GISdevelopment.net ---> AARS ---> ACRS 1990 ---> Agriculture and Forestry

Grass resources investigation with Remote Sensing in lichuan country

Sun Jiabin, Lu Jian, Guan Zequn and Ma Jiping
Sun Jiabin, Lu Jian, Guan Zequn and Ma Jiping
Wuhan Technical University of Surveying and Mapping
Wuhan, China


Abstract
In the test area of Lichuan country of Hubei province in China, we used multi data (spectral, textural and non-remote sensing) to investigate and evaluate the grass resources. Its reliability is more than 90%. This paper describes the results of the research which will provide a scientific for the local government to develop and utilize grass resources and to plan, and mange livestock farms.

General situation of test region
We select Lichuan country of Hubei province in China as a test area. It is located in El 108021' to 109019' and NL 29042 to 30040'. The number of image lines are from 1650 to 5749 and number of pixels from 620 to 3767 in Pss-Row of 126-39. TM. It is at the intersection of Daba and Wulin mountains. It is located in the wet climatic zone of north subtropical zone an south temperate zone. Spring is later and autumn is earlier than that of other areas. Its area is 4612 km2. The highest elevation point is 2041m and the lowest elevation point is 315m. However most land is between 1000m - 2000m. The climate and plants appear vertically different distribution. Usually crops are the staple plants intermingled with forest and waste land in the area lower than 1200m. There are barren hills, brush covered slopes and grasslands, except that the woodlands and farmlands appear in the area higher than 1200m. Large grassland with area more than 7000 ha. are distributed on the wide and smooth top of mountain and mountain slopes. There are various kinds of lush grasses growing, which are suitable for developing large livestock farms.

Grassland taxonomy in the south of China
The international taxonomy of the grassland in southern part of china can be classified on 3 levels - CLASS, GROUP, TYPE.

At the first level grassland is divided into 5 classes :

(1) Grassland
(2) Bush-grassland
(3) Wood-grassland
(4) Meadow
(5) Odd Pieces of grassland.

There are only three classes (1), (2), (3) in Lichuan Country.

At the second level every class is divided into 3 groups :

(1) High mountain group, in which the terrain is higher than 1200m.
(2) Mid-mountain group, in which the terrain is between 800m and 1200m
(3) Low mountain group, in which the terrain is lower than 800m

At the third level grassland TYPEs are determined on the basis of grass format.

Classification of grass resources
  1. Classification Based on Spectral Feature

    The spectral bands TM3, TM4 and TM5 and used in classification of spectral feature. They are determined by feature selection. The algorithm of classification is maximum likelihood method:


    The results of classification are shown in table 1.

    Table 1 : the confuse matrix of 7 classes
    Output

    Classes    		 1
    2
    3
    4
    5
    6
    7
    other    		 Real classes 
    1 2 3 4 5 6 7
    82.0 5.3 0.6 0 0 0.6 0
    11.3 86.6 1.70 0 0 0 0
    0 2.4 78.6 0 0 0 3.7
    0 0 0 81.6 0 0.9 0
    0 0 0 0 84.8 0 17.0
    0 0 0 0.4 0 87.3 0
    0 0 14.4 0 9.4 0 74.6
    6.7 5.7 4.6 7.9 5.8 11.2 4.6
    Number pixel 247 246 1119 280 831 322 753

    Where:

    1. Rice (store water in all season)
    2. Rice (9stoe water in some season)
    3. Forest
    4. Town
    5. Grassland
    6. Water
    7. Bushland

  2. Classification Board on Textural Feature

    Some classes of grassland are still seriously confused after classification based on spectral feature, such as bushland, vegetables, wood-grassland and grassland or bushland. But their texture feature are quite different. So we use the texture measuring to distinguish the various confused classes. For this reason, every 4x4 or 8x8 window of image with confused classes will be transformed into spatial frequency domain, one by one.

    F(u,v) =FFT (f(x,y))

    Then we calculate their texture:

    TXF = dsi / Fsi (0,0)

    Or the texture in spatial domain:

    TXs = dsi / Msi

    Table 2 shows some measures of texture in 2 typical areas (Qiyu mountain and Fubao mountain).

    Table 2: Texture measures of 5 classes
    Class Measure of texture
    TXS TXF
    Wood grassland 12 12
    Forest 6 15
    Grassland 8 10
    Vegetable plot 10 11
    Bush-grassland 8 18

    Base on the use of feature of spectrum and texture in image processing 3 classes of grassland at first level are extracted.

