The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
obvious from the map that the sampling units that belong to the 
same layer are relatively centralized in the area. 
3.1.2 Selecting background data for stratified sampling: 
As soon as the background data for stratified sampling is 
selected, sampling units are also determined. RSAC selects two 
kinds of background data to do stratified sampling. One is 
multiyear statistic data collected by local governments and the 
other is the latest land use data in vector format. 
Processing statistic data: Because statistic data mainly comes 
from local governments, borderlines of administrative units 
such as counties generally have to be selected to act as 
sampling units when the statistic data is used to stratify 
sampling. The main step is to average the multiyear data of 
each county, whether making ascending or descending data 
array totally based on the method of stratified sampling. 
Stratified sampling steps should correspond to the following 
presentation. Each county should be marked with layer sign. 
Processing vector data: When the land use data is selected as 
background data to stratify sampling units, there are some 
things to consider. Firstly, it is necessary to select sampling unit 
such as the frame of relief map used by RSAC. Secondly, it is a 
key step to combine the vector data using land use dada with 
frame data of relief map in GIS, shown below as an example. 
Thirdly, it is to calculate surveyed crop area such as rice, 
soybean, com, cotton, or wheat, etc, that distributes in every 
frame of relief map. The final step is to sort in ascending or 
descending order according to the crop area of every unit if 
necessary. 
China 
Sou Hi China 
and mmwiunt umpM 
Figure 2. Layers distribution of late rice of 2007 in China. 
3.1.3 Methods of stratified sampling: After the layer 
amount is determined, the next step is to stratify. Two stratify 
ways used by RSAC are called Frequency Accumulation Means 
(FAM) and Systematic Clustering Means (SCM). 
Middle part China 
FREQUENCY ACCUMULATION MEANS (FAM) 
RSAC mainly uses this method to stratify sampling units in area 
survey of cotton, rice, soybean, etc. The essence of the method 
includes dividing data into groups according to a certain step 
length between two groups in ascending or descending data 
array, calculating the amount of sampling units of each group, 
which is called frequency, accumulating frequency and square 
root of frequency of each group, getting the total accumulating 
value of square root by adding up square roots of frequency of 
total groups, obtaining the step length between two layers 
through dividing the total accumulating value of square root by 
layer amount such as six, which is an equal step length method. 
The thresholds of layers have been shown in table 2 and they 
will be used for segment point marking each layer. 
When the background data is processed to be either in 
ascending or descending data array, the following steps stay the 
same whether it is statistic data or land use data, as shown 
below by the example of the process of stratified sampling of 
early rice of 2007 in China. Table 2 shows the key points: the 
first column on the left side is codes list of frame of relief map, 
the second is the crop area distributed in frames of relief map. 
codes of 
relief 
map 
Area of 
rice(ha.) 
Frequency 
/00 
Accumulating 
Vx/09 
Thresholds 
of layers 
Layers 
7500123 
0.04 
0 
0.00 
1 
6491261 
2.3 
7 
2.65 
1 
6490013 
4058.72 
1 
7500252 
4060.22 
3 
9286.06 
9288.83 
1 
7500314 
4063.05 
1 
9316.52 
2 
7490053 
6547.84 
2 
7491214 
6550.25 
2 
18570.81 
18577.65 
2 
8501173 
6554.17 
3 
6500141 
9534.13 
3 
7491283 
9537.54 
2 
27842.35 
27866.48 
3 
7500861 
9561.18 
1 
27884.69 
4 
7490562 
13310.9 
1 
37078.12 
4 
9501293 
13322.8 
1 
37123.62 
37155.30 
4 
7500283 
13328.1 
1 
37169.14 
5 
8491412 
19009.5 
1 
46379.36 
5 
8501382 
19027.2 
1 
46427.30 
46444.13 
5 
849.174 
19033.9 
1 
46475.24 
6 
8500182 
40091.7 
1 
55732.95 
55732.95 
6 
Table 2. The process of stratified sampling 
Grouping: Dividing data into groups according to a certain step 
length between two groups. RSAC has selected 4.5 hectare as 
the step length between two groups, which is about one ten- 
thousandth of a frame area, produced 1550 groups from the 
sampling units of 2497, marking at the position of each group 
before the threshold of next group in the data array using code, 
which position can be called Position of Group Threshold(PGT), 
correspondingly producing a new column which is omitted in 
table 2.