A GIS using remotely sensed data for identification of ...

1 downloads 0 Views 373KB Size Report
(3), MASCHMEDT David (4), BISHOP Lyall (2). (1) CSIRO Land & Water, Waite Road, Urrbrae, SA 5064, Australia. (2) University of South Australia, North ...
Scientific registration n° : 2134 Symposium n° : 17 Presentation : poster

A GIS using remotely sensed data for identification of soil waterlogging in southern Australia Un SIG utilisant des données de télédétection pour identifier l’engorgement des sols en Australie du Sud DAVIES Philip (1), BRUCE David (2), FITZPATRICK Robert (1), COX James (3), MASCHMEDT David (4), BISHOP Lyall (2) (1) CSIRO Land & Water, Waite Road, Urrbrae, SA 5064, Australia (2) University of South Australia, North Terrace, Adelaide, SA 5000, Australia (3) CRC for Soil & Land Management, Waite Road, Urrbrae, SA 5064, Australia (4) Primary Industries South Australia, Waite Road, Urrbrae, SA 5064, Australia

Abstract The up-scaling capability of active microwave remote sensing combined with GIS techniques and soil landscape 1:50,000 scale maps was assessed for predicting soil waterlogging at catchment and district scales. We tested the methods in an area of approximately 7500 ha in the Mt Lofty Ranges of South Australia. Soil dielectric constant (a surrogate for volumetric soil moisture) was derived from L-band, polarimetric AIRSAR data acquired for the area. At the same time, a process-based terrain attribute, ln(As/tanβ) was derived from a hydrologically correct DEM of the area. This topographic wetness index provides a computed terrain attribute that is an effective means of spatially determining the trends of soil moisture in the landscape. Initial results suggest that soil landscape units with attributes of aquic conditions are relatable at both catchment and district scale to areas identified from the remote sensing and terrain analysis as being prone to waterlogging. With further detailed analysis, we believe that a set of tools for quantifying certain pedological and hydrological processes in catchments with similar landscapes and vegetative conditions can be developed. Introduction Soil waterlogging and salinity are increasing in catchments in the Mt Lofty Ranges of South Australia as a result of historical land clearance. Consequently, identification of areas prone to waterlogging and salinity in the landscape is essential for effective regional landuse planning using more refined management strategies (e.g. Fitzpatrick et al., 1997). The qualitative agreement between the spatial distribution of soil moisture at catchment scale derived from active microwave remote sensing, computed terrain attributes and GIS analysis of existing digital soil mapping was studied. The aim is to 1

assess this broad scale information as a means of up-scaling our detailed understanding of soil-water-landscape processes from detailed studies at the sub-catchment scale. This will enable us to assess large areas of the Mt Lofty Ranges and predict areas that are prone to waterlogging. Active microwave soil moisture remote sensing methods accurately provide estimates of soil moisture under a variety of topographic and vegetation cover conditions to a depth of about 5 cm (Wang et al., 1986; Pultz et al., 1990; Engman and Chauhan, 1995). Recent studies (Lin et al., 1994; Giacomelli et al., 1995), have related topographic attributes derived from digital elevation models (DEM’s) to airborne synthetic aperture radar (AIRSAR) backscatter characteristics. Terrain analysis techniques within hydrological models are used to estimate catchment flow characteristics (Moore et al., 1992). Topographically derived indices can play a major role as useful and effective methods of determining a catchments susceptibility to water-driven land degradation. Topographic wetness indices such as, ln(As/tanβ) of Beven and Kirkby (1979) are often applied in catchment hydrological models. Despite being based on a number of assumptions (e.g. the water table is parallel to the ground surface, recharge is spatially uniform and there is non-variable drainage), this particular index has proven to be particularly powerful in predicting soil moisture (Quinn et al., 1995). Topographic parameters are a potential method of mapping spatially, the soil water content in a landscape as a derivative of a DEM, bearing in mind the importance of soil, vegetation and weather. Soil landscape unit (SLU) data exists in digital form for the entire Mt Lofty Ranges at a scale of 1:50,000. SLU’s represent areas of land (usually between 10 ha and 10,000 ha) formed on similar geological/parent materials, with characteristic topography, and a defined and limited range of soil types. SLU documentation includes descriptions of topography, component soils and the main features affecting agricultural land use. Land classification tables are also included, which summarise a range of key attributes, for each mapped SLU. A number of these attributes, when combined can be indicative of aquic conditions. Each of the three spatial analysis techniques described above bring a different predictive capability to the determination of soil moisture in the landscape. This study shows how these techniques can be related within the framework of a GIS to enhance their capabilities to predict waterlogging. Methods Study area description The study area is located east of the township of Mt Torrens, in the Mt Lofty Ranges of South Australia and is approximately 75 km2 in extent. The climate is Mediterranean and representative of the eastern part of the Mt Lofty Ranges, with a mean average annual rainfall of 680 mm. The generalised topography of the area consists of valley floors surrounded by rolling hills (Cox et al., 1996). A regional northeast - southwest topographic high east of the township of Mt Torrens bisects the area, with catchments to the west of this high, draining into the Onkaparinga catchment system and catchments to the east forming part of the Murray-Darling Basin catchment system. Soils are mostly Xeralfs with abrupt textural boundaries between sandy and loamy textured surface horizons (A and E horizons) and clayey subsurface horizons with 2

