Detection of Surface Temperature Anomalies in the Coso Geothermal Field Using Thermal Infrared Remote Sensing

Transcription

1 GRC Transactions, Vol. 31, 2007 Detection of Surface Temperature Anomalies in the Coso Geothermal Field Using Thermal Infrared Remote Sensing Mariana Eneva 1, Mark Coolbaugh 2, Steven Bjornstad 3, and Jim Combs 4 1 Imageair, Inc., San Diego, CA, meneva@imageair-inc.com) 2 Great Basin Center for Geothermal Energy, University of Nevada Reno, mfc@unr.nevada.edu) 3 Geothermal Program Office, Naval Air Weapons Station, China Lake, CA, steven.bjornstad@navy.mil 4 Geo Hills Associates, Reno, NV, geohills@mac.com) Keywords Geothermal, remote sensing, thermal infrared, TIR, satellite, Coso ABSTRACT We use thermal infrared (TIR) data from the spaceborne ASTER instrument to detect surface temperature anomalies in the Coso geothermal field in eastern California. The identification of such anomalies in a known geothermal area serves as an incentive to apply similar markers and techniques to areas of unknown geothermal potential. We carried out field measurements concurrently with the collection of ASTER images. The field data included reflectance, subsurface and surface temperatures, and radiosonde atmospheric profiles. We apply techniques specifically targeted to correct for thermal artifacts caused by topography, albedo, and thermal inertia. This approach has the potential to reduce data noise and to reveal thermal anomalies which are not distinguishable in the uncorrected imagery. The combination of remote sensing and field data can be used to evaluate the performance of TIR remote sensing as a cost-effective geothermal exploration tool. Background Remote sensing can be used as a cost-effective tool to explore large areas for geothermal potential. Useful applications include mineral mapping (e.g., Kratt et al., 2006), geobotanical markers (e.g., Pickles et al., 2001), and identification of thermal anomalies (Coolbaugh, 2003: Coolbaugh et al., 2007; Calvin et al., 2002) associated with geothermal activity. Here we report results from the search for thermal anomalies using images collected by the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) mounted on the Terra satellite. This instrument has been collecting data since early It is important to note that the thermal infrared (TIR) imagery contains a number of artifacts that are not connected to geothermal activity. Among several factors, topographic slope, albedo, and thermal inertia contribute significantly to the observation of numerous non-geothermal anomalies and considerable data noise. High albedo, i.e. higher reflectance, is associated with less energy remaining for heating, southern slopes receive more of the flux of solar irradiance than the northern slopes, and thermal inertia is related to how fast a material gets heated and cools off. The surface temperature expression of these three factors has no relevance to possible genuine geothermal anomalies. The effect of albedo and topographic slope can be reduced through modeling of the heat energy using visible-near infrared (VNIR) data and digital elevation models (DEM). The effect of thermal inertia can be neutralized by combining TIR satellite images collected as daytime/nighttime pairs. Corrections for these three factors have been shown to increase the number of remotely sensed thermal anomalies by an order of magnitude compared with the TIR images without such corrections (Coolbaugh et al., 2007). We focus on the Coso Geothermal Power Project (Monastero, 2002) in the central part of eastern California, with the intent to extend this work north-northwest, towards the Mammoth Geothermal Power Project. The region between these existing geothermal fields is suspected to have significant geothermal resources. We use mainly thermal infrared (TIR) ASTER data (wavelengths 8 to 12 µm). However, a visible-near infrared (VNIR) ASTER product (0.52 to 0.86 µm) also assisted us in the correction for artifacts in the TIR data. Close to the type of application featured here, Calvin et al. (2002) used a daytime/nighttime pair of ASTER TIR scenes over the Brady Hot Springs geothermal area (Nevada) to identify a thermal anomaly associated with a nearby fault. Comprehensive corrections for topographic slope, albedo, and thermal inertia have been first carried out by Coolbaugh (2003). Eneva et al. (2006) used these techniques to analyze a nighttime/daytime pair of ASTER TIR scenes collected over Coso in August This work confirmed that such corrections reduce noise and reveal thermal anomalies that are not seen in the uncorrected TIR imagery. Since we used archived data, no concurrently collected field data existed in August 335

