| Radar Imagery for Precision Crop and Soil Management
By: M. Susan Moran, Daniel C. Hymer, Jiaguo Qi, Yann Kerr Studies over the past 25 years have shown that measurements of surface reflectance and temperature are useful for monitoring crop and soil conditions. By the year 2000, there will be about 10 Earth-observation satellites supporting optical sensors with the spatial, spectral, and temporal resolutions suitable for many farm management applications. These optical sensors provide information in the reflective and thermal emissive portions of the electromagnetic spectrum. In a multitude of studies, this information has been used for such important farm applications as scheduling irrigations, predicting crop yields, and detecting certain plant diseases and insect infestations.
Although optical remote sensing is a powerful farm management tool, there are some serious limitations that have restricted farm management applications. An alternative to the use of optical remote sensing for farm management is the use of radar backscattering data obtained from Synthetic Aperture Radar (SAR) sensors. There are currently four SAR sensors aboard polar-orbiting satellites, and there are plans for two more by the year 2000.
SAR sensors measure the spatial distribution of surface reflectivity in the microwave spectrum. The radar transmits a pulse and then measures the time delay and strength of the reflected echo (i.e., amplitude and phase measurements). The amplitude is called the radar backscatter. The scattering behavior of the SAR signal is governed by the dielectric (non-conductive) properties of both soil and vegetation. In addition, soil roughness, leaves, stalks, and fruit create a geometric configuration of the scattering elements that influences the wavelength, direction, and polarization of the incident wave.
The adoption of SAR technology is poorly understood, in part, due to the lack of research on the response of SAR backscatter in relation to soil and plant conditions. Unlike optical sensors, SAR systems provide the advantages of cloud penetration, all-weather coverage, day/night acquisitions, and signal independence of the solar illumination angle. These features allow SAR images to meet the rigid data requirements involved with site-specific farm management decisions.
To capitalize on the abundance of ongoing research covering the applications for optical sensors in site-specific crop monitoring, and develop a better understanding on how SAR images can benefit agriculture, a long-term study was conducted from 1995 to 1997. To facilitate the study, Landsat TM sensor and the ERS-2 SAR sensor images covering the University of Arizona Maricopa Agricultural Center in central Arizona, were acquired. The information obtained from multispectral reflectance and temperature measurements made with the TM sensor was used to interpret the signal received by the ERS-2 C-band SAR sensor. In particular, focus was placed on the determination of within-field variations in ¥soil roughness (related to tillage, subsidence or erosion); ¥vegetation density (related to seeding, crop vigor and pest infestations); ¥surface soil moisture condition (related to monitoring irrigation efficacy, soil texture); plant litter (related to erosion control). BACKGROUND AND THEORY
In the reflective region of the optical spectrum, discrimination of crop growth and plant status is generally accomplished by assessing the reflectance of red and near-infrared reflectance of the plant canopy. Simply put, plants absorb red radiation and scatter near-infrared radiation resulting in a large difference between red and near-infrared reflectance. In contrast, when monitoring bare soil, the multispectral reflectance measured in the red and NIR spectrum is often quite similar. This difference between plant and soil reflectances is often enhanced by computing a ratio of visible and near-infrared reflectances, termed a Vegetation Index (VI). A commonly-used VI is the Soil Adjusted Vegetation Index.
In the thermal region, remotely sensed measurements of soil and foliage temperature have been linked to soil moisture content, plant water stress, and plant transpiration rate. The sensitivity of surface temperature to plant and soil moisture conditions is related primarily to the heat loss associated with evaporation and transpiration. As such, the thermal signal (Ts) is related to the percentage of the site covered by vegetation and the water status of the vegetation and soil (i.e., EvapoTranspiration or ET).
In the microwave region, specifically the C-band SAR wavelength, it is generally assumed that backscatter (s0) is directly related to surface roughness, soil moisture, vegetation density, and the SAR sensor configuration. EXPERIMENT
The site of the Agricultural SAR/Optical Synergy (ASOS) study was the University of Arizona Maricopa Agricultural Center (MAC). MAC is a 770 hectare (ha) research and demonstration farm located about 48 km south of Phoenix. The demonstration farm is composed of large fields (up to 0.27 x 1.6 km) in which alfalfa is grown year-round, cotton is grown during the summer, and wheat is grown during the winter. A data management system is in place to archive planting, harvesting and tillage information, and the times and amounts of water, herbicide and pesticide applications. Since the predominant irrigation method for the MAC demonstration farm is flooding, each field is dissected into level-basin "borders".
