Altona, Manitoba, Canada , SIR-C/X-SAR Experiment
T.J.Pultz, R.J.Brown
Canada Centre for Remote Sensing
Y.Crevier
Intermap Technologies Ltd.
J.Boisvert
Agriculture and Agro-Food Canada
Q.H.J.Gwyn
University of Sherbrooke
R.Protz
University of Guelph
Objectives
The Canada Centre for Remote Sensing (CCRS) participated in the SIR-C/X-SAR mission by conducting an experiment during April 10-18 and October 2-8, 1994 over an agricultural test site centred on Altona, Manitoba, Canada. The project was a co-operative venture amongst CCRS, Agriculture and Agri-Food Canada and the Universities of Sherbrooke, Guelph and Waterloo.
The overall objectives of the experiment were to: i) evaluate the capabilities of multi-date, multi-parameter SAR data to estimate soil moisture in an agricultural environment for a variety of soil types in the spring and fall; ii) develop models which provide a better understanding of the relationships between soil moisture and texture, surface roughness and radar frequency, polarization and incidence angle; iii) evaluate the use of a change detection approach for soil moisture monitoring and; iv) gain a better understanding of the information content of polarimetric data.
Data Acquired
Table 1 indicates the SIR-C acquisitions over
the Altona test site. The SIR-C SAR data were processed to create
a multi-looked (1 look in range x 3 look in azimuth) images (Chapman,
1994). The number of looks was selected to create images with
relatively square pixels and to maintain resolution. The multi-looked
images were then filtered by applying a 5 by 5 GAMMA filter which
reduces the coherent noise while preserving strong scatters and
structural features (roads, field boundaries) (Lopes et al.,
1993). The filtered images were then rectified in UTM cartographic
projection with a rigorous mathematical modelling method which
takes into account all the distortions inherent to the platform,
the sensor, and the earth (Toutin, 1995). For SIR-C/X-SAR, a
mean altitude value, representative of the flat area (15 m maximum
difference in height), was used to produce accurate ortho-images
for the study area.
All fields within the study site were photographed
and qualitatively described as to the surface roughness characteristics
and amount of crop residue cover. Infrared air photographs were
acquired over the site on April 10 and October 10.
Meteorological data were recorded using a
portable weather station deployed at the Altona Flying Club. The
station recorded hourly data on air temperature, soil temperature
(measured at the surface and depths of 3 and 7 cm), humidity,
wind speed and direction, solar radiation and precipitation.
Twelve bare soil fields were sampled to determine
soil moisture during each SAR data acquisition. The fields were
selected so as to get a wide range of soil textures, which increased
the variability associated with soil moisture conditions. Each
field was sampled at 5 sites, with a minimum distance between
sites of 50 m., along a transect which typically followed the
long axis of the field. At each of the sample sites 3 replicate
samples were taken at depths of 2.5, 5, 10 and 15 cm. Soil moisture
values were obtained using traditional gravimetric sampling, time
domain reflectometry instruments (TDR) and a portable dielectric
probe (PDP). Gravimetric samples were used to estimate the soil
moisture of the top 2.5 cm of the soil profile. The TDR instruments
were used to sample soil dielectric values over depths of 5, 10
and 15 cm profiles. These data are then converted into average
volumetric soil moisture content using the relationship established
by Topp et al. (1980). The PDP measures the complex dielectric
constant over a small depth (approximately 1 cm) and as such was
used to generate a depth profile from the soil surface to a depth
of 8 cm at 2 cm intervals. This profile can then be converted
into a volumetric soil moisture profile using algorithms developed
by Brisco et al. (1991). The gravimetric samples were also
used for further soil analysis which included bulk density, organic
matter content and particle size distribution.
Surface roughness measurements were obtained for each of the fields sampled for soil moisture using an instrument developed at CCRS which is described by Brisco et al. (1989). This instrument is designed to measure roughness (RMS height and correlation length) along 50 cm lengths on the ground using a photographic technique. Measurements were made at 3 sample sites within each field. The roughness measurements were made along the range look direction for the ascending and descending SIR-C orbits.
The percent crop residue cover on the soil moisture sampled fields was estimated using a simple technique called the knotted rope method. With this method, a seven meter rope with 50 equally space knots is laid diagonally across the tillage rows. Each knot that touches or lies directly above a piece of residue counts for 2% residue cover. These data were collected at each of the 3 surface roughness sample locations.