2 Figure 1. ASTER coverage of the Coso KGRA (white polygon) extracted from two Level 1B VNIR images collected at 11:45 am (daylight savings time) on August 22, 2006 (RGB: R=band 3, G=band 2, B=band 1). Production area is shown with a red outline. Yellow arrows point to locations of temperature probes along three lines in the Coso Hot Springs (CHS), Devil s Kitchen (DK), and Wheeler (WH) areas. Green arrows show locations of reflectance measurements. Magenta arrow points to the location where an infrared thermometer was installed However, Coolbaugh et al. (2007) showed these to be very important in the interpretation of remote sensing TIR imagery. We organized two field experiments in July and August 2006 concurrent with the collection of new daytime/nighttime ASTER pairs over Coso. We collected subsurface and surface temperature data, reflection data for albedo estimates, and atmospheric profiles to perform atmospheric corrections of the imagery. The contribution of the July 2006 field experiment was significantly reduced due to clouds uncharacteristic for that time of the year in this area. The July 2006 nighttime ASTER scene turned out to be useless, while the quality of the daytime scene was marginal at best. This was the reason to organize the second filed experiment in August 2006 that was significantly scaled-down compared with the first field trip. At that time a good quality daytime/nighttime pair of ASTER scenes was collected concurrently with surface temperature and reflection measurements. Figure 1 shows part of the August 2006 daytime scene marking the locations of the various field measurements during both field trips. Although weather and logistics reduced the usefulness of the field data we collected, they are presented here as learning experience and in order to demonstrate what is possible in future work. 336 Methodology The methodology used here has been described in detail by Coolbaugh (2003) and Coolbaugh et al. (2007). A brief summary is provided for the sake of clarity here, but the reader is referred to the above publications for details. A simplified heat energy model based on net surface radiation flux is used to correct for albedo and topographic slope. This flux is approximated by (1-A)*M(Z)*cosZ, where A is the ground albedo, cosz is a measure of the topographic slope (calculated as so-called shaded relief ), Z is the sun zenith angle, and M(Z) is the atmospheric transmission depending on Z. The albedo A is obtained as a weighted average from three band-dependent albedo estimates A w =(R w b w )/(k w *cosz ), where the subscript w indicates any of the three ASTER VNIR bands, R w is taken from the AST_07 product (surface reflectance), and the constants k w and b w are estimated from field reflectance measurements as described by Coolbaugh (2003) and Coolbaugh et al. (2007a). These constants would have been 1 and 0, respectively, if it were possible to correct the AST_07 scene perfectly for atmospheric absorption and scattering effects. The heat flux is further integrated over time to model changes in the intensity of light and the position of the sun relative to the topographic slopes over the course of a day, such that E ~ (1-A)*Σ t [M(Z) t *cosz t*d t ]* t t, where E is the solar energy absorbed per unit area; t t is the time interval for each component of the sum; and D t is a time decay factor ranging from 0 to 1, which is inversely proportional to the time gap between a given position of the sun and the time the imagery was acquired. This simplified model accounts for heat dissipation and is used to calculate pseudo-temperature images that are subtracted from the AST_08 (surface temperature) daytime and nighttime images. Both the daytime and the nighttime scenes are affected by differential heating during the day, even at predawn times. This is particularly important in our case, because the Terra satellite passes only several hours after sunset. The final step in the technique is to neutralize the effect of thermal inertia by summing up the corrected daytime and nighttime AST_08 images, using weighing coefficients minimizing the variance of the sum (Coolbaugh et al., 2007). Data and Analysis Satellite Thermal Infrared (TIR) Data Compared with previous TIR remote sensing, ASTER is a unique instrument providing multispectral images in 14 different bands, of which three visible and near-infrared (VNIR) channels (wavelengths 0.5 to 0.9 µm) at 15-m spatial resolution and five thermal infrared (TIR) channels (8 to 12 µm) at 90-m