mottled (Btg horizons) and/or sodic (dispersive Btn horizons) properties. Waterlogging often develops in winter in many of the soils, particularly those in the lower slopes to flats. Fritsch and Fitzpatrick (1994), have determined the dominant sequence of soil types (Soil Survey Staff, 1996) occurring down slopes as including: moderately well drained Xeralfs (e.g. Lithic Haploxeralfs) on the crests and upper-slopes, through somewhat poorly drained Xeralfs (e.g. Ochreptic Haploxeralfs, Typic Palexeralfs and Aquic Natrixeralfs) on the mid- to lower-slopes, with very poorly drained Xeralfs (e.g. Mollic and Typic Natraqualfs) represented on the flats. The landcover of the area is predominantly pasture with the majority of the native tree vegetation being cleared during the last century. AIRSAR data processing Microwave energy incident upon an unvegetated soil may be scattered, transmitted or absorbed. Ulaby and Dobson (1989) assessed the degree each of these processes occur and the directional characteristics they acquire are determined by several factors (dielectric properties of the soil medium - related to soil moisture, surface roughness, the incidence angle, frequency and polarisation). All of these factors are interrelated and both frequency and incidence angle dependent. Of importance in soil moisture studies is the fact that the dielectric properties of the soil determines the amount of microwave energy reflected by the soil surface (expressed as the backscattering coefficient, σo). It is this relationship that forms the basis for measuring soil moisture by microwave techniques. For this study multi-polarisation C-, L- and P-band data from the NASA/JPL airborne synthetic aperture radar (AIRSAR) instrument was acquired over the study area for early spring, 1993. The processed data for the site was received from JPL as a 12 band composite image with pixel size of approximately 6.7 m × 8.3 m. The characteristic effects associated with data of this type, i.e. near-range compression, brightness fall-off and radar speckle were all present in the image and it was necessary to correct for some of these geometric and radiometric errors prior to processing for soil moisture determination. In addition the image was resampled to a 10 m pixel size for effective comparison with the other datasets. Soil dielectric constant derivation Dubois et al. (1995) suggested an empirical approach for the modeling of radar backscatter (σo) from the variables of incidence angle (θ), soil dielectric constant (ε), wave number (k), surface roughness (s) and wavelength (λ). The model, which utilises co-polarised data and is expressed in the following relationships: σ 0hh = 10−2.75

cos1.5 θ 0.028ε tan θ 10 ( ks sin θ )1.4 λ0.7 sin5 θ

σ 0 vv = 10 −2.35

cos3 θ 0.046 ε tan θ 10 (kssinθ)1.1 λ 0.7 sin3 θ

was implemented for the L-band of the AIRSAR data. The inversion model is described in more detail by Bruce (1996) and utilises three input raster data sources and produces two output raster images. The first input consisted of the corrected AIRSAR image which had been despeckled and geometrically corrected. The second input consisted of an image of local incidence angle deduced from the radar geometry combined with the