The ASOS study was conducted in two parts. A retrospective study was conducted based on existing images in the European Space Agency (ESA) and EROS Data Center (EDC) archives. These images from 1995 and 1996 were ordered with the intent of determining field soil moisture, vegetation cover, and tillage conditions based on the response of the optical and SAR signals, and validating these determinations with the field notes archived by the MAC Farm Manager. A second study was conducted in which the TM/SAR image pairs were captured on three dates in May, June and July of 1997. During all three overpasses, arrangements were in place for one field to be flood irrigated such that a large portion of the field was saturated, and, for contrast, a large portion was completely dry. A kenaf crop was planted in May, and by the June overpass dates, the green leaf area index was 0.3; by the July overpass, the GLAI was 1.5. Vegetation and soil moisture conditions were also monitored in two fields of alfalfa at various growth stages with a variety of soil moisture conditions.
For a number of reasons, the ASOS study did not go as smoothly as planned. First, there were few TM/SAR pairs available in the ESA and EDC archives. Images were available for only November and December 1995 and 1996. During this time of year, there was very little farm activity, and the only crops were alfalfa and emergent wheat. Compounding this setback, though ERS-2 SAR and Landsat TM images were ordered for May, June and July 1997, only one SAR/TM image pair was received. The reason for the failure to obtain the additional images, as ordered, was never revealed. RESULTS AND DISCUSSION
Using all MAC fields in the five 1995-1997 images that had a record of distinctive within-field differences in tillage, soil mois- ture, and vegetation density, it became apparent that results were comparable for fields of similar surface conditions. Based on this observation, four fields were selected. All four fields (numbered 1-4 for reference herein) had a notable increase in the SAR backscatter from one end of the field to the other. The increase in ÆNs0 in Field 1 was due to the increased scattering of the SAR signal resulting from soil roughness. In Field 2, the increase in ÆNs0 resulted from a decrease in the alfalfa crop density due to a recent harvest. In Field 3, ÆNs0 increased due to the change in soil moisture and the sensitivity of the SAR signal to the dielectric quality of the surface. (In the C-band wavelength, the dielectric constant of water is about 80 and that of dry vegetation or soil is about 2-3.) In Field 4, ÆNs0 apparently increased due to an increase in dry crop litter cover from 15 to 75%.
Response of the optical data to the four different field conditions corresponded well. Based on data for the fields, we found that the optical data were also useful for discriminating "mixes" of effects of roughness, vegetation and soil moisture. For example, in the SAR image acquired in November 1995, two adjacent fields of alfalfa showed no difference in SAR backscatter. Yet, we computed large negative values of ÆNTs and ÆNSAVI. Based on the optical response, we postulated that one of the fields was recently harvested and had a low soil moisture content; the other was near full vegetation cover and had been recently irrigated. As a result, the high backscatter associated with low crop cover was offset by the low backscatter associated with high soil moisture content, and ÆNs0~0. Overall, the ÆN indices worked well to discriminate the causal relation between surface conditions and SAR backscatter. Though results for only four fields are illustrated here, similar results for several more fields showed that this method has potential for interpretation of SAR imagery with coincident optical imagery. These results also illustrated the sensitivity of Landsat TM and ERS-2 SAR imagery to differences in tillage, surface soil moisture, vegetation density, and dry crop litter. CONCLUDING REMARKS
The objective of this study was to investigate the utility of SAR images for precision farm management applications. These preliminary results showed that the SAR backscatter was sensitive to differences in field roughness (related to tillage), vegetation density, surface soil moisture, and plant litter. Furthermore, optical imagery obtained coincident with SAR imagery allowed a better understanding of the interactions of the SAR signal with soil and plant surfaces. Thus, it may be possible to model SAR backscatter based on optical measurements rather than the time-consuming ground truth measurements of surface roughness and GLAI. Future work on this data set will be focused on compiling the SAR, optical, and field information necessary to develop a relation to facilitate interpretation of the SAR image. About the Authors:
M. Susan Moran, Daniel C. Hymer and Jiaguo Qi are research scientists with the USDA-ARS U.S. Water Conservation Laboratory in Phoenix, Arizona; and Yann Kerr is a research scientist with the Centre d'Etudes Spatiales de la Biosphere Toulouse, France. Back |
