The use of remote sensing data to extract information from agricultural land with emphasis on soil salinity.
The Rio Grande Valley on the United States-Mexico border is exceptionally vulnerable to agricultural production losses due to accumulation of salts. Contribution of salts from the agricultural return flow and capillary rise from the high watertable continue to concentrate sulfate and chloride salts in soil surfaces with continuous variations in space and time. Spatially reliable techniques are needed to set base lines and verify changes in salt distributions and their controlling factors. Historically, the application of remote sensing in mapping and detecting salt-affected soils has utilised one or more of the following methods: (1) interpretation of aerial photographs (e.g. Shi and Xie 1988); (2) interpretation of colour composite films from satellite data, visual or by optical density method (e.g. Sharma and Bhargava 1988); (3) computer interpretation of digital satellite data of Landsat generations such as Landsat TM, and Multi-Spectral Scanner MSS (e.g. Sehgal et al. 1988; Rao et al. 1991; Saxena et al. 1991); and (4) computer analysis of the hyperspectral data such as Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) bands, and field spectroradiometric data.
Previous research has indicated that salts in soil and evaporates can be distinguished remotely by examining absorption features of the remote sensing spectra (Hunt and Salisbury 1971; Hunt 1982; Crowley 1991; Csillag et al. 1993; Drake 1995; Clark 1999; Howari et al. 2002). By visual interpretation of Landsat TM false colour compositions or aerial photographs, image elements such as dull white and bright white patches can be correlated with ground data to determine degree of salinity-alkalinity classes (Rao et al. 1991; Verma et al. 1994). Using surface soil samples from San Joaquin Valley, California, and the Carpathian Basin, Hungaria, Csillag et al. (1993) analysed laboratory spectroradiometer data using a modified stepwise principal component band selection method to separate 13 classes of the soil salinity status. They reported that the importance of high spectral resolution in recognition accuracy of salinity status decreased from 91, 90, and 88% as band width increased from 10, 20, and 40 nm, respectively, over both locations. They identified 35-42 narrow bands, between 500 and 2400 nm, which led to a recognition accuracy of 100% for the San Joaquin Valley. In this paper, ground, laboratory, airborne, and satellite-borne data were used to show the type and distribution of salt crusts in the Rio Grande Valley by using digital image processing and spectral extraction and matching techniques as well as field observations.
The study area
The study area is located west of El Paso, Texas, in the Rio Grande Valley, and lies between lat. 31[degrees]45' and 32[degrees]15'N and long. 106[degrees]45' and 106[degrees]30'W; further towards New Mexico obvious salt crusts close to White Sand were also considered. The climate is semi-arid with a mean annual rainfall of 200 mm. Humidity is low, and the evaporation rate is high, with the gross annual pan evaporation rate averaging 2000 mm. Cotton, chili, alfalfa, and pecan are the major crops in the Rio Grande Valley. Physico-geographically, the area is a part of an old flood-plain of the Rio Grande River. Soil development in the west of El Paso is believed to have occurred over different stages. During the first stage, the area was essentially a closed shallow playa, and the sediments from ancestral Rio Grande and adjacent areas have accumulated. The change in the course of the Rio Grande was followed by drying of the playa; this indicated the second stage of soil formation in the study area (Lovejoy and Cornell 1996). However, soils in the study area are divided into 5 units that are discussed below in this paper based on field observations and published literature such as USDA (1971).
Data and methodology
An integrated approach involving image interpretation coupled with field spectroradiometric and soil analyses was followed to detect salt-affected soils in the study area. The approach consisted of visual and spectral interpretation of multi-date, multispectral, and hyperspectral data. Ground truth verification and soil studies were accomplished during the dry season. In this study, Landsat enhanced thematic mapper (ETM+), airborne visible/infrared imaging spectrometer (AVIRIS), colour infrared film (CIR), and spectrometric data (GER 3700) were used as the main data sources. AVIRIS is a unique optical sensor that delivers calibrated images of the reflected spectral radiance in 224 contiguous spectral channels with wavelengths from 400 to 2500 nm. Landsat data (path 33 and row 38) were acquired on 12 September 1999, and AVIRIS data on 28 May 1998; these data were obtained from the Pan American Center for Earth and Environmental Studies (PACES), whereas the El Paso CIR, 1-m resolution, were obtained from the Natural Resource Conservation Service of the USDA.