3

DEM described above. This derivation computes the slope and aspect of each DEM cell and uses this as input to a trigonometric solution for the local incidence angle in the direction of the SAR pulse. The final input to the model is a binary mask image in which all objects with large backscatters, particularly in the VV polarisation, are identified so as to not be used in the output calculations. These objects, which are usually stands of native trees, buildings, etc., are masked out of the output data as they are not bare soil (or soil covered by grass). Creation of digital elevation model The primary modelling structure used in the analysis of catchment topography is the DEM, of which there are three primary representations: grids, triangular irregular networks and digital contours (see Weibel and Heller, 1991 for a detailed review of elevation model structures). To determine the spatial distribution of a topographic index for the study area a hydrologically correct, grid DEM was created using the method of Hutchinson (1989). The interpolation technique imposes morphological constraints on the DEM using existing drainage, such that elevations decrease monotonically down each stream line and ridges and streams are represented more accurately (Hutchinson, 1993). The inputs to the interpolation are essentially drainage and elevation data and prior to input, the required coverages were edited and cleaned to remove any errors of logical consistency. For this study a DEM with cell size of 10 m × 10 m was created as the most suitable for simulating geomorphic and hydrological processes (Zhang and Montgomery, 1994), while reducing the bias towards large topographic index values associated with coarser grid sizes (Quinn et al., 1995). Development of topographic wetness index The topographic index, ln(As/tanβ) was calculated for the study area to represent the geomorphic processes associated with soil moisture and it’s spatial distribution in the landscape. The variables of the index (As = specific catchment area; tanβ = local slope angle) were evaluated on a cell by cell basis from the DEM, using the procedure of Hutchinson and Dowling (1992). GIS analysis of soil landscape unit coverage The land classification tables associated with the digital 1:50,000 scale SLU coverages, summarise a range of key attributes for each mapping unit. These attributes, or land qualities include: drainage, water erosion potential, scalding, salinity and recharge potential. The classification ranks each of these land qualities on a numeric scale, according to eight, generically defined class limits. Each successive class implies an increasing level of management input to overcome any limitation(s) identified by the classification from Class I (very low level of limitation) to Class VIII (extreme level of limitation). For this study, two attributes (salinity and drainage) were identified as being indicative of aquic conditions (i.e. soil landscape units prone to waterlogging). These were selected as inputs to a weighted index model with salinity given a greater weighting as it was considered an indicator of more permanent wetness in the landscape (Cox et al., 1996). 4

Results The results for each of the three derived datasets are presented as Figures 1 - 3. The datasets were co-registered and georeferenced to the same coordinate system and projection to allow for qualitative analysis. Output from the AIRSAR inversion model (Figure 1) consists of an image of soil dielectric constant, pseudo-coloured for values between 0.1 and 31.4.

Figure 1. Soil dielectric constant map.

5

Figure 2. Topographic wetness index map.

The topographic wetness index is shown in Figure 2 displayed as a classified image with values from 1 - 25, while the map of aquic soil index is shown as Figure 3.

Figure 3. Aquic soils index map showing increasing susceptibility to waterlogging.

