Analysis of interpretation
properties of forest lands on various scale TM imagery1
Li Jiquan Research
Institute of Forest Resource information Techniques, Chinese Academy of
forestry
Abstract This is paper described
the radiation between TM image scales and its interpretation properties of
forest lands and purposed optimal TM image scale for forest survey in
"Three north Shelter forest Region".
General situation of study
area and applied information source Kangbao county is located at
the north part of Zhangjiakou prefecture, Hebei province. Its geographical
position is 41 25'-42 09 N and 1114 11 114 57 'E belong to the south east
part of Mongolian Plateau and the elevation is from 1300 m to 1700 m the
annual precipitation in 350 mm and its is very windy and dusty Kangbao
County is a transit region from aired steeple to forest grasslands and the
vegetation type is simple [1] is no natural forest in Kangbao County and
small pieces of artificial forest are scattered nearby the villages and
small towns however the covering area of arm land wind break forest
network is extensive LANDSAT 5 false color composite TM image scale
1:1000,000 1:200,000 and 1:500,000 imaged in 8th Sep 1897 were employed in
this research.
Forest interpretation and area measurement on TM
imagery
- The timing of TM image employed in the research is very opportune by
this time most crops have been harvested The outlines of forest lands
and windbreak forest network are clearly discernible on TM images which
affords is a favourable conditions to improve the forest interpretation
precision(FRITP).The interpretation keys are as follows:
forest
lands----with obvious regular shape. The tone on 1:100,000 scaled TM
image is fresh and light red colour; and on 1:200,000 and 1:500,000
scaled TM images is dark red color. Windbreak forest network----with
clear netted and belt shape. The tone on 1:100,000 scaled TM images is
red color and dark red color appears on 1:200,000 and 1:500,00 scaled TM
images .
Visual interpretation and plot delineation of forest
lands and wind break forest network were carried out respectively on
1;100,000 and 1:200.000 as well as 1:500.000 scaled TM images according
to the records obtained during fields survey in Kangbao County and
relevant specialized maps and materials as reference.
- The 'KP-90 N" model digital planimeter was employed to measure the
area of forest lands To counter of some forest area smaller than the
minimum measurable area of planimeter (ie 10 ha. On 1:100.000 scaled
interpretation map) the following method was adopted to solve the above
problem i.e several adjacent small piece forests were constituted in to
larger plot for area measurement, and a DOT-GRID with dense dots was
employed to measure the weight of every small piece forest than their
actual area could be calculated respectively.
Fig.1 TM image on the scale of
1:100,000
Fig.2 TM image on the scale of
1:200,000
Fig.3 TM image on the scale of
1:500,000Examination of correct interpretaton percentage
of FOREST plot (CIPFP) on TM imagesIn order examine the CIPFP
(wind break forest network was not included here the lowing steps were
taken.
- A random table [2] was employed to carry out sampling on 1:100.000
and 1:200.000 scaled TM images respectively The population on 1:100.000
and 1:200.000 scaled TM images are compared of 332 and 285 interpreted
forest plots respectively the sampled forest them 50 forest plots each
of them are randomly sampled to form two large samples. The sampled
forest plots were marked on topographical map respectively according to
their ordinal number and geographical position There are only 18
interpretable forest plots on 1:500:000 scaled TM image in Kangbao
County therefore they were all marked on topographical map for
examination.
- The CIPFP on 1:100.000 1:200.000 and 500,000 scaled TM images were
examined respectively using 1:34.000 scaled serial photography taken in
1984 (after setting up the ground interpretation keys) and making
reference to the geographical position of sampled forest plot marked on
topographic map.
- Taking the area weight of each sample plot in to account the
calculation of CIPFP is more reasonable on the basis of area.
After examining to the random samples taken from the populations
of 1:100.000 and 1:200.000 scaled TM images, the results of CIPFP were
arranged in table1.