  3. Classification Based on Non-Remote Sensed Data

    At this is second level of classification, each class is classified into 3 groups by DHM and soil types, and natural extension of grassland should be considered at the same time. Finally, a grass resources distribution map of Lichuan country at scale 1:100000 has been done by computer. Table 3 shows the area of 9 groups in grass resources map.

    Table 3. Area of 9 groups grassland in Lichuan country
      Grassland
    (ha)    		 Bush grassland
    (ha)    		 Wood grassland
    (ha)    		 Total
    (ha)
    High mountain groups 34444 30432 51277 118153
    Mid mountain groups 807 825 7954 9388
    Low mountain groups 877 20 0 897
    Total 35928 31277 59231 126438

    A comparision between results of the computer classification and from field survey in this area are given in Table 4.

    Table 4. Area comparison for each kind of grassland between the computer classification and field identification.
    Classes of grassland Area of computer classification
    (ha)    		 Area of field survey *1
    (ha)    		 Comfortable percentage
    (%)    		 Statistical are *2
    (ha)    		 Comfortable percentage
    (%)
    Grassland 35928 37090 96.87 38423 93.81
    Bush grassland 31277 28361 90.68 26172 93.6
    Wood grassland 59231 60840 97.36 63018 93.99
    Total 126436 126291 99.89 127613 99.07

    *1 From livestock burean of Lichuan,
    *2 From livestock burnean of Hubei

  4. Classification Based on Grass Format

    In the third level classification we use traditional methods such as
    1. Collecting sample-square of grass.
    2. Making specimen of forage grass.
    3. Recognizing and extracting superior species of forage grass in grass format.

    Table 5. show grass types of 3 large stretches of grassland.
    Name of grassland Elevation (m) Forage grass tape
    Hanchi 1910-1951 Anaphalls contocta-Arthraxon hispidium erlophorum consum pteridium aquilinum vbunum macrocephalum fortums Artemisia aplacea hance.
    Mashan 1500-1700 Miscanhus sinesis arderspteridium aquillinum anaphalls contocta paspalum thunbergi
    1700 Anapalis contocta - Preridium aquilin amcacalia tanquita - Arthraxon hispidus imperate cylindrica.
    Qiyushan 1543-1784 Miscanthus sinesis arderss cotonease adpressusbois eriophorum canosum - Artemisia aplacea hanc-anaphallis salix dunnil ochheld arthaxon hispidus - Pterdius aquilinum.
Evaluation of grass resources
First we used various kinds of data to build a database for grass resources evaluation this country, These data are form remote sensing images and non-remote sensing data such as hydrology, soil types, landuse types, elevation, slope, climate, agriculture economy etc and then we used computer to do the grass resources evaluation with various mathematical analysis models. The results will provide a scientific basis for the local government to develop and utilize grass resources.
  1. Region Similarity Analysis Model.

    On the basis of geographical network system, Lichuan country is divided into 57 grids (10x10 km). Corresponding data in database are stored for every grid.

    Generally, there are two steps for region similarity analysis:

    Step 1:  To formulate fuzzy similarity matrix for 57 grids


    X(i,k) is the kth index in the ith grid
    A(i,j) is similarity - degree of the ith and jth grid

    Setp 2: By using maximal brace three and fuzzy cluster, Lichuan country is divided into 7 regions and their natural condition and direction of development are similar. See Fig. 1.

  2. Region Evaluation Model for Livestock-Farming Suitability

    The single condition parameter suitability is expressed as

    And the general suitability is


    Where
    a(i), b(i), is not permitted value, satisfaction value, and weight respectively; d Iij) is the actual value of the jth parameter of the ith grid. According to the modular analysis by computer, we get the final result and point out that thirteens areas of Lichuan county are suitable for construction and development of livestock farm. See Fig.2.




    Conclusion

    1. The classification reliability will we improved effectively if using multi-data to recognize grassland.

    2. The evaluation grass resource by using multi-criteria will make more objective, and reasonable result.

    References

    1. Yang Kai, Sun Jiabing, et. al: Principle and Methods of Image Processing in Remote Sening, Beijing, Publishing House of Surveying and Mapping. 1988.

    2. Liao Guofan, Xiang Yuanquin, et. al: Grass Resource of Hubei Province, Livestock Bureau of Hubei Province, 1985.

    3. Robert A Schowengert: Techniques for Image Processing and Classification m Remote sensing, New York Academic Press 1983.

    4. Yang Kai, Chen Jun: The use of Auxiliary information in the classification of the Remote sensed data, Remote Sensing of Environment 1986.

    5. Sun Jiabing et. al: Proceedings of ISPRS. Comm. III. May. 20 1990.