Several image processing methods are available to extract information from images (e.g. Adams et al. 1986; Gupta 1991: Verma et al. 1994; Campbell 1996; Metternicht and Zink 1997). For simplicity, supervised and spectral extraction techniques were used. The classification of the spectral classes, and spectral extraction for the Landsat ETM+, AVIRIS, and aerial photographs were performed using ENVI and ERDAS image processing softwares. The images were calibrated and geo-referenced at the PACES of the University of Texas at El Paso. Details on the aforementioned image processing procedures can be found in Adams et al. (1986), Gupta (1991), Campbell (1996), and Metternicht and Zink (1997). Spectral matching of the GER 3700 data with the associated spectral library collected for this study was performed with the correlation coefficient algorithm using WinFirst spectral processing software. The spectral library and database used in this study contain reflectance spectra from NASA, USGS, Drake (1995), Clark (1999), Crowley (1991), and Howari et al. (2002).
The spectral reflectance measurements of salt crusts (n = 56) collected from the Rio Grande Valley during the summers of 1998, 1999, and 2000 were performed with a high-resolution spectroradiometer (Model GER 3700) in a wavelength range of 400-2500 nm following the procedure of Howari et al. (2002) and Howari (2002). The samples (target) were placed to completely fill the field of view (FOV) of the instrument. The optical sight of the instrument was used to align targets to the instrument's FOV. The operators wore non-reflective clothing during measurements, and measurement taken away from any reflective objects. Replicates of the measurements and samples were taken (n = 112). The average recorded standard deviation, standard error, and sample variance of the spectral reading of the same sample were 0.0084, 0.0020, and 0.00071 nm, respectively, whereas the corresponding values for the duplicates samples were 0.3187, 0.0438, and 0.1186 nm, respectively. The standard deviation for reflectivity varied, but averaged 0.179% of the means. Field samples of spectrally tested salt crusts were previously tested and identified under the microscope and analysed for physico-chemical properties at Texas A and M Agricultural Research Center, El Paso, using the standard methods reported by Hesse (1972), Ehlers (1987), and Nesse (1991).
Results and discussion
The soil in the study area can be classified according to USDA (1971) and field observations into several units with different physico-chemical properties (Fig. 1 and Table 1). The first unit (U1) consists of deep loamy soils located along the eastern foothills of the Franklins. The second unit (U2) is deep loamy sand, mostly structureless. The third unit (U3) consists of sandy loam soil containing a very hard layer of caliche; this soil unit has reddish brown sandy loam. The fourth unit (U4) is deep gravelly sandy loam or loam developed on deep gravelly or stony sediments. The valley soils (V) vary in texture, but can be best characterised as sandy loams (Vinton) to silty clay loams (Harkey).
[FIGURE 1 OMITTED]
Utilising the Landsat ETM+ (red, green, and blue--RGB--bands), and CIR (Fig. 2), as well as physco-chemical properties and field observations of soil and vegetation cover, the soils were classified in terms of vegetation and salt crusts as follows: (G1) uniform crop stand by deep red and smooth soil textures; (G2) barren landscape with weak vegetation cover; (G3a) completely barren land with salt efflorences; and (G3b) barren landscape with moderate vegetation cover. G3 in Fig. 2c indicates that crusted saline surfaces are generally smoother than non-saline surfaces and appear in images as spots of white efflorescence. Similar observations were reported by Everitt et al. (1988) in other parts of Texas. The morphology and the physico-chemical properties indicated that salinity increases from G1 to G3a and G3b (Table 1). The classification of the different spectral classes of Landsat ETM+ was able to distinguish between the different targets in the Rio Grande Valley (Fig. 2). As seen in this figure, the salt-affected land in the left of the image appears as white strips, with the light colour tone representing alfalfa and the dark colour tone representing pecan trees, the differences in the colour tone being related to shade. Pecan trees have more shade and are therefore associated with dark colour tone in the image.
