GISdevelopment.net ---> AARS ---> ACRS 1990 ---> Poster Session Q

A study on the application of aerial Remote Sensing technique to calculate uncultivated coefficient of cropland in Tibet

Liu Jiyuan Yu Zhiqian
Institute of Remote Sensing Application,
Chinese Academy of Science,
Beijing China


Abstract
The remote sensing technique has been used successfully in the surveying and calculation of uncultivated coefficient of cropland in Tibet. The key technical steps include: (1) Taking sampling strip aerial photographs (2) Calculating the value of uncultivated coefficients of sample fields in the infrared colour aerial-photographs by digital image processing systems and other techniques (3) setting up the uncultivated coefficient indexing table in different geomorphologic conditions and types of various cultivated lands, etc. (4) Using infrared colour Air-photo and indexing table together to calculate the uncultivated coefficient of each fields. This method has reduced manpower and expenditure. The study also provided efficient methods to investigate the cultivated areas in other mountainous regiosnand plateaus in China.

Because of limitation of natural condition and affection of cultivating history, vast cultivated lands in the mountainous regions and plateaus in our country are irregular, and mixed with a great quantity of uncultivated components, such as balks interval paths, irrigation channels and small broken pieces of wasteland, etc., There are many problems to survey and calculate cultivated land precisely even using large scale maps and images, therefore, it makes land investigation very difficult. As a result, it is also difficult to estimate agricultural yield without accurate area of cultivated land. In order to find a way to measure cultivated lands by the remote sensing technique, we have applied colour infrared aerial photographs and indexing tables together to determine uncultivated coefficient. This method completely took advantage of remote sensing techniques and image processing systems. In addition, only a small group of investigators was demanded to deal with the critical technical problems, and the final solution was simple and easy to be mastered. Using this method, satisfactory results have been gotten in landuse investigation of whole Tibet.

Taking sample colour infrared aerial photographs and getting indexing tables for uncultivated coefficient calculation
In the method of using remote sensing technique to determine uncultivated coefficient of crops, taking colour infrared aerial photographs and setting up indexing tables are critical steps.
  1. Data Acquirement and Field Measurement

    1. Taking colour infrared aerial photographs

      To obtain uncultivated coefficients, the infrared colour aerial photographs were taken by way of sampling strip. The strip of sampling flight should cover regions including all types of geomorgraphic condition and cultivated lands, and every types should be repeated a few times. The time we chose was from June to August when the crops are the most flourishing which will make the colour of cultivated lands in photographs most distinct from that of uncultivated ones. The sample strips included every kind of cultivated lands in the east and the south Tibet, with a total length of 1, 698.6km and an area of 15, 798.2 km2

    2. Spectral reflection measurement and data analysis

      The 100 AX multichannel radiometer was used to measure the spectral reflectance's of cultivated and uncultivated lands in Tibet. The data analysis of the spectral reflectance included: (i) calculating the vegetation index of every kind of land, (ii) calculating clustering tendency of spectral reflectance's of every type of land, (iii) estimating the confidence interval of the spectral reflection. The confidence interval was calculated under the hypothesis that the spectral reflectance follows the normal distribution. The value of the confidence interval was worked out by the following formula:


      In which m is the average value of the spectral reflectance of one kind of land, d is the standard deviation, n is the sample number, m is equal to 0.05 and t0.05 can be found a t-test table . The results of the confidence interval are shown in Table 1.

      Table 1. The distributed area of the spectral reflectance Of crops and uncultivated fields in Shanan Prefecture of Tibet
      TABLE 1 IS MISSING PAGE Q_3_2

      It can be seen from the results the results that using the characteristic data of the three-dimension spectral reflection, we are able to effectively distinguish the lands with crops from balks which are bare or sparsely covered with grass, by means of either computer automatic classification of annual discrimination. However, when the balk is completely covered with grass, it is difficult to catch, we have to do manual discrimination on the high-resolution aerial photographs. This result has built a critical foundation of the method to calculate uncultivated coefficients indoors.

    3. Sampling measurement in the field

      In order to determine the accuracy of the uncultivated coefficient gained by indoor methods, we selected a subsample from sample lands. These subsample lands included every type of lands, and each land had been measured in the fields. Using theodolite and small surveyor's tables, the maps of subsample lands were drawn on the scale of 1:500 or 1:1000. Then, using KP-90 electric planimeter and micro-computer digitizing measurement system, the areas of cultivated lands and whole sampling area were calculated, and finally the uncultivated coefficients were calculated.