Results and Discussion
Environmental Effects
The effects on radar backscatter of two environmental events were evaluated. The first was a frost event which occured on April 10 and the second was a heavy rain (75 mm) on October 7. In order to compare data acquired at different incidence angles, the effect of incidence angle on radar backscatter needs to be evaluated under stable ground conditions where the only varying parameter is the SAR imaging geometry configuration. For this analysis, October data which presented relatively constant soil moisture conditions between October 2-6 and which had incidence angles varying from 38 to 54 were used.
First, surface roughness for the site had to be characterized, since the decrease in radar backscatter with increasing angle of incidence is a function of the the surface roughness of the target. Surface roughness was classified using a method presented by Ulaby et al. (1982) which takes into account the wavelength. A surface may be considered smooth if the index kd < 0.2, and may be considered very rough if kd > 1.0, where the wave number k = 2/, is the wavelength and d is the standard deviation of the surface height. It was observed that all the sampled fields were categorized as being very rough at C-band and rough at L-Band. The relationship between radar backscatter as a function of incidence angle was therefore computed for these classes of surface roughness. This relationship was then used to correct the imagery for incidence angle effects.
Since each of the sampled fields were classified as very rough at C-band and rough at L-Band (Ulaby et al., 1982), the curves derived are appropriate only for these roughness classes. For example, a decrease in backscatter of 3dB was observed between incidence angles of 38 and 54 at C-HH. Assuming that this relationship is linear over this range of incidence angles, the rate of change is 0.2dB/ of incidence angle. This rate of change in radar backscatter as a function of incidence angle is in agreement with scatterometter observations reported by Brisco et al. (1991) over a range of incidence angles from 30 to 60.
Using these relationships the data were corrected for incidence angle effects so that comparasons of data acquisitions could be made at various incidence angles with minimal concern for angular effects. This correction permits the evaluation of changes in backscatter related only to changes in environmental conditions. The incidence angle corrected field mean C-HH and L-HH temporal curves are illustrated in Figure 1 (April) and Figure 2 (October).
As can be seen of Figure 1 there is a decrease in radar backscatter from April 11 to April 18 which reflects the drying of the soil over this period. However, the data for April 10, at which time soil moisture values were at their highest, display values typical for drier soil conditions as observed later in the experiment. This behavior is the result of a frost event which resulted in freezing of the soil surface. During the night and early morning the portable weather station recorded temperatures just below freezing at a soil depth of approximately 2 cm. The SAR data acquisition was at 8:55 local time
Table 2 indicates the average decrease in radar backscatter for all of the sampled fields between April 10 (frozen layer) and 11(unfrozen). The observed change in radar backscatter for data collected at an incidence angle of 33 on April 10 and at 38 on April 11 are shown as well as the values corrected for incidence angle. The greatest decrease in backscatter occurred in the L-Band observations, specifically at HH polarization (3.7 dB). The C-HV data displayed the largest decrease of the C-Band observations (2.8 dB). The difference between the fwo frequencies may be related to the actual conditions during the SIR-C data acquisition at 8:55. The PDP data collected at approximately 8:00 to 8:40 reflect the frozen conditions but by 8:55 the temperature was increasing and thawing the soil surface nonuniformly over the site. As such, the SAR data were most likely collected during the period when although the 0-2 cm soil layer was still primarily frozen, there was some thawing beginning to occur at the soil surface. Therefore, the C-Band data with a shallower soil penetration depth related to the shorter wavelength relative to L-Band may be in part responding to this thawing. The greater sensitivity observed at C-HV suggests there may be a larger volume scattering component, which would be more sensitive to the composition of ice, water, air and soil particles within the 0-2.5cm layer.
Figure 2 illustrates the relatively unchanging soil moisture conditions and radar backscatter over the period of October 2-6. The heavy rain event which occured prior to the October 7 acquistion is cleary evident in the C-Band data (5-6 dB increase in backscatter) but not in the L-HH data. Table 3 shows the average increase in radar backscatter between October 6 (dry) and 7 (wet) due to the rain event. The observed values collected at an incidence angle of 54 on October 6 and at 56 on October 7 are shown as well as the values corrected for incidence angle. Only, HH and HV data were collected on this data take. Although the L-HH backscatter values do not reflect the rain event the L-HV backscatter did increase approximately 4 dB. One possible explanation for this is that the fields are behaving as specular reflectors at this large incidence angle and high moisture content for HH polarization but not at the HV polarization. As such, the scattering mechanisms have changed and the increase in soil moisture does not result in an increase in backscatter due to the specular reflection from the target. The fields do not behalve as specular reflectors at C-Band due to the greater relative surface roughness at the shorter C-Band wavelength.
