|
|
|
Agriculture / Soil
|
Rice Crop Monitoring Using RADARSAT - simulated SAR Imagery
Gordon C. Staples1, Stephane Rossignol1, Dennis Nazarenko1, Greg Elms1, Chao Wang2, Huadong Guo2, Ron Brown3, Brian Brisco4
1 RADARSAT International, Richmond, British Columbia, Canada
2 Institute of Remote Sensing Applications, Beijing, China
3Canada Centre for Remote Sensing, Ottawa, Ontario, Canada
4Intera Technologies, Ottawa, Ontario, Canada
Abstract
Airborne SAR imagery, acquired of Sihui, Zhao Qing, China in November 1993, was modified to produce a RADARSAT simulated Standard Mode (25 m resolution) image, Ground truth information was used to identify mature rice crops, rice stubble, banana crops, and aquaculture ponds. Land use class distributions from the RADARSAT simulated image were analyzed. Comparison of the mean DB and histograms of each land use class indicated significant differences, but with out a priori knowledge of land use, unsupervised classification was not possible. Orbati planning software predicting RADARSAT standard mode coverage indicated an adequate number of viewing opportunities were available to support operational multitemporal observations during the rice-crop life cycle.
1. Introductiond
China, India and Indonesia are the leading rice producing countries with China and India accounting for 50% of the world's supply of rice (1). The socioeconomic importance of rice combined with urban growth and the requirement for cost effective monitoring of the rice crop have lead these countries to investigate methods to augment traditional land based monitoring with space borne sensors. Effective rice monitoring programs should provide information on the growth stage of the rice, the planted and fallow acreage, and the crop yield. Previous studies using space borne optical sensors (Landsat and SPOT) provided high resolution multi spectral imagery for crop discrimination and rice crop growth stage identification. Multi spectral Advanced Very High Resolution Radiometer (AVHRR) imagery can provide extended spatial coverage at the expense of reduced resolution. Optical sensors, however, are limited to reasonably could-free viewing conditions which poses a significant constraint for operational rice crop monitoring.
The cultivation cycle as defined by identifies five stages for wetland rice growth: fallow, flooded, growing crop, ripening crop, and stubble investigations using ERS-1, SAR and airborne X-band SAR data have identified the rice growth stages based on the backscatter variability. Using VV polarised ERS-1 SAR, studies showed that low backscatter from specular reflection in the flooded stage increased dramatically due to corner like reflections and volume scattering during the growing through ripening rice crop stages, combined with constraints requiring observations at appropriate time intervals, makes rice monitoring a viable activity following the launch of the RADARSAT satellite.
With the opening launch of RADARSAT in early 1995, we identified two major objectives for the rice crop study. The first objective focused on technical capability issues - understanding the utility of RADARSAT data to monitoring rice crops and to delineate and differentiate rice fields from the crops and land use classes. This objective was investigated using a RADARSAT simulated image of th Sihui Country, Zhao Qing, China. The second objective dealt with ligistical capability issues - developing an operational scenario for rice crop monitoring using a orbital planning software routine which predicts RADARSAT coverage as a function of user - selected incidence angles, resolution and coverage frequency.
The outline of this paper is as follows. Section II summarizes the geography of Sihui Country, followed by descriptions of the data sources and data analysis techniques in Section III. In addition, the Intera SARPAC RADARSAT simulation process is briefly discussed. Section IV presents image analysis and processing results and observations. Sections V develops the operational coverage of RADARSAT over the study region as a function of incidence angle and beam modes. The final section offers concluding comments.
2. The Study Area
Sihui Country, located in the southeast corner of china, is in the Southern Uplands geographic region, this region is fairly mountainous with the only level terrain located in the delta of the River Xijiang. Nutrient-rich soils and a tropical climate support active agriculture along the delta, but in the mountainous regions, little land can cultivated, even by terracing. Rainfall in the southeast corner is the heaviest in China and averages 100 to 200 cm per year. Rice is a dominant crop in this region , and the precise dates for planting and harvesting vary depending on the weather and cultivation practices. The season begins typically in May or June and concides with the onset of the main monsoon showers. Traditional varieties of rice require four to five months to reach the harvest stage, while modern varieties require about three months to reach harvest. Typically, two rice crops are grown during the summer months while a third crop is usually planted in driest months just prior to the monsoon season. The availability of water for irrigation, however is the limiting constraints on the number of crops planted per year.
Fig1. Teh Zhao Qing study area. The town of Gangli is indicated by the letter G
3. Data
A. Airborne SAR Data
As part of the Globe SAR program, airborne SAR imagery was acquired along the Rive Xijiang on November 20, 1993 from the CV-580 aircraft operated by the Canada Centre for Remote Sensing (CCRS). The CCRS CV-580 aircraft acquired Nadir Mode C-band HH polarized imagery with a 6m nominal resolution (range and azimuth) and 4 m nominal pixel characterized by nadir viewing in the near range with approximately 760 incidence angel in the far range, and a swath width of roughly 22 km. Post processing of the airborne SAR image included modification to create a simulated RADARSAT standard mode image. The airborne data are initially converted from slant to ground range and corrected for antenna pattern effect. The SARPAC routine takes the airborne C-band HH polarized image as input, and using a bilinear interpolation procedure, outputs a RADARSAT - simulated standard Mode image (25 m resolution and 12.5 m pixel spacing). The SARPAC routine also applies a "speckle ball" to the image to simulate radiometric characteristics of RADARSAT.
Once launched the RADARSAT standard mode data (100 km x 100m scene size) will consist of seven distinct beam positions, with stepped incidence angles that vary from 200 in the near range to 490 in the far range, encompassing a total swath width of 500 km. Although the airborne data are acquired in incidence angles from nadir to approximately 760 in the far range, the RADARSAT simulated image is readiometrically balanced to represent incident angle variations through the larger incident angle range of the RADARSAT standard mode (approximately 400 to 490)
|
|
|
|
|
|
|
