data set has endured. I consider it like over working metal, If you over work metal it slowly be comes hardened, brittle and more difficult to manipulate and mold. Every time you perform an operation on a GIS data set, through limitations of data structures (computer theory not relevant to this forum - i.e. rounding error etc), small errors are introduced, this error can be compounded depending on the process you are performing, and other data sets involved. This is more prevalent with floating point and double precision numbers, the type most often used in calculations.
Hence, the importance of keeping a record of processes and the data sets associated with each process, and for this information to be stored in Metadata.
Unfortunately this is unavoidable, but can be minimized by keeping a clean copy of all base data sets and performing analysis from these, and not using overworked data sets
- Data Types used for storing attributes - This one is commonly overlooked. Understand your data, they types of information you need to keep, what it will be used for and the current and future precision you think the data will maintain. I've seen many databases and data set that truncate floating point data to 3 decimal places to 'save space' only to find out that their calculations are not what they should be. This relates and contributes to a Compound Error but is more importantly, usually a gross misunderstanding of the data captured, what it represents, and the importance of precision. If you do truncate data, put the reason why in your records and Metadata!
GIS data is never going to match results from ground truthing, purely because you're more inclined to use precision instruments for verification, i.e. Theodolite for surveying and area calculation.
If your area calculations are off by a very large margin, then look at the scale, the age, and processes performed on the data set. It could be a colleague may have deliberately or accidentally altered the data set for a reason.
Hope this helps