Paddy Rice Monitoring with RADARSAT-1
2.0 Site and Data Description
2.1 Study Site
The study site is near the city of Zhao Qing (23°02'N, 112°29'E) in the Guangdong Province of South China. This area was used as a test site for the GlobeSAR project and numerous RADARSAT images of the area were acquired (Yun et al, 1995). The site is characterized by its flat topography and wide variety of agricultural crops and plants including rice, bananas, sugar cane, and euryale ferrox. The fields are generally small
and are clearly delineated by the ditches, dikes and roadways which separate them. The weather is cloudy and rainy most of the year, making the soil moisture content quite high. This site has been used for studies in the past and is generally considered to be representative of most regions of southern China where rice production is the dominant representative of most regions of southern China where rice production is the dominant form of agriculture.
The diverse nature of the study site makes it in interesting target for RADAR. Figure 2 shows a RADARSAT image (beam mode S5) of the area of Zhao Qing. This image was acquired on June 5, 1997 at the peak of the growth cycle.
Figure 2: RADARSAT Image of Zhao Qing study site
Rice paddies are found throughout the region as non-uniformly shaped fields, ranging from 50-150 meters wide (A). Land cover classes such as banana crops and grasslands appear very bright as there is a lot of scattering effects related to signal interaction with the trees and large leaves. The response from targets produces a bright signal as seen at point C. The water targets, and aquaculture fields remain dark as a result of specular reflectance (B). There are many other land uses in this area of China, however for the purposes of this study, only the above mentioned classes will be considered.
2.2. Data
The data set used in this project was collected between April and July 1997 and represents a complete rice growing season from flooding to harvest. RADARSAT beam mode Stander 6 was collected every 24 days, based on the repeat orbit of the satellite. This provided a set of five images over the growing season. Due to the dynamic nature and rapid growth rate of rice crops, three more images (Standard 5) were collected between the peak and the end of the growing season to attain a more complete representation of the growth cycle. All the images are descending passes thus eliminating error due to diurnal and time of day effects. The table below summarizes the data collected and analyzed for this study.
Table 1: RADARSAT Data Set
| Date Beam |
q |
Growth Stage |
| 04/25/97 |
S6 |
q=41°-46° |
Flooded/Planted |
| 05/19/97 |
S6 |
q=41°-46° |
Planted |
| 06/05/97 |
S5 |
q=36°-42° |
Vegetative |
| 06/12/97 |
S6 |
q=41°-46° |
Vegetative |
| 06/29/97 |
S5 |
q=36°-42° |
Harvested |
| 07/06/97 |
S6 |
q=41°-46° |
Harvested |
| 07/23/97 |
S5 |
q=36°-42° |
Harvested |
A land cover map was provided by the Institute of Remote Sensing Applications (IRSA) in China. This map, along with ground data and crop type information, were used to determine sample sites of each of the land cover types which were then used for the extraction of the data.
As mentioned earlier, three main land cover classes were chosen for preliminary analysis: rice, bananas, and water. In addition, the response from grassland was considered, as well as a comparison of aquaculture and water targets. It was expected that the water (including aquaculture and river locations) would exhibit little or no change in its dark response throughout the temporal sequence. Similarly, the bananas were anticipated to remain bright throughout the period due to the unchanging nature of the canopy at this time of year. It was also expected that the dynamic nature of the rice growth curve would be seen in the spectral response of the crops over the time sequence. These three classes thus represent a good combination of expected target responses. The results of these hypotheses will be discussed in section 4.0.