  2. The Plan for Calculating Uncultivated Ceofficients Indoors by the Use of Aerial Photographs

    1. Calculating uncultivated coefficient by computer automatic classification.

      The IRSA-II and I2S-101 image processing systems were used to classify the digitized aerial photographs supervisoryly and discriminate the cultivated lands from uncultivated ones. Then the rations of these kinds of land are calculated. Using this method, we could determine uncultivated coefficients accurately when uncultivated lands was bare or sparsely covered with grass.

    2. Uncultivated coefficient calculation under microscpo

      Using ten-time aerial measurement microscope, we measured the length of cultivated and uncultivated parts of the sample lands. Then uncultivated coefficients can be got from tills measurement Because of manually enlarging and discriminating , this method can be used to treat every kind of lands, However the accuracy of this method is poorer than the two formers. This is not only because both the power of the magnifying glass and the smallest scale are definite, but also because the results are worked out by statistics.

      We compare the results gained by the three indoor methods above with results gained by measuring in site. The accuracy analysis of these three methods are shown in Table 2.

      Table 2. Accuracy analysis results by comparing uncultivated coefficients measured indoors with data measured in site.
      TABLE 2 IS MISSING PAGE Q_3_4

      From Table 2, we can see that the largest difference of each method is less 4%, satisfying the technical demand.

      In order to reduce the effect of topographical project, we used long-length camera; in addition, we selected those, which were near the main focus on the photographs as sample lands.

  3. Discriminating geomorphologic condition and types of cultivated lands, editing indexing table

    After the uncultivated coefficients of all sample lands were obtained a study on the method of using sample photographs to calculate the uncultivated coefficients for every pieces of cropland was carried out. The study is described in the following two aspects.

    1. Establish an indexing classification system.

      As an indexing classification system, it should include all kinds of cultivated lands in the investigated region. And every element, included in the system, should be distinguished early in the aeira photographs. So we set up an indexing classification system as Table 3 shows.

      Using the three methods described above, we determined the uncultivated ceofficients of 600 pieces of sample fields. Therefore, every column of each grade contained numbers and uncultivated coefficients of 10-20 pieces of sample fields.

    2. The sampling color infrared photographs and indexing tables of every prefecture

      According to the in-situ investigation and discrimination of aerial photos, we determined the second grade-classified types of every prefecture. Considering the demands of classification system, sample lands of different prefecture were selected according to its own condition. These samples were selected from the cultivated lands whose uncultivated coefficients have been calculated. Also it was required that the samples had enough numbers. Then, according to the demand of each prefecture, the sample photographs were made into copies. By matching the indexing tables of uncultivated coefficients with their corresponding photographs, a complete comparing indexing system of uncultivated coefficient was set up.

    TABLE 3. IS MISSING PAGE Q_3_5
A Practice of using sampling indexing table method to calculate non-cultivation coeffecient, its accuracy and benefit analysis
  1. An Utilization of the Method Using Air-photos and Indexing Table Together

    The method which used color infrared air-photos and indexing table together to calculate uncultivated coefficient is described below: First, get information of investigated fields by air-photos interpretation or measuring in the field. These information include: geomorphologic conditions, slop, irrigation, drainage, with terraced fields or not, and so on. Then, find a similar sample field of the same type from color infrared photos and indexing table. Use the coefficient of this similar sample fields as its coefficient. Or, make a suitable correction for the given coefficient by comparing the small difference between the investigation for the given coefficient by comparing the small difference between the investigated field and the similar sample fields. This correction will make the coefficient accord with the demands more. Having a simple training, all investigators in the fields can well master this method.

  2. Accuracy Analysis of the Method of Using Air-Photos and Indexing Table Together

    This method have been used to investigate parts of Tibet. After measurements in the field have been done, a random inspection for different geomorpholgic conditions and type of cultivated lands was made. The results are listed in Table 4.

  3. Benefit analysis

    The method using air-photos and indexing Coefficient is a completely new method. Its potentiality of application is determined by both its accuracy demands and its practical benefit. As a result, we make Table 5 in order to give a basic estimate of benefit.

    TABLE 4. IS MISSING PAGE Q_3_6

    TABLE 5. IS MISSING PAGE Q_3_6

    Table 5 shows that in a special region with wide area, difficulties of success and scattered croplant, such as Tibet Plateau, after forming systematic method to calculate the uncultivated coefficient of croplaned, in the investigation, the method which using sampling IR aeiral photos and indexing table to compare and analysis, instead of the traditional method which surveying pitch by pitch in field, saves field work in large quantity, lessens labor intensity and , in the meantime, saves funds, So, the method has been spareaded in the cultivated land in the cultivated land investigation in fully range of Tibet, whether the cultivated lands are concentrated or scattered.