• No results found

The process of dividing the survey population into homogeneous subgroups is called stratification.

Stratification in relation to target variables enables a concentration of resources. Land is not likely to have much or any agriculture, such as deserts, virgin forest, national parks, military reservations, and central urban areas, can be separated from the areas of primary interest and sampled at a low rate, or not at all. Stratification can also be used in cases where different segments sizes need to be used (Cotter and Tomaczak, 1994).

When the objective of using permanent boundaries conflicts in actual practice with the objective of obtaining homogeneous sampling units, permanent boundaries take prece- dence. For example, cultivated fields can be located at the base of a mountain but a good boundary does not exist to include these fields in the appropriate strata, and, therefore, they my be placed in a <15 percent cultivated strata with better boundaries. Roads and rivers make good strata boundaries, while intermittent streams and field edges do not and rarely be used. The geographic features most frequently used for strata boundaries ranked from highest to lowest quality area

• Paved highways,

• Secondary all weather roads,

• Local farm to marked roads,

• Railroads and,

• Permanent rivers and streams (Cheng et al., 1989).

The stratification is performed on a country-by-country basis for administrative purposes. Each stratification analyst works a country until its completion. Stratification generally begins with determining the urban and agric-urban strata for the country. The agriculture areas are then stratified.

The criteria for stratification should be directly associated with the information required. The main stratification criteria for the cultivated land and pastures of an area frame is based on a single characteristic, namely proportion of land cultivated. Additional strata, usually substrata, are formed, using special site or crop specific information. Therefore the strata, in agriculture areas, are defined by proportion of cultivated land, predominance of certain crops, average size of cultivated fields and special sites of agriculture activities (FAO, 1996).

Improving Land Use Survey Method Using High Resolution Satellite Imagery Chapter-5

Table 5-1 displays the complete set of land-use categories which were used in the devel- opment of Missouri’s are frame in 1987. The strata for general cropland can vary slightly depending on the amount of cultivation and agriculture activity in the state (USDA, 1995).

Table 5-1:Land use strata codes and definitions

The area frame stratum definitions used in Honduras (FAO, 1998) are shown in the fol- lowing table.

Table 5-2: The area frame stratum number and definitions.

Another example of stratification used in Sanmatenga province, Burkina Faso (Leeuwen et al., 2000) is as follows.

Stratum Code

Definition 11 General cropland, 75% or more cultivated. 12 General cropland, 50 - 75% cultivated. 20 General cropland, 15 - 49% cultivated.

31 Ag-Urban, less than 15% cultivated, more than 100 dwellings per square mile, residential mixed with agriculture.

32 Residential/Commercial, no cultivation more than 100 dwellings per square mile.

40 Range and pasture, less than 15% cultivated. 50 Non-agricultural, variable size.

62 Water

Stratum Number Stratum Definition 1 61% to 100% Cultivated 2 31% to 60% cultivated 3 0% to 30% cultivated

4 Mostly forest land

5 Non-agricultural land

6.1 Urban areas

6.2 Agro-urban areas

7 Permanent water

Improving Land Use Survey Method Using High Resolution Satellite Imagery Chapter-5

Table 5-3: Descriptions and codes strata

Landsat TM satellite images are considered most suitable for a visual stratification due to their pixel resolution (30m) and the large coverage of 180 x 180 km, which makes it pos- sible to cover a whole province of large parts of it with one image. Nine SPOT images for comparison would be necessary to cover the same area and which would result in higher costs per km2 (Leeuwen, 2000).

Aster image has a potential for a visual stratification due to its resolution (15m) and for reducing cost due to free availability.

In this case, before going to collect the ground truth data, stratification was done by broad visual interpretation with the knowledge in remote sensing regarding image char- acteristics of the satellite image (Aster) in image processing software, Erdas Imagine. Some image characteristics were confirmed with referred to the updated topographical maps. Therefore, different land use land cover characteristics were identified. Non- agricultural areas such as villages, water, escarpment, hills, desert, Zones without agriculture etc were identified by visual interpretation of the image. For stratification, at first, these non-agricultural areas were separated from the areas where agricultural plots are exist.

Agricultural plots were clearly identified on the Aster image, which provides better op- tion for a visual interpretation due to their high pixel resolution (15m). The interpretation elements used for identifying agricultural plots and the methodology followed will be discussed in chapter 8. In this case, the main stratification criteria, for the agricultural

Stratum

Description code Lowland 1

Slopes and rock outcrops 2

Upland plain > 50% fields 3.1

Upland plain and slopes 2 to 50% fields 3.2

Upland plain, < 2% fields 3.3

Excluded zones (villages, water, nature reserves, irri- gated plots, Zones without agriculture)

4

Improving Land Use Survey Method Using High Resolution Satellite Imagery Chapter-5

land of an area frame is based on proportion of agricultural plots. According to this crite- rion, the study area was divided into four different strata as follows.

Table 5-4: Table showing strata descriptions, which are based on proportion of agricultural plots, strata code, total area and agric plot area

To build a strata map, first, the image was saved in a vector layer. Then, using vector tools, the polygons and lines were drawn. Next, the acquired data set were checked using clean-up operations for consistency and completeness. Finally, the topological structure of the vector layer was build. The steps of creating a strata map are shown in Figure 5-4. The stratified map with five different stratums created in image processing software fol- lowing the procedure, as described above, is shown in Figure 5-6.

Stratum Description Code Total Area (Ha) Ag: Plot area (Ha) % Ag: Plots Ag: Plots > 40% 1 6635 3221 49 Ag: Plots 20%- 40% 2 10281 3080 30 Ag: Plots 1%- 20% 3 37428 6521 17 Excluded zones (villages, water,

escarpment, hills, desert, Zones without agriculture)

4 149812 0 0

Improving Land Use Survey Method Using High Resolution Satellite Imagery Chapter-5

Figure 5-4: The steps were followed to create a strata map

Related documents