When I started this blog, I attended to summarize everything into one story. However, I realized that I might have to split this into two parts. So, this section will be part 1 where I will explain the next steps of my radar research and the data processing that I have been up to these past couple of weeks. Part 2 will discuss more about the radar images I have produced and my thoughts about how to interpret the quantified radar data.
For the past few weeks, I have been preparing for my annual PhD meeting where I sit down with my research committee and review the work I have completed since my comprehensive exam in November 2018. I had to submit a report reintroducing the aim(s) and objectives of my PhD thesis, list my first author, co-author and conference publications, and summarize completed and planned work for the coming academic year. In addition to the report, I had to create a presentation for my committee and present it on the day of my meeting. During my time writing the report, I started to think about new methods to interpret radar data to better discuss the backscatter mechanisms of lava flows. As most of you know, I use synthetic aperture radar (SAR) datasets to study the surface roughness and morphology of lava flows on Earth to understand the emplacement of lava flows on the Moon and Mars. To get more context on my research, I would recommend reading my past blogs ( (1) Craters of the Moon Geological and RADAR Maps, (2) Volcanic Analogues, Surface Morphology and Roughness of Lava Flows: An Open Discussion, and (3) Holuhraun: Wherefore Art Thou Lava Flow). I thought about how to better support my interpretations of radar data, understanding the types of backscatter mechanisms occurring with lava flow surfaces and trying to identify which mechanisms are dominant over a particular lava flow. I looked over my literature review reading list and re-read a paper by Campbell (2012) titled “High circular polarization ratios in radar scattering from geologic targets”. The paper discusses the types of backscatter mechanisms that are responsible for certain circular polarization ratio (CPR) values (ratio between the power of transmitted same-sense polarization signals and signals with an opposite-sense polarization) and how to interpret results extracted from radar polarization datasets. Re-reading this paper made me realize that I should look deeper into the radar data I am using for my research to better understand the results and to improve my interpretations. From that point, I got to work…
After my annual PhD meeting on the 15th of October, I began looking over my radar data. I am focusing on data collected from the NASA/JPL UAVSAR-L (24-cm wavelength) instrument that scanned the surface of the 2014-15 Holuhraun flood basalt in Iceland in May 2015. The lava flows at Holuhraun exhibit diverse surface roughness and morphology making it a perfect location to study radar interpretation of lava flows, and in addition it has analogous surface features to Martian lava flows. The data is available on the UAVSAR JPL website (https://uavsar.jpl.nasa.gov/cgi-bin/data.pl. You can download .grd files (gridded data) on all of the radar polarization data products (e.g. HHHH, VVVV, HHHV, etc). Campbell (2002) and Neish and Carter (2014) provide an excellent overview of radar polarization states and scattering matrixes. All of the datasets provide us with different information about the scattering mechanisms and the physical properties of the lava flows surfaces. I wanted to look at the data more closely. I began by converting the data to files that are readable on ArcGIS. Using an Interactive Data Language script written by Dr Catherine Neish, I converted the .grd files to .dat files (a type of text file). The .dat files are not readable on ArcGIS, so I needed to convert them to an image type file. For my work, it is a USGS ISIS3 cube file. Below is an example of the command line I used in USGS ISIS3 to convert the .dat files to .cub files.
ascii2isis from=PolarzationDataProduct_HHHH.dat to=PolarizationDataProduct_HHHH.cub samples=5881 lines=29724
(Note, the sample and line values are found in the UAVSAR-L metadata)
After each radar polarization dataset was converted to a cube file I ran them through a reduce command function to decrease the speckle noise errors in the data. Speckle noise describes the variability between shifting constructive and destructive interference patterns among the numerous surfaces within each radar resolution cell (Campbell 2002). Since each surface produces a reflected signal with its own amplitude and phase, each cell will comprise a collection of different radar echo data. This over exaggerates the echo from each resolution cell increasing the error. The reduce function in ISIS3 reduces a cube data to a new size by scaling the lines and samples up or down. The pixel values can either be averaged or the closest pixel to the window center will be selected to accommodate the new assigned cube data proportions. For more information about how the reduce function works see the following site, https://isis.astrogeology.usgs.gov/Application/presentation/Tabbed/reduce/reduce.html.
After the reduce command was completed, I uploaded the radar polarization products (both the original and reduced) onto ArcGIS 10.7.1. I wanted to compare the original and reduced cube files to see any differences in the results. Now I know a couple of people will ask why I did not use the qview function on ISIS3 to view the cube files to see if the commands worked correctly. I did before uploading the images onto ArcGIS to make sure the images were converted correctly. I wanted to check again on ArcGIS since I will be using this program for most of my data analysis. I noticed that the reduced datasets had less bright pixels than the original datasets. This was a good sign. However, I would not know if it worked until I extracted quantitative results from the datasets. Below are examples of the radar polarization datasets.
Both images are HHHH (horizontal-like transmitted and horizontal-like received polarization) images of the 2014-15 Holuhraun flood basalt. The first image in the slideshow is the polarization product without having the reduced function applied. The second image has had the reduced function applied, scaling down the samples and lines by a factor of 2. The reduced image has fewer white pixels than the original image.
With the radar polarization datasets successfully converted to cube files and uploaded to ArcGIS I was able to begin producing a CPR image of Holuhraun. The values from CPR images are more appropriate to use when studying the surface roughness of lava flows. To produce CPR images, I required the same-sense transmitted polarization signals (SC) and opposite-sense polarization (OC) (CPR = SC/OC) data. To get the SC and OC I calculated the Stoke Matrix’s W11, W14 and W44. The equations below show the steps required to obtain the Stoke Matrix’s and then OC, SC and CPR.
W11 = 0.25 * (HHHH + VVVV + (2 * HVHV))
W14 = -0.5 * (HHHVI + HVVVI)
W44 = 0.5 * (HVHV – HHVVR)
Note: The I and R denotes the imaginary and real component of the Stoke polarization state
OC = W11 – W44
SC = W11 + (2 * W14) + W44
CPR = SC / OC)
These calculations were carried out using the Raster Calculator in ArcGIS (http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/raster-calculator.htm) to produce raster datasets of each product. The final CPR images can be observed below (includes a CPR image created from polarization products that had not been reduced and a products that had been reduced).
To keep this post sweet and short I am going to end it here and talk more about the radar polarization and Stoke Matrix datasets in part 2. Part 2 will be posted at the end of the month and it will include a plethora of radar images :D
See you all next time!
References
Campbell, B. A. (2002). Radar remote sensing of planetary surfaces. Cambridge University Press.
Campbell, B. A. (2012). High circular polarization ratios in radar scattering from geologic targets. Journal of Geophysical Research: Planets, 117(E6).
Neish, C. D., & Carter, L. M. (2014). Planetary radar. In Encyclopedia of the Solar System (pp. 1133-1159). Elsevier.