! 2004 
ithout 
itasets 
d that 
ied on 
Jated- 
es of 
errain 
ent is 
ipdate 
; will 
RY 
most 
ea of 
ıtputs 
The 
nly if 
ntain 
ch of 
[ will 
ngle- 
l| the 
nulti- 
area 
tified 
rface 
nt a 
The 
te of 
odel 
ints, 
km). 
:d to 
ame 
ified 
w of 
ition 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
  
je p $e m selection Took indo Hep | 
    
  
  
  
Layers 
f] Central Terrain T 
© BA cntyspl 
  
  
  
  
  
  
  
  
  
      
&) Centzal Terrain 
is A atyspl 
  
  
  
  
  
  
hA 
ar? 
4 
S "cr 
Euer Jani Jar E 
pe AG An Ef “fo {487 vu |A- bru] | rin + kon Als A B70/A- 8-7 4- 27 
ug I MEME" [7 [0462363 696380.16 Meters | ET 
5a 5b 
  
  
   
    
  
IeeuuogesmkosM: 
Ele] 
  
Central Tenain 
E B otyspi 
    
  
  
  
  
{= PE nu 
  
5c 
Figure 5: An example of 
  
% Untitled - ArcMap - dod E LES 
| Be Ek Mew Insert Selection Took Mns eb 
  
    
Déega|ieex|ec|t DS mie ae een Pa 
| ions = Lori] ERE | 
= £ Layers 
M Central_Terrain 
& À atysp! 
  
  
  
  
  
  
TZ poat Joe 70 jA- rr =| 
(TE [i07335.96 £39718.45 Meters | 
  
  
5d 
Terrain application in surface visualization (source data: courtesy of MassGIS, Commonwe alth of 
Massachusetts Executive Office of Environmental Affairs). 
The DTM shown in Figure 5a contains only about 60000 points 
and was generated in several seconds. When zooming into a 
sub-area of interest, more data will be needed in order to 
provide a more detailed view. However, in the meantime, the 
extent of the query area has become much smaller, resulting in 
higher resolution, but a potentially smaller (or an acceptably 
sized) DTM (31000 points in Figures Sb, 28000 points in Figure 
5c, and 28000 points in Figure 5d). The 407 and AH can be 
calculated automatically for each zoom (and pan) operation. In 
this process, users can specify what level of vertical resolution 
to use at given scale by associating scale-ranges with pyramid 
layer indices. 
Users can use the DTM generated as the result of a query to 
perform surface anal: ses. There may be, however, cases where 
a DTM cannot be created because of system constraints. This 
can happen if the query «xteiit is too b:g, and high resolution is 
required, as in calculating volume and area, generating 
contours, profiles, and view-shed, all across the whole terrain 
extent. This problem can be solved by performing such tasks 
tile by tile (or a sub-group of tiles by a sub-group of tiles), and 
then unifying the results. 
5. CONCLUSIONS 
This paper has presented an efficient approach for GIS users to 
handle large terrain data and model surface applications. 
Because only measurements and rules are stored in a database, 
users can take the advantages of TIN and GRID structures 
without sacrificing storage or losing information. The tiling 
scheme makes it possible to perform large-scale tasks that 
require working on a DTM of high resolution. It also helps to 
achieve spatial coherence, thus speeding up spatial queries, and 
reducing disk 1/0 and network traffic. Vertical indexing 
provides another contribution to further speed up spatial 
queries. 
Storing terrain data as feature classes in a feature dataset allows 
them to be integrated with 2D data and be shared by other 
applications, such as Topology and Geometric-Network (Zeiler,