by Ghassan Mikati
Salinity has always been a major problem for agriculture in semi-arid irrigated areas. Salinity has been a potential hazard in crop production and a major factor in yield reduction. All over the world, agricultural lands have been put out of production due to extensive salinization. To cope with this problem, methods for reclaiming saline soils have been researched and implemented. Drainage and leaching of the root zone are some of the most commonly used practices to reduce salinity and keep agricultural lands in production. However, salinity is a recurrent problem, and areas with a potential to develop salinity should be closely monitored to avoid excessive problems due to salt accumulation in the root zone. The time at which a salinity problem is detected in the field during the growing season and the promptness with which the problem is treated are crucial for guaranteeing a good yield.
Remote sensing can be a fast and useful tool in diagnosing and predicting salt-related crop productivity problems (Rahman 1994). The availability of multispectral satellite imagery from commercial satellites such as the Landsat Thematic Mapper (TM), the MultiSpectral Scanner (MSS), and the Systeme Probatoire pour l'Observation de la Terre (SPOT) originated the development of applications for detecting soil information with improved high spatial resolution (Su,1989). Satellite imagery, however, has a relatively coarse resolution that makes it inadequate for use in delineating problems on a small scale, essential for a good salinity monitoring.
In order to use satellite imagery to identify and map salinity affected areas, a methodology need to be developed that includes a way to overcome the spatial resolution problems of space- based multispectral imagery. One possibility is to use high resolution airborne multispectral imagery in a carefully controlled experiment that would include concurrent satellite imagery. In this way, the effects of salinity in crop growth (decrease in biomass and leaf area)could be better assessed and the confounding issues such as irrigation non-uniformities as well as soil physical properties could be examined and addressed.
Airborne multispectral videography is a new technology which can provide low-cost high resolution imagery for monitoring irrigated agriculture and can be used to provide a better understanding of the limitations and advantages of existing commercially available satellite imagery. The USU airborne multispectral video/radiometer remote sensing system (Neale, 1991) acquires high resolution multispectral imagery and thus will be used to help delineating salinity on satellite imagery by using its high resolution video imagery as reference areas on the satellite image.
Data is collected using the USU airborne multispectral video/radiometer system (Neale,1991). The data is taken from a flight altitude of approximately 4000 m (13,000 ft) above sea, which resulted in an approximate 2000m above the ground level. This cameras are equipped with 16 mm camera lens at focal aperture of 11. The elevation of the flight, the camera lens and the focal aperture resulted in pixel sizes of 2 meters. The direction of the flight lines was East-West and were planned to result in approximately 35% overlap between adjacent flight lines. The data were collected in June 1992, under clear sky conditions.
The video images of each spectral band tape will be digitized with a special software (DIAQUEST) and a frame-grabbing board (TARGA+) using a personal computer. The image video tapes will be played on a super VHS video cassette editor that is controlled by a software on the computer. The single band images from the videotapes that represent the green, red, near infrared and the thermal infrared are displayed on an RGB monitor connected to the TARGA+ board. The images will thus be monitored and digitized as is desired. The video images will thus be digitized using a frame-grabbing board and software (Neale, 1991). The analog video images will be digitized so that images following each other will have at least 60% overlap, a necessary step for further processing. After grabbing, the digital images (TGA files) will be converted to ERDAS compatible format (LAN files).
Figure 1: Grabbed images in the Green, red and NIR respectively
Figure 2 : Two registered 3band images. Notice the overlap.
The procedure for rectification starts by identifying Ground control points (GCP) for each strip of images on the digital frames and on the orthophotoquad maps using the curser and a digitizing tablet to select corresponding points from the image and map, respectively. Then the strip is rectified to the topographic map using the nearest neighborhood scheme. A RMS error of 0.7 or lower will be maintained for this process. The rectification step was required to give the images real coordinates (Universal Transverse Mercator UTM) and to make it possible to calculate distances and areas directly from the images.
Figure 3 : 3band images stitched together
The second step will be to rectify the small stitched frames to 7.5 minute USGS (orthophotoquads by identifying ground control points (GCP) on the image and on the map. A RMS error for the transform of 2.5 or lower will be maintained. The last task will be to stitch again the small rectified frames with each other forming a large unit map.
Supervised classification was performed on the mosaicked maps using five different classes for surface soil conditions.
The classes are:
- Areas of healthy vegetation
- Areas of surface water and waterlogged unhealthy vegetation
- Areas of moist bare soil
- Areas of dry bare soil
- Areas of visible surface salinity
The following figure shows a classified image strip along with its legend.
PLEASE NOTE : Class 3 is the waterlogged/unhealthy vegetation.