2. Collecting land use data by sampling methods 
Land use data may be obtained by a complete survey or by using 
sampling techniques. In the first case the surfaces of the 
different land use units are determined by using a planimeter 
or by digitizing the land use boundaries. This procedure can 
become very time consuming in territories with small parcels 
of different land uses like in many parts of Switzerland. The 
application of sampling techniques is less time consuming as 
it allows to determine the land uses from selected sample units. 
These sample units may be either points, lines (traverses) or 
areas, they can be arranged regularly (systematic sample) or 
randomly (random sample). For the area measurement, point and 
line samples are known to be most suited (4). If point samples 
are used, preference should be given to systematic samples 
(5), as there are several advantages compared to random 
samples: 
- the original sample grid can be used again (permanent 
samples) if a land use inventory will have to be updated. 
This allows a more reliable detection of changes in land use 
than random point samples would do 
- it is also possible to get detailed statistical land use 
data of areas which are not necessarily administrative 
units 
- data can also be easily displayed in form of couputer maps 
- besides the actual land use data, further information, e.g. 
about geology, soil conditions, the gradient of slopes, 
elevation, etc. can also be referred to the same grid. 
The accuracy of a sampling survey depends mainly on the total 
amount of the sample points and on the land use pattern. If 
aerial photographs are used for sampling, there is an error 
due to the photo interpretation which will have to be 
considered besides the sampling error itself. 
An estimation of the accuracy of aereal samples can be 
achieved by the following formula (6, 7): 
(100-p) 
oc [2] =r eR [1] 
In this formula n means the total amount of sample points 
used within the area of investigation. p is the fractional 
part of the phenomena to be estimated (e.g. percentage of 
forest within a community). If k=1,66G is the mean square error 
in percent, referring to the entire area of investigation 
(confidence interval of 66%). 
    
   
  
  
   
   
  
  
  
  
  
    
  
   
   
   
   
   
  
  
  
   
  
    
  
    
   
  
  
  
     
In ce 
inves 
follc 
nw is 
phenc 
use 1 
with 
disti 
fall 
overe 
cate 
with 
erro: 
These 
of tl 
accu: 
In c: 
aeri: 
picti 
cons: 
elev: 
regu. 
There 
- The 
ti: 
bo! 
- In 
th 
us‘ 
th 
- Th 
th 
us