2004 
— 
ient", 
nsing 
ysical 
"nsity 
TION 
ARY 
is, an 
onent 
18.sh 
ch for 
>mote 
shtm 
  
CROP AREA ESTIMATION USING REMOTE SENSING ON TWO-STAGE STRATIFIED 
SAMPLING 
Wu Bingfang, Li Qiangzi 
Institute of Remote Sensing Applications, Chinese Academy of Sciences, Beijing, China, 100101, wubf@irsa.ac.cn 
Commission VI, WG VII/3 
KEY WORDS: crop acreage, remote sensing, and stratified sampling 
ABSTRACT: 
In this paper, we analyses problems that remote sensing technique met in China. And provide a new methodology on stratified 
sampling technique. During the stratification procedure, physical factors, such as temperature, precipitation, soil type, sun eradiation 
was considered as well as proportions of main crop types. And then, we first estimate crop proportion using cluster sampling assisted 
by remotely sensed images. Secondly, we estimate crop type proportions of different crop types using transect sampling and GVG 
survey system. Here, transect sampling is a two-stage sampling in fact. In the first step, PSUs were selected randomly from a 4KM * 
4KM area frame. And in the second stage, road segments were selected to survey crop type proportions. At last, crop area was 
calculated under the support of current 100,000-scaled land resource database. And a case study of early rice area estimation in 2003 
showed this methodology was efficient and accuracy enough to meet the running of CCWS. 
1. INTRODUCTION 
Monitoring, estimating and forecasting agricultural production 
are very important for the management of world / regional or 
local food demand and supply balance for social security. China 
has made series of progresses from 1983 after tackling key 
problems with states plan and researches in some institutes and 
universities. Monitoring crop types has increased to including 
wheat, maize and rice with monitoring area increased to 
including 11 provinces. In 1998, China built a CCWS running 
system for crop early monitoring, predicting and estimating. 
Every year, CCWS publish the hot information on crop growing 
status, crop acreage, crop production, crop structure changing 
and multiple-plant index et al. 
In China, crop acreage estimation using remote sensing always 
met the following problems. (1) Some methods that have great 
precise at small area become invalidate at country level. (2) 
Crop planting and harvesting time are all concentrated in the 
period from April to October every year, time is relatively 
shorter when carrying out crop acreage estimation. So crop 
acreage estimating-methods must be effective and time saving. 
As a result, some method with a large amount of processing 
successfully applied in small area cannot meet the request of 
large region. (3) Some method that successfully applied in 
small area cannot be applied because of their high-cost. 
Therefore, country level crop acreage estimation methodology 
must have the following properties: 
- Accuracy enough to support decision. 
- Speedy enough to meet the time that application section 
required. 
- Frugal enough not to go beyond the ability that application 
section could bear. 
-- Having unified criterion and can be operated easily. 
In this paper, we mainly introduced the crop acreage estimating 
methods used in China Crop Watch System, CCWS. Supported 
by the arable area data in Chinese Resource and Environment 
Database, we constructed a feasible methodology using a two- 
sage sampling to meet Chinese crop estimating request. And 
283 
six years operating practice showed that, this crop acreage 
estimating technique is accuracy, reliable and practically. 
2. METHDOLOGY 
Estimating crop acreage at large region using remotely sensed 
data in operating always meet the following 3 obstacles when 
there is no other techniques assistant. 
---- Remotely sensed data cannot cover the country completely 
in crop growing stage. Not only because the input reason, but 
also the excessive processing-time. For the running of CCWS, 
after image received, only there is 1 or 2 weeks left for image 
pre-processing and analyzing before results publishing, and we 
have no ability to process a large numbers of images required 
for crop acreage estimation in China such a large area. 
---- It's difficult to obtain enough images with proper date and 
spatial resolution for meteorological conditions and limited time 
slot. Sometimes, it is always cloud-covered or raining and we 
cannot receive cloud free or high quality images. Nevertheless, 
image-receiving date is extremely limited for crop acreage 
extraction. In China, the most appreciate image date for 
estimating rice acreage is the 2-5 week after moving. For winter 
wheat acreage estimation is in March and April. For maize 
acreage estimation in Northeast China is the last 10 days in 
August or the first 10-day in September. And for Northeast 
soybean acreage estimation is during the first 10-day in August. 
Because of this, the ratio of acquiring proper remotely sensed 
data is much lower. 
---- Because of the special land-use system, rotation system and 
cadastral pattern in China, there exists the particular cross- 
planting and inter-planting phenomena everywhere. Early rice, 
single rice and late rice are planted crossly in south China. 
Spring wheat, maize and soybean are interlarded in Northeast 
China. Maize and cotton are also interlarded in North China. As 
a result, it’s very difficult to successfully identify crop patches 
by crop type from remotely sensed data directly. As a result, 
crop discrimination using remotely sensed data has maintained 
a very low accuracy in China. It's very difficult to extract the