3 a Band 1 Band 2 Band b Relative Reflectance AST_ ast07_b1 ast07_b2 ast07_b Wavelength (nm) A*cosZ' Figure 2. Reflectance measured at various locations with a hand-held ASD FieldSpec spectroradiometer. Wavebands 1 to 3 correspond to ASTER VNIR. (a) Individual reflectance curves at various locations. Dark to light surfaces are represented by increasing reflectance. (b) ASTER reflectance versus product of shaded relief (related to topographic slope) and albedo measured in the field. Fitted straight lines are used to estimate coefficients described in text. resolution. The most frequently used ASTER data product is Level 1B (radiance at sensor). Higher-level data products calculated from Level 1B include surface kinetic temperature (AST_08) and surface reflectance corrected for atmospheric effects (AST_07), both used in this work. The ASTER scenes are of size ~60-km X 60-km. The satellite passes every 16 days over a given site, although scenes are not necessarily collected with each passage. However, special scheduling makes it possible to collect a daytime/nighttime pair separated by a short time interval. The shortest possible time interval in California is ~36 hours. The nighttime image in the second of the two pairs (i.e., the only useful one) was collected at 11 p.m. on August 20, and the daytime imagery was collected at 11:45 a.m. on August 22 (daylight savings times). Figure 1 shows part of the daytime ASTER Level 1B VNIR image, covering the Coso known geothermal resources area (KGRA) and the current production area. Distinct hydrothermal features are marked with yellow arrows - the Coso Hot Springs (CHS), Devil s Kitchen (DK), and Wheeler (WH) areas. The ASTER scenes were orthorectified using a USGS 30-m digital elevation model (DEM). Orthorectification is important in this case because the surface elevations in the Coso area vary rather significantly, from 720 m to 1550 m in the area depicted in Figure 1. Field Data In order to correct the AST_08 images for albedo, we collected reflectance data at 15 locations over surfaces of different color and texture, around the time of daytime satellite passage in August For this purpose we used a hand-held ASD FieldSpec spectroradiometer to measure the reflectance of contrasting surfaces such as bare soil, volcanic rock, red cinder, gravel, dirt road, and dry and live sage bush. The locations of reflectance measurements are marked in Figure 1 with green arrows. These field data were used to evaluate the k w and b w coefficients (Figure 2) in the relationship between the spaceborne reflectance data (AST_07) and the product of albedo and topographic slopes - see section Methodology above, and for details Coolbaugh et al., We estimated k w to be between 0.50 and 0.55 and b w between 0.11 and 0.16 for the three bands. Results Figure 3 shows the final product of the analysis that is a corrected for albedo, topographic slope and thermal inertia. This is a combined scene from the daytime and nighttime August 2006 AST_08 orthorectified corrected images. The weighing factor that produced minimum variance was found to be for the nighttime image (and consequently, for the daytime image). That is, the combined scene in Figure 3, overleaf, contains about 2/3 of the nighttime and 1/3 of the daytime scenes in order to neutralize the effect of thermal inertia. This is a work in progress and we are still working on optimizing this application. Although Figure 3 shows clear elevated temperatures at the three known hot areas at Coso (marked with CHS, WH, and DK in Figures 1 and 3), there are broad areas of elevated temperatures that are not associated with geothermal activity. In future work we will also strive to remove the effect of elevation (generally cooler temperatures at higher elevations), microclimate (e.g., temperature inversions producing warmer surface temperatures in valleys), and vegetation. Additional Field Data Additional field data were collected, especially in July 2006, but because of the poor quality of the concurrently collected ASTER images at that time, they could not be used in this analysis. However, they depict interesting aspects of the geothermal field and are therefore briefly discussed here. Temperature loggers were used in July 2006 to record subsurface temperatures in 13 locations grouped along three 337

4 Figure 3. Corrected AST_08 (surface temperature). White and red outlines mark KGRA and production areas, respectively. Blue to red indicate increasing temperatures. The CHS and WH areas are clearly captured as anomalies with elevated temperatures (yellow arrows), and the DK area is also suggested. Broad areas of higher temperatures are associated with topographic planes and valleys with likely nighttime temperature inversions not accounted for at this stage of modeling. Figure 4. Subsurface temperatures measured at 13 locations (blue circles in map insets) across hydrothermal features, as indicated. Distance is measured from the first probe along each line, starting within features and moving outwards. Photo insets show a probe and the surroundings of the CHS-2 probe. Map insets, with legend in middle top, courtesy of Bethiah Hall, USN GPO (China Lake, CA). lines across distinct hydrothermal features in the Coso Hot Springs (CHS), Devil s Kitchen (DK), and Wheeler (WH) areas. The locations of the three lines are marked with yellow arrows in Figure 1 and additional details are shown in Figure 4. The temperature probes were installed at depths 0.70 to 0.90 m, as rock permitted, 50 m apart for each line. The recordings lasted about a week during which subsurface temperatures remained constant for any given site. As could be expected, in each of the three groups, the temperatures decrease in the direction from the center of a hydrothermal feature towards its periphery. In July 2006 we also measured surface temperature with a hand-held infrared camera FLIR P65. This was mostly done at distances 2 to 3 m from the locations where the temperature probes were installed, as well as from a distance towards the surrounding rhyolite domes and hills. Although we could not compare the FLIR images with the cloudy ASTER data from that time, they are rather unique and it is instructional to examine the small-scale spatial distribution of surface temperature (Figure 5). An Omega OS534E infrared thermometer was installed in the area of the Coso Wash in August 2006, pointed at bare soil and logging data over a two day period. Figure 1 shows 338