[FIGURE 2 OMITTED]
However, in most of the investigated groups except G1, the lack of vegetation or scattered vegetation on salt-affected soil (SAS) surfaces makes it possible to detect salt at the surfaces, especially as in the case of Sunland Park (Fig. 3). Salt frosting of the ground is quite evident in this location (Fig. 3). Salinity of the soil often exceeds 4 dS/m in soil saturation extracts, and that of the groundwater usually exceeds 6 dS/m. Salt crusts around this area range in size from 12 to 40 [m.sup.2]. In the vicinity of this location only halophytes such as tamarix (salt cedar) and saltbush grow in a scattered pattern. As can be observed from Fig. 3, the salt-affected land is often associated with high reflectivities. Salt-affected soils tend to have a spotty type of vegetation, mostly yellowish in the RGB bands, and appear on the grey scale as light tone with smooth texture. Salinity and associated water-nutrient deficiency decrease plant pigmentation, resulting in a condition known as chlorosis or yellowing, which in theory could be used to indicate salinity status of soil covered with vegetation. However, a remaining challenge is to differentiate between yellowing resulting from salinity stress and that from nutrient deficiency such as nitrogen, which is unlikely to be present in the study area. Often in the near infrared, an inverse relationship is observed between reflectance and salinity, since salt content induces less plant cover, which appears in the form of decreased density, leaf area index, and height (Richardson et al. 1976; Everitt et al. 1977; Mougenot et al. 1993). The problem of direct identification of SAS occurs in other areas where the size of the crust is small (<1 [m.sup.2]) or where there are no evident crusts, or the area is covered with salt-tolerant vegetation. The radiometric measurements indicate that the main factors affecting the reflectance or the signals gathered by the sensor are mostly the moisture content and slope (Fig. 4) as well as roughness, quantity, and mineralogy (Howari et al. 2000b, 2000c, 2002).
[FIGURES 3-4 OMITTED]
The spectra of salt crusts collected from the Rio Grande Valley were tested against the library of collected and known spectra. Both the overall shape of the spectrum and absorption bands are important in explaining the spectral properties of soil and salt crusts. The shape of a spectrum can be characterised by the albedo, the continuum slope, and the intensity distribution. Although albedo is the primary source of variability, absorption features play a significant role in relation to specific chemical characteristics. There are 5 regions of the mean spectra of the spectrometer that exhibit distinct absorption features and high variability among the salt crusts (around 1000, 1400, 1900, 2200, and 2300 nm) (Fig. 5). Those absorption features are due to the vibration processes of salts (Hunt et al. 1971; Howari 2002; Howari et al. 2002). The spectroscopic results reveal the presence of gypsum and halite (Fig. 5).
[FIGURE 5 OMITTED]
Although the spectra extracted from the AVIRIS image of the upper Rio Grande Valley look like gypsum spectra (e.g. Fig. 6), they cannot be considered typical; the overtones, or combination tones, from the fundamental vibrations of the water molecules, produced a series of bands affecting the infrared spectrum between 1000 and 2500 nm, e.g. 1464, 1550, 1750, 1978, and 2300 nm (Howari et al. 2000a). However, not all absorption features can be detected in the spectra extracted from AVIRIS. It is likely, therefore, that the extracted spectra are mixed. Indeed, ground-truthing indicated that zones of gypsum and gypsum contaminated with quartz and clay were present. In Fig. 6 the different colours coincide with the different zones of salts or evaporate phases in the ground. About 83% of the desk survey and interpretation activities of remote sensing data that took place at the remote sensing laboratory (PACES) were confirmed in a satisfactory manner on the ground, mainly due to the fact that salt crust is evident on the surface in many locations. However, previous and ongoing studies (Howari et al. 2000b, 2002) indicated that some interlinked factors, when present, might make the identification of salt crusts a problematic process. These factors include layering or coveting of salt, grain size and/or moisture content effects, as well as washout or rainfall effects. The presented information on the remote sensing of types and occurrences of salts can help agricultural scientists and engineers to manage salinity problems affecting the ecosystem and watershed (e.g. Massoud 1990; Tanji 1996; Summer and Naidu 1998; Summer 2000).