Soil Moisture Effects
Correlations between radar backscatter and measured soil moisture using the combined April and October datasets are presented in Table 4. Incidence angle was included as an independant variable in the multiple correlation calculations in order to account for the decrease in radar backscatter with increasing angles of incidence. As stated, the April 10 data take was not used due to the frozen conditions and the October 7 data take was also excluded from the analysis as the heavy rain made the site inaccessable for detailed ground data collection. The strongest correlations (r = .84) were obtained for the 0-2.5 cm layer at HH polarizations and was approximately the same at C- and L-Band. The HV and VV polarizations had very similar relationships with soil moisture. All polarizations displayed a decrease in soil moisture correlation with increasing soil profile depths.
The effect of soil texture and roughness on radar backscatter were then evaluated using particle size analysis. The sites were classified into soil classes according to the textural triangle of the Canadian System of Soil Classification (Agriculture Canada, 1987). Subsequently, the relationship between measured soil moisture for each soil class (sandy- loam, clay-loam and silty-clay-loam) and radar backscatter were computed. The results of the regression computed between measured soil moisture (0-2.5 cm) and radar backscatter for each soil class are shown in Table 5. Very small changes were observed between the slopes and intercepts of the regressions for the three soil type. As such, it can be concluded that within the range of soil texture measured over the Altona site there is no soil texture effect on radar backscatter. Engman and Chauhan (1995) also concluded that the effect of soil texture on the estimation of soil moisture from radar backscatter is relatively small and can be neglected.
It is well known that surface roughness has an affect on radar backscatter. As such, roughness effects within the data were evaluated by examining the improvement in correlation when roughness (RMS height) is included in the multiple regression. The results are shown in Table 6.
It was observed that the correlation with
soil moisture increases when radar backscatter was considered
as a function of both moisture and roughness for VV and HV polarizations.
As such, it can be concluded that data collected at HH polarization
is less sensitive to surface roughness than VV or HV data. This
observation was also reported by Geng et al. (1996) in
their analysis of C-Band airborne data.
As observed by Geng et al. (1996), the contribution of roughness to backscatter was still quite small. This suggests that roughness did not produce a strong overall effect on radar backscatter. However, this observation is probably due to the analysis being based on a limited range of rough surfaces. It is likely that the spatial variation of backscatter due to soil surface roughness would be more significant if the test sites had included both smooth and rough surfaces.
Conclusions
Environmental events such as frost, which result in a lower soil dielectric constant and subsequent decreased radar backscatter, and rain, which results in increased radar backscatter can be monitored. However, care must be taken as some soil targets at high moisture contents may behave as specular reflectors at longer wavelengths, such as L-Band at large incidence angles. As a result soil moisture effects may be masked.
The strongest correlation between radar backscatter and soil moisture were observed for a soil surface layer of 2.5 cm. The C- and L-Band data displayed a very similar correlation with soil moisture with the highest correlation occurring with HH polarized data. The correlation with soil moisture decreased as the sample depth increased. Soil texture effects on the estimation of soil moisture from radar backscatter are relatively small and can be neglected.
The sensitivity to surface roughness is greater
at HV and VV than at HH polarization. However, it is possible
that roughness could be neglected when measuring soil moisture
over relatively short periods of time at a given site without
significantly reducing the sensitivity of the relationship between
radar backscatter and soil moisture.
References
Agriculture Canada Expert Commitee on Soil Survey (1987), The Canadian System on Soil Classification. 2nd ed. Agric. Can. Publ. 1646. 164 pp.
Brisco, B, Brown, R.J., Cihlar, J., Branker, R.S.W. (1989), A Field Instrument for Surface Roughness Measurement. International Geoscience and Remote Sensing Symposium/12th Canadian Symposium on Remote Sensing, Vancouver, B.C, July 10-14, pp. 1177-1180.
Brisco, B., Brown, R.J., Snider, B., Sofko,
G.J., Koehler, J.A. and Waker, A.G., (1991), Tillage Effects on
the Radar Backscattering Coefficient of Grain Stubble Fields.
International Journal of Remote Sensing, Vol. 12, No.11, pp.2283-2298.
Brisco, B., T.J.Pultz, R.J.Brown, G.C.Topp
and W.D.Zebchuk, 1991. "Dielectric Constant Measurements
of Soil with Portable Dielectric Probes and TDR Techniques".
Journal of Water Resources Research, Vol.28, No.5, pp. 1339-1346.
Engman T.E., Chauhan, N. (1995), Status of microwave soil moisture measurements with remote sensing. Remote Sensing of Environment, No.51, pp.189-198.