Table 1. A schedule of CIPFP on TM images
Scale of TM images |
1 : 00,000 |
1 : 200,000 |
|
Number of sample plots |
Total area(ha.) |
Percentage(%) |
Number of sample plots |
Total area(ha.) |
Percentage(%) |
Correct interpretation |
46 |
827 |
92.3 |
47 |
971 |
91.4 |
Erroneous interpretation |
4 |
69 |
7.7 |
3 |
91 |
8.6 |
Grand total |
50 |
896 |
100.0 |
50 |
1,062 |
100.0 | All of 18 interpretable
forest plots on 1:500,000 scaled TM image were examined however the area
weight of correctly interpreted forest plots is large and that of
erroneously interpreted forest plots is small in general in general
therefore the results of CIPFP obtained from 1:500.000 scaled TM image
was higher than that gained from random sampling So this result was used
for reference only. Results and Analysis
- The above examination of CIPFP on 1:100.000 and 1:200.000 and
1:500.000 scaled TM image enable us to preliminary find out the degree
of interpretation accuracy within the extent of minimum delineation area
(MDA) on this three kinds of TM image scale in order to systematically
analyze interpretation properties of forest lands on Various scaled TM
image (VSTMI) the 332 interpreted forest plots of Kangbao county were
sorted out according to the specified area limits on this basis of
ground resolution of TM images the number and area of forest plots
within each area limits and their percentages occupied in total number
and area of forest plots were calculated and a diagram was drawn for
understanding the distribution patterns on the aspect of area and
quantity in this region. It can be seen from table 2 and fig. 4.
The small forest plots less than32 is 283b in number hat makes
up 85.3% of the total number of forest plots and the total area of
forest plots less than 32 hais 4,226 ha which makes up 45.9% of the
grand total area forest plots in kangbao county so they greatly affect
the FRITP on VSTMI.
Table 2 Number and area of forest plots and their
percentages within various area limits.
Area limits(ha.) |
Number of forest plots |
Percentage(%) |
Total area(ha.) |
Percentage(%) |
³3.0----<4.5 |
11 |
3.3 |
38 |
0.4 |
³4.5----<8.0 |
36 |
10.8 |
217 |
2.4 |
³8.0----<12.5 |
67 |
20.2 |
675 |
7.3 |
³12.5----<18.0 |
70 |
21.1 |
1042 |
11.3 |
³18.0----<24.5 |
61 |
18.4 |
1221 |
13.3 |
³24.5----<32.0 |
38 |
11.5 |
1033 |
11.2 |
³32.0----<40.5 |
27 |
8.1 |
980 |
10.6 |
³40.5----<50.0 |
4 |
1.2 |
177 |
1.9 |
³50.0----<100.0 |
13 |
3.9 |
852 |
9.2 |
³100.0 |
5 |
1.5 |
2990 |
32.4 |
Grand total |
332 |
100.0 |
9225 |
100.0 |
Fig. 4
Diagram of forest distribution pattern sorted out according to the
specified area limits.
- In order to further analyze FRITP on VSTMIaccording to the minimum
area of forest plot that can be practically interpreted and delineated
as well as interpretation experience the number and total area of
interpreted forest plots on TM images with 9 kind scales were arranged
in Table 3 and a diagram e\was drawn for finding out the loss of FRIM on
VSTMI see table 3 and fig 5.
It can be seen from Table 3 and
Fig.5 show that the along with reduction of TM image's scales,
information of some small artificial forest plots loses continuously and
their interpretation properties of forest lands reduce rapidly . The
table 3 and fig 5 show that loss percentages of the FRIM is low on
1:500.000 and 1:200.000 scaled TM are 3.3 % and 14.2% respectively and
the loss of total area of forest plots are 0.4% respectively but
beginning at 1:250.000 scaled TM image the loss percentage of FRIM image
increase rapidly along with the reduction of TM image scale. This shows
that using 1:250.000 or smaller scaled TM image to carry out FRIT can
not satisfy the precision requirements of forestry production in forest
lack region with small artificial forests scatteredly distributed so
these scaled TM images can only be used for reference materials.
Table 3. Statistical table of number and area of
interpretable forest plots on TM images 9 kind scales.
Scale of TM iamge |
Minimum delineated area(ha.) |
Number of forest plots |
Total area(ha.) |
1 : 100,000 |
3.0 |
332 |
9225 |
1 : 150,000 |
4.5 |
321 |
9187 |
1 : 200,000 |
8.0 |
285 |
8970 |
1 : 250,000 |
12.5 |
218 |
8295 |
1 : 300,000 |
18.0 |
148 |
7584 |
1 : 350,000 |
24.5 |
87 |
6032 |
1 : 400,000 |
32.0 |
48 |
4967 |
1 : 450,000 |
40.5 |
22 |
4019 |
1 : 500,000 |
50.0 |
18 |
3842 |
Fig. 5
Diagram of loss percentages of forest information on variously scaled TM
images.