Discussion and conclusions

6

Initial assessment of the results obtained show good qualitative agreement between the three derived datasets and with areas of salinisation and ground water discharge known from detailed studies at the sub-catchment scale. More detailed analysis of the derived datasets is currently underway in an attempt to quantify any meaningful relationships and maximise the predictive capability of these techniques for determination of soil moisture in the landscape. In isolation, each of these spatial analysis techniques can provide information on the status of soil moisture in the landscape. This study demonstrates the benefits of relating the different techniques using a GIS to enhance their predictive capabilities. Given that determination of soil moisture content and its spatial and temporal variation is of great importance for identification of areas prone to waterlogging and salinity in the landscape, there is value in developing these methods further. Acknowledgements The authors would like to thank the CSIRO Office of Space Science and Applications (COSSA) for funding of this work under the CSIRO Earth Observation Centre program. Thanks also go to the staff of the Geographic Information Services unit at PISA for provision of the soil landscape unit data. References Beven K.J. and Kirkby M.J. (1979). A physically-based variable contributing area model of basin hydrology. Hydrological Sciences Bulletin. 24, 43-69. Bruce D.A. (1996). Soil Moisture from Multi-Spectral, Multi-Polarising SAR. In Proceedings of 8th Australasian Remote Sensing Conference, Canberra, March. Cox J.W., Fritsch E. and Fitzpatrick R.W. (1996). Interpretation of soil features produced by ancient and modern processes in degraded landscapes: VII. Water duration. Australian Journal of Soil Research. 34, 803-824. Dubois P.C., van Zyl J.J. and Engman E.T. (1995). Measuring soil moisture with imaging radars. IEEE Transactions on Geoscience and Remote Sensing. 33, (4) 510-516. Engman E.T. and Chauhan N. (1995). Status of microwave soil moisture measurements with remote sensing. Remote Sensing of Environment. 51, (1) 189-198. Fitzpatrick R.W., Cox J.W. and Bourne J. (1997). Managing waterlogged and saline catchments in the Mt. Lofty Ranges, South Australia: A soil-landscape and vegetation key with on-farm management options. Catchment Management Series. CSIRO Publishing, Melbourne. Fritsch E. and Fitzpatrick R.W. (1994). Interpretation of soil features produced by ancient and modern processes in degraded landscapes: I. A new method for constructing conceptual soil-water-landscape models. Australian Journal of Soil Research. 32, 889-907. (colour figs. 880-885). Giacomelli A., Bacchiega U., Troch P.A. and Mancini M. (1995). Evaluation of surface soil moisture distribution by means of SAR remote sensing techniques and conceptual hydrological modelling. Journal of Hydrology. 166, 445-459. Hutchinson M.F. (1989). A new procedure for gridding elevation and stream line data with automatic removal of spurious pits. Journal of Hydrology. 106, 211-232.

7

Hutchinson M.F. and Dowling T.I. (1992). A continental hydrological assessment of a new grid-based digital elevation model of Australia. In: Terrain Analysis and Distributed Modelling in Hydrology. (Eds) K.J. Beven and I.D. Moore. John Wiley, Chichester. Hutchinson M.F. (1993). Development of a continent-wide DEM with applications to terrain and climate analysis. In: Environmental Modelling with GIS. (Eds) M.F. Goodchild, B.O. Parks and L.T. Steyaert. Oxford University Press, New York. Lin D.S., Wood E.F., Beven K. and Saatchi S. (1994). Soil moisture estimation over grass-covered areas using AIRSAR. International Journal of Remote Sensing. 15, (11) 2323-2343. Moore I.D., Grayson R.B. and Ladson A.R. (1992). Digital terrain modelling: A review of hydrological, geomorphological and biological applications. In: Terrain Analysis and Distributed Modelling in Hydrology. (Eds) K.J. Beven and I.D. Moore. John Wiley, Chichester. Pultz T.J., Leconte R., Brown R.J. and Brisco B. (1990). Quantitative soil moisture extraction from airborne SAR data. Canadian Journal of Remote Sensing. 16, (3) 56-62. Quinn P.F., Beven K.J. and Lamb R. (1995). The ln(a/tanβ) index: How to calculate it and how to use it within the topmodel framework. Hydrological Processes. 9, 161-182. Soil Survey Staff (1996). Keys to Soil Taxonomy, 7th edition. USDA, Washington. Ulaby F.T. and Dobson M.C. (1989). Handbook for Radar Scattering Statistics for Terrain. Artech House, Norwood. Wang J.R., Engman E.T., Shiue J.C., Rusek M. and Steinmeier C. (1986). The SIR-B observations of microwave backscatter dependence on soil moisture, surface roughness and vegetation covers. IEEE Transactions on Geoscience and Remote Sensing. GE-24, (4) 510-516. Weibel R. and Heller M. (1991). Digital terrain modelling. In: Geographical Information Systems: Principals and Applications. (Eds) D.J. Maguire, M.F. Goodchild and D.W. Rhind. Longman. Zhang W. and Montgomery D.R. (1994). Digital elevation model grid size, landscape representation and hydrologic simulations. Water Resources Research. 30, (4) 1019-1028. Keywords : waterlogging, GIS, remote sensing, AIRSAR, terrain analysis Mots clés : engorgement, SIG, télédétection, AIRSAR, étude de terrain

8