Geng, Huaibin, Gwyn, Q. Hugh J., Brisco, Brian,
Boisvert, Johanne and Brown, R.J., (1996), Mapping of Soil Moisture
from C-band Radar Images. Canadian Journal of Remote Sensing.
Vol.22, No.1, pp.
Lopes, A., Nezry, E., Touzi, R., Laur, H.
(1993), Structure detection and Statistical Adaptive Speckle
Filtering in SAR Images. International Journal of Remote Sensing,
Vol. 14, No. 9, pp. 1735-1758.
Topp, G.C., Davis, J.L., Annan, A.P., (1980),
Electromagnetic Determination of Soil Water Content : Measurements
in Coaxial Transmission Lines. Water Resources Res., Vol. 16(3),
pp. 547-582.
Toutin, Th. (1995) Multisource data fusion
with an integrated and unified geometric modelling. EARSeL Journal
"Advances in Remote Sensing" Vol. 4, no 2, October 1995,
pp. 118-129.
Ulaby, F.T., Moore, R.K., Fung, A.K. (1982),
Microwave remote sensing: active and passive. Volume II: Radar
remote sensing and surface scattering and emission theory. Artech
House, pp.457-1064
Publications
Boisvert,J.B., T.J.Pultz, Y. Crevier, R.J.
Brown, B.Eilers, 1995. "Potential of Multi-date Imagery for
Soil Moisture, Texture and Drainage Classification : Preliminary
Results". 17th Canadian Symposium on Remote Sensing, Saskatoon,
Saskatchewan, June 13-15, pp.511-515.
Crevier, Y., T.J.Pultz , T.Toutin, 1995. "The
Influence of Data Integration Methodology on Multi-Source SAR
Image Radiometry and Soil Moisture Estimation". 17th Canadian
Symposium on Remote Sensing, Saskatoon, Saskatchewan, June 13-15,
pp.448-453.
Crevier, Yves, T. J. Pultz, R. J. Brown, 1996.
"Towards the Use of Multi-Beam RADARSAT Data for Soil Moisture
Estimation: Results from the CCRS SIR-C/X-SAR Experiment".
26th Remote Sensing of Environment/18th Canadian Symposium on
Remote Sensing. Vancouver, B.C., March, 25-29, pp.320-323.
Crevier, C., T.J.Pultz (in press). "Analysis
of C-Band SIR-C Radar Backscatter Over a Flooded Environment,
Red River, Manitoba". Third International Workshop on Applications
of Remote Sensing in Hydrology, Greenbelt, Maryland. October 16-18.
Pultz, T.J, R.J.Brown, J.Boisvert, Y.Crevier,
R.Duncan, H.McNairn, D.Mullins, D.Randall, D.Wood, P.Vincent,
1994. "The CCRS SIR-C/X SAR Soil Moisture Experiment".
In G.W.Kite, A. Pietroniro and T.J.Pultz (ed.), Application of
Remote Sensing in Hydrology, Second International Workshop, Saskatoon,
Saskatchewan, 18-20, October, 1994, NHRI Symposium No.14, pp.34-42.
Pultz, T.J, R.J.Brown, J.Boisvert, H.Gwyn,
R.Protz, 1995. "SIR-C/X-SAR Observations of Soil Moisture
over The CCRS Altona. Manitoba Test Site". IGARSS'95, July
10-14, Florence, Italy, pp. 990-993.
Pultz, T.J., Y.Crevier, R.J.Brown, J.Boisvert,
1997. "Monitoring of Local Environmental Conditions with
SIR-C/X-SAR". Remote Sensing of Environment, Vol.59, No.4,
pp.248-255.
Figure 1a. Temporal C-HH Backscatter, April
10-18, 1994
Figure 1b. Temporal L-HH Backscatter, April
10-18, 1994
Figure 2a. Temporal C-HH Backscatter, October
2-7, 1994
Figure 2b. Temporal L-HH Backscatter, October
2-7, 1994
Table 1. SAR Data Acquisition
over Altona, Manitoba, 1994
Table 4. Multiple Correlation
Statisitics for Radar Backscatter and soil moisture and incidence
angle. April and October data combined. All correlations significant
>95% confidence level
| Soil type | ||
| Sandy-loam | ||
| Silty-clay-loam | ||
| Clay-loam |
Table 5. Regression Equations
between measured soil moisture (0-2.5cm) and C-HH and L-HH radar
backscatter for three soil texture compositions.
Table 6. Multiple Correlation
Statisitics for Radar Backscatter and 0-2.5 Soil Moisture, Incidence
Angle and Soil Surface Roughness (RMS). (* not significant)