An error-correcting syntax analyzer for tree languages with substitution errors, called structure-preserved error-correcting tree automaton (ECTA), is studied. Substitution errors are defined in terms of transformations which can easily be accommodated to linguistic notion. Let L be a tree language, for a tree ß not in L, the essence of ECTA is to search for a tree α in L such that the cost sequence of error transformations needed to transform α to ß is the minimum among all the sentences in L. A LANDSAT data interpretation problem is used as an example to illustrate the operation of ECTA.