5 its location (pink arrow) and Figure 6 shows the recorded surface temperatures, as well as images of the thermometer installation. The AST_08 and AST_07 products incorporate a standard atmospheric correction. However, better correction is achieved if specific profiles of atmospheric temperature, water vapor, and pressure are available. We collected such data by launching radiosondes about half an hour before the two satellite passages in July This was done with the assistance of the ASTER team at the Jet Propulsion Laboratory (JPL). Due to expense and logistic difficulties, radiosondes were not re-launched in August 2006, the only time when we got clear ASTER images. Figure 5. Surface temperature measured with a hand-held FLIR-P65 infrared camera close to the times of satellite passages. Left nighttime images; right daytime images. Top vicinity of the CHS-1 probe (probe and demarcation pole are cooler). Note hotter surface due to ground disturbance around probe. Bottom mud pots near the CHS-2 probe. Figure 6. Surface temperature measurements with an infrared thermometer over two days in August Grey diamonds indicate temperatures at the times of satellite passage. Insets show thermometer installation, with a portable power supply and a fence and ribbons to deter burros. Conclusions ASTER TIR remote sensing data and various field measurements were recently collected in and over the Coso Geothermal Project in eastern California. A model is being applied to the TIR imagery that takes into account albedo, topographic and thermal inertia effects. Although simplified, this model is capable of eliminating false thermal anomalies and revealing thermal signals not seen in the uncorrected imagery. These techniques are thus promising as an exploration tool and we will be further applying them to search for thermal anomalies outside known geothermal areas. Due to unfavorable weather conditions, the ASTER images from July 2006 could not be used, and some of the field data we collected could not be utilized for comparison purposes as initially intended. However, the field data depict complementary aspects of the Coso geothermal field and are instructional in view of future applications. 339

6 Acknowledgments USN Geothermal Program Office (GPO) - Frank Monastero made it possible to collect field data at Coso, and Dave Meade and Chris Page helped in the process. Jet Propulsion Laboratory (JPL) - Leon Maldonado helped to make arrangements for the collection of the ASTER nighttime/daytime pairs, while Mike Abrams, Elsa Abbott and Howard Tan were instrumental in the radiosonde launches. San Diego State University (SDSU) - Doug Stow and Pete Coulter provided the ASD FieldSpec spectroradiometer for the reflection measurements. This project is funded by the California Energy Commission (CEC) and is partially matched by a NASA grant. References Coolbaugh, M.F. (2003). The Prediction and Detection of Geothermal Systems at Regional and Local Scales in Nevada using a Geographic Information System, Spatial Statistics, and Thermal Infrared Imagery. Ph.D. Thesis, University of Nevada Reno. Coolbaugh, M.F., C. Kratt, A. Fallacaro, W.M. Calvin, and J.V. Taranik (2007). Detection of geothermal anomalies using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) thermal infrared images at Brady s Hot Springs, Nevada, USA, Remote Sensing of Environment, 106, Eneva, M., M.F. Coolbaugh, and J. Combs (2006). Application of satellite thermal infrared imagery to geothermal exploration in east central California, Geothermal Resources Council Transactions, 30, Kratt, C., M.F. Coolbaugh, and W.M. Calvin (2006). Remote detection of Quaternary borate deposits with ASTER satellite imagery as a geothermal exploration tool, Geothermal Resources Council Transactions, 30, Monastero, F.C. (2002). Model for success: An overview of industrymilitary cooperation in the development of power operations at the Coso Geothermal Field in southern California. Geothermal Resources Council Bulletin, Pickles, W.L., P.W. Kasameyer, B.A. Martini, D.C. Potts, and E.A. Silver (2001). Geobotanical remote sensing for geothermal exploration. Geothermal Resources Council Transactions, 25,