[FIGURE 6 OMITTED]
Remote sensing has a potential application for rapid and large scale mapping of salt-affected lands. In the study area, uneven growth of plant cover and the presence of bare spots were evident, and both indicated different ground salinity status. The salt-affected soils with patchy vegetation appeared mostly yellow in combined RGB bands and appeared on the grey scale as light tone with smooth texture. Salt-affected soils were also often associated with higher reflectivity. There were 5 regions of the mean spectra that exhibited distinct absorption features and high variability in the salts. They were located around 1000, 1400, 1900, 2200, and 2300 nm, and can be used for salt identification. Additional research should focus on exploring additional remote sensing data to study the spatial variability of salt crusts of soils, and their controlling factors, and resolve end-points from mixed spectra.
Table 1. Selected physico-chemical parameters of the investigated soils Salinity (dS/m) Soil Permeability Water holding Means Max. CV unit (mm/mm of soil) capacity (mm/mm of soil) U1 15-5 2.5 3.8 11.1 127 U2 15-500 2 1.5 2.6 51 U3 15-500 1-2.5 1.8 3.3 50 U4 15-5 1.25-2.5 1.7 2.3 43 V 15-5 5-15 4.6 8.0 37
The author extends his sincere thanks and appreciation to the Pan American Center for Earth and Environmental Studies at UTEP for technical assistance, and Texas Water Resource Institute for funding part of this research.
Adams JB, Smith MO, Johnson PE (1986) Spectral mixtures modeling: new analyses of rock and soil types at the Viking sites. Journal of Geophysical Research 91, 8098-8112.
Campbell A (1996) 'Introduction to remote sensing.' (The Guilford Press: New York)
Clark RN (1999) Spectroscopy of rocks and minerals and principles of spectroscopy. In 'Remote sensing for earth sciences: manual of remote sensing'. 3rd edn (Ed. AN Rences) Vol. 3, pp. 3-52. (John Wiley and Sons Inc.: New York)
Crowley JK (1991) Visible and near-infrared (0.4-2.5 micron) reflectance spectra of playa evaporate minerals. Journal of Geophysical Research 96, 16231-16240.
Csillag F, Pasztore L, Biehl LL (1993) Spectral selection for characterization of salinity status of soils. Remote Sensing of Environment 43, 231-242. doi: 10.1016/0034-4257(93)90068-9.
Drake NA (1995) Reflectance spectra of evaporite minerals (400-2500 nm): applications of remote sensing. International Journal of Remote Sensing 16, 2555-2571.
Ehlers EG (1987) 'Optical mineralogy:.' Vol. 1, Theory and techniques, Vol. 2 Mineral description. (Blackwell Scientific Publication: Palo Alto, CA)
Everitt JH, Gerbermann AH, Cuellar JA (1977) Distinguishing saline from non-saline rangelands with Skylab imagery. Programmatic Engineering and Remote Sensing 43, 1041-1047.
Everitt JH, Escobar DE, Gerbermann AH, Alaniz MA (1988) Detecting saline soils with video imagery. Photogrammatic Engineering and Remote Sensing 54, 1283-1287.
Gupta RP (1991) 'Remote sensing geology.' (Springer-Verlag: Berlin)
Hess PR (1972) 'Textbook of soil chemical analysis.' (Chemical Publisher: New York)
Howari FM (2002) Spectroscopy of evaporates. Periodico di Mineralogia, an International Journal of Minerology. Crystallography and Geochemistty 71, 192-200.
Howari FM, Goodell PC, Miyamoto S (2002) Spectral properties of salt crusts formed on saline soils. Journal of Environmental Quality 31, 1453-1461.
Howari FM, Goodell PC, Miyamoto S, Penn B (2000a) Spectral properties of treated silty clay loam and fine sandy loam with gypsum, halite and their mixtures: Remote Sensing 2000. In 'Laboratory spectroscopy to remotely sensed spectral observation'. pp. 2-27. (Soil Science Society of America, Corpus Christi, Texas; Blackland Research and Extension Center: TX)
Howari FM, Goodell PC, Miyamoto S, Penn B (2000b) Spectral properties of gypsum, halite and their mixtures. In '14th International Conference and Workshop on Applied Geologic Remote Sensing'. Las Vegas, Nevada. (Abstract) http://notesmi.altarum.org/Conf/GeologyPC.nsf/14_TechProg?Open FrameSet
Howari FM, Goodell PC, Miyamoto S, Penn B (2000c) Spectral reflectance characteristics of salt deposited on the soil surfaces. In '2nd EARSel Workshop on Imaging Spectroscopy, ITC, The Netherlands'. Proceedings on CD. Vol. 40, pp. 1-11. (ITC: Enshide, The Netherlands)
Hunt GR (1982) Spectroscopic properties of rocks and minerals. In 'Handbook of physical properties of rocks'. Vol 1 (Ed. RS Carmichael) pp. 295-385. (CRC Press: Boca Raton, FL)
Hunt GR, Salisbury JW (1971) Visible and near-infrared spectra of minerals and rocks. I. Silicates. Modern Geology 2, 23 30.
