In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
636
imagery with the producer’s- and user’s accuracy of the
classified tree species.
Table 5 shows that best agreements for the classification based
on the summer 2007 images are obtained for Picea abies (89%)
and Larix decidua (78%). The confusion matrix clearly reveals
that especially deciduous trees are misclassified. But also Pinus
sylvestris is confused with Picea abies. Acer sp. and Fraxinus
excelsior are generally overestimated.
July 2007 Classified as
Field
Aa
Pa
Ps
La
Ac
Fs
Fe
PA
Aa
28
5
0
0
1
6
4
0.64
Pa
2
68
2
3
1
0
0
0.89
Ps
0
5
12
1
2
1
0
0.57
La
0
4
0
14
0
0
0
0.78
Ac
1
0
0
0
19
10
12
0.45
Fs
6
4
0
1
11
54
15
0.59
Fe
4
1
0
0
5
8
38
0.67
UA
0.68
0.78
0.86
0.74
0.49
0.68
0.54
Table 5. Confusion matrix for tree species classification using
the explanatory variables from July 2007 ADS40-SH40
imagery with the producer’s- and user’s accuracy of the
classified tree species.
Table 6 shows that five of seven tree species are classified with
accuracies > 74% when using the Summer 2008 images
including the NIR band.
The obtained accuracy for Acer sp. remains very low because it
is often misclassified as Fagus sylvatica and Fraxinus excelsior.
August 2008 Classified as
Field
Aa
Pa
Ps
La
Ac
Fs
Fe
PA
Aa
29
3
1
0
0
1
2
0.81
Pa
2
78
2
5
0
2
2
0.86
Ps
0
7
16
0
0
1
0
0.67
La
0
4
0
23
0
0
2
0.79
Ac
0
1
0
0
12
14
8
0.34
Fs
2
1
0
1
7
64
11
0.74
Fe
3
0
0
0
8
8
84
0.82
UA
0.81
0.83
0.84
0.79
0.44
0.71
0.77
Table 6. Confusion matrix for tree species classification using
the explanatory variables from August 2008 ADS40-SH52
images with the producer’s- and user’s accuracy of the
classified tree species.
4.2 Predictive mapping
The tree species which have been modelled with > 90 %
probability in the 2007 and 2008 images are depicted in Fig. 3.
At first glance, a visual image analysis suggests that the
agreements in most parts of the site are good. However, a more
detailed image inspection confirms the results of tables 4-6 and
indicates that Acer sp. is often misclassified as Fagus sylvatica
or Fraxinus excelsior - independently from the acquisition date
of the images. Fig. 3 also shows that the predictions of the tree
species slightly vary in each of the maps. The underestimation
of Larix decidua is clearly visible in the May images whereas
the overestimations of Fraxinus excelsior and Fagus sylvatica
are clearly visible in the July 2007 images.
$0 ■ Fagus syfvallca
3 Motor« m Acer sp. I
I Pinus syt.ei.xrts
I A birrs albs
Larix decktu»
0 40 80 f№ Fas us sylvatica ill Fraxlnut excelsior MB Pirns sylvestris
mif^c==>*s*en gjrn Acer so m met a abies SOI Abies mbs
É3 Larix deddua
WB Fagus sylvatica
am Acer sp.
Frsxmut excelsior MH Pinus sylvestris
ma Pice a abies MM Ablet alba
’CCiLed* deddua
Figure 3. Tree species classification maps based on imagery
from May 2007 (top), July 2007 (middle) and August 2008
(bottom) with the largest shadows.
5. DISCUSSION
The potential and the limits of classifying seven tree species has
been tested using multi-temporal multispectral ADS40 imagery.
The most significant achievement is the demonstration that
spring ADS40 imagery as input for the classification of tree
species is superior to the summer imagery - even without the
additional information of the NIR band. Summer imagery from