WHO’S ON FIRST? attempts to 
A SURVEY OF COINCIDENT FEATURE MANAGEMENT STRATEGIES polygons, 
world of d. 
James C. Fass 
The Geonex Corporation 1.2 Impli 
St. Petersburg, Florida, USA 
Th 
ABSTRACT objects is 
called leve 
Coincident features occur in the real world all the time. A street centerline, for example, may fall each real-w 
on a county boundary or a Census block may be a portion of a Postal Zip Code area. Interestingly, data. Ear 
approaches for managing coincident features in GIS databases have been slow in coming and frequently Intergraph 
involve data storage and processing overheads which discourage their effective use. In this paper we will this strateg 
survey several algorithms for maintaining coincident features with their relative advantages and associated a unique r 
reference ( 
level assigr 
drawbacks. Finally, we will look at an approach employed recently by the Geonex Corporation which 
manages coincident features as auxiliary FIFO (first-in, first-out) stacks for selected elements and its impact 
  
on data automation techniques. some adva 
user-define 
1. DATA LINKAGE As we shall see, the problems are not the WI 
same for everyone. Some simpler methods for coincident 
The occurrence of geographically coinci- data storage are actually more conducive to represented 
dent features has a way of causing logical coincident feature management than the more each level 
wrinkles in many of today’s GIS applications. In complex GIS applications. The key is to look at world featu 
the worst cases, coincident feature handling how graphic data and attribute data are linked Th: 
routines are what we might expect if Abbot and together. The method of data linkage employed simplicity. 
Costello were the applications analysts: what's by one's GIS applications will determine the this is an id 
visible on top depends on who put it on the extent to which coincident features will require attribute (fe 
coverage first unless what's on second is related special handling. all feature 
to the first coverage... and so it goes. The heart information 
of the problem has to do with the often conflicting 1.1 Raster Data Linkage as the. f 
requirements to maintain one-to-one relationships Intergraph 
between graphic objects and attribute records Raster data users generally don’t worry as de facto 
while maintaining one-to-many relationships about how to store geographically coincident and most ( 
between graphic objects and real-world features. 
Real-World 
Features 
Band 1 
Raster 
Images 
Band 2 
  
User Documentation 
Band 1 = A or B 
  
  
Band 2 — C or D 
  
  
  
Figure 1. Raster Coincidence 
  
144 
features mainly because the concept of a feature is 
usually only a product of the user's interpretation, 
not a product of the data structure. A user of 
remote sensing data for land-use classification, for 
example, may think of a group of similarly 
attributed pixels as a crop of corn. By focussing 
on another band of information in the same grid, 
the user may notice that this crop is coincident 
with a patch of wet soil. The two bands of 
information are physically stored for each and 
every pixel in the grid regardless of the user’s 
recognition of certain features within each band 
(See Figure 1). In this sense, we could say that 
the data are ambivalent to the user’s interpretation 
as long as they remain in raster form. 
At this stage, there are no special handling 
requirements for coincident features simply 
because there is nothing in the data structure 
corresponding to the user’s concept of a feature. 
Additional bands of information can be added with 
no affect on existing bands. As soon as the user 
data in thes 
  
Real-Wori 
Features 
  
Fi;