Hunt GR, Salisbury JW, Lenhoff CJ (1971) Visible and near-infrared spectra of minerals and rocks, IV. Sulphides and sulphates. Modern Geology 3, 1-4.
Lovejoy EP, Cornell WC (1996) 'El Paso's geological past.' (Texas Western Press: The University of Texas, El Paso, TX)
Massoud II (1990) 'Salinity and alkalinity as soil degradation hazards.' (FAO: Rome)
Metternicht G, Zink JA (1997) Spatial discrimination of salt and sodium affected soil surfaces. International Journal of Remote Sensing 18, 2571-2586. doi: 10.1080/014311697217486.
Mougenot B, Epema GF, Pouget M (1993) Remote sensing of salt affected soils. Remote Sensing Review 7, 241-259.
Nesse WD (1991) 'Introduction to optical mineralogy.' 2nd edn (Oxford University Press: New York)
Rao BR, Dwivedi RS, Venkataratnam L, Ravishankar T, Thammappa SS (1991) Mapping the magnitude of salinity in part of the Indo-Gangetic plains Uttar Pradesh, northern India using Landsat TM data. International Journal of Remote Sensing 12, 419-25.
Richardson AJ, Gerbermann AH, Gausman HW, Cueller JA (1976) Detection of saline soils with Skylab multispectral scanner data. Photogrammetric Engineering and Remote Sensing 42, 679-684.
Saxena RK, Verma KS, Barthwal AK (1991) Assessment of land degradation hazards in Etah district, Uttar Pradesh, using Landsat data. Photonirvachak, Journal of Indian Society of Remote Sensing 19, 83-84.
Sehgal JL, Saxena RS, Verma KS (1988) Soil resources inventory of India using image interpretation technique. In 'Remote sensing is a tool for scientists. Proceedings of 5th Symposium of the Working Group for Remote Sensing'. Budapest, Hungary. pp. 17-31.
Sharma RC, Bhargava GP (1988) Landsat imagery for mapping saline soils and wetlands in north-west India. International Journal of Remote Sensing 9, 39-44.
Shi Yuanchun C, Xie AG (1988) Remote sensing for surveying and mapping of salt-affected soils. In 'Proceedings of the International Symposium on Solonetz Soils. Problems, Properties, Utilization. Osijek, Yugoslavia'. (Ed. M Adam) pp. 104-110. (Agricultural Faculty, University of Osijek: Yugoslavia)
USDA (1971) 'Soil survey, El Paso County, Texas.' (Soil Conservation Service USDA, and Texas Agricultural Experiment Station: TX)
Summer EM, Naidu R (1998) 'Sodic soils, distribution, properties, management, and environmental consequences.' (Oxford University Press: Oxford, OH)
Summer ME (2000) 'Handbook of soil science.' (CRC Press: Boca Raton, FL)
Tanji KK (1996) 'Agricultural salinity assessment and management.' (ASCE: New York)
Verma K, Saxena R, Barthwal A, Deshmukh R (1994) Remote sensing technique for mapping salt affected soils, International Journal of Remote Sensing 15, 1901-1914.
Manuscript received 11 November 2002, accepted 6 June 2003
Faculty of Science, Geology Department, UAE University, PO Box 17551, Al-Ain, UAE; Present address: Southwest Applied Earth and Environmental Services, POB 155144, Irving, TX 75015, USA.
|Printer friendly Cite/link Email Feedback|
|Publication:||Australian Journal of Soil Research|
|Date:||Dec 1, 2003|
|Previous Article:||Effect of soil amendment with bauxite Bayer process residue (red mud) on the availability of phosphorus in very sandy soils.|
|Next Article:||The hydrology of Vertosols used for cotton production: I. Hydraulic, structural and fundamental soil properties.|