- The above analysis results have displayed the extent of FRIM loss on
VSTMI according to the previous experience of forest interpretation on
variously scaled images and the examination results of CPIFP on
1:1000,000 and 1:200.000 scaled TM images it can be concluded that CIPFP
are very close when the interpretation is carried out within the scope
of respective minimum interpretable and delineative area of VSTMI
therefore FRITP on several scaled TM images were reckoned on the basis
of combination the mean of FRITP on VSTMI table 4 and fig. 6.
The Changing trend of FRITP of VSTMI can be from fig. 6 which
can provide reference for optimum scale selection of TM image for
conducting FRIT three North shelter forest region and to accomplish the
purpose of meeting the precision requirement of forestry production and
scientific research as well as reducing the raising working efficiency
to make economic benefits[3]. ConclusionsThe above
research results show that applying TM image to carry out FRIT is greatly
restricted by TM image scale in forest lack region of China especially in
small and scattered artificial forests dominantly distributed area such as
Kanhbao county so some conclusions obtained as follows.
- The optimum scale of TM image for FRIT in this area id 1:100.000 the
MDA can be 3 ha loss of FRIOM is very little is 92.3% A satisfying
results can be gained if the FRIT is carried out on 1:500.000 scaled TM
image the MDA is 4.5 ha and the FRITP can be 91.5%.
Table 4. Table of reckoned FRITP of TM image with nine
kind scales.
Scale of TM image |
Keeping percentage of FRIM (%) |
CIPFP (%) |
FRITP (%) ** |
1 : 100,000 |
100.0 * |
92.3 |
92.3 |
1 : 150,000 |
99.6 |
91.9 |
91.5 |
1 : 200,000 |
97.2 |
91.4 |
88.9 |
1 : 250,000 |
89.9 |
91.9 |
82.6 |
1 : 300,000 |
82.2 |
91.9 |
75.6 |
1 : 350,000 |
65.4 |
91.9 |
60.1 |
1 : 400,000 |
53.8 |
91.9 |
49.5 |
1 : 450,000 |
43.6 |
91.9 |
40.0 |
1 : 500,000 |
41.7 |
91.9 |
38.3 |
in the table :
|
*------ |
The loss of FRIM on 1 : 100,000 scaled TM image is very
small so, it is neglected here.i.e. suppose keeping percentage
of FRIM on 1 : 100,000 scaled TM image is 100.0%.
|
**------ |
FRITP = keeping percentage of FRIM *
CIPFP. |
Fig. 6
Diagram of FRITP of TM image with nine kind scales.
- Applying 1:200.000 scaled TM image conduct FIRT can meet the
precision requirement of forestry production MDA is 8.0 ha the loss of
FRIM is a few and its FRITP is 89.9% Moreover the costs of purchase
information resource and the office work day for forest survey can be
reduce and the economic benefit is higher.
- TM images with scales of 1:250:000 1:300.000 1:350.000 1:400.000
1:450.00 1.500.000 can not be used to carry out FRIT in three north
Shelter forest region precision too much so the precision requirements
of forestry production can not be satisfied.
In the light of
change trend of FRITP and VSTMI shown in diagram of FRITP there is
possibility on applying 1:250.000 scaled TM image to conduct FRIT in the
forest area with higher forest covering percentage therefore the
research should be further carried out. References
- Zhu Junteng, Natural Resources in "Three North Shelter-Forest
Region" and Integrated Agricultural Division, Forestry Publishing
House(1985), pp 153-154.
- Beijing Forestry College, Mathematical Statistics, Forestry
Publishing House(1979), pp 416-417.
- Donald T. Lauer and Andrew S. Benson, Classification of Forest Lands
with Ultra-High Altitude Small Scale False Color Infrared Photography,
Proceedings Symposium IUFRO S 6.05 Freiburg(1973), pp
156-158.
---------------------------------- 1. Mr Li
Yingguo and Ms. Wu. Honggan Bai Suhua (Research Institute of Resources
Information, Chinese Academy of Forestry) partly participated in this
research work. |