Awards & Nominations
Castillo has received the following awards and nominations. Way to go!
Global Finalists Honorable Mentions
Crystal Clear: Working with Tropospheric Noise in InSAR Images
High-Level Project Summary
In this project, the team has developed a solution to reduce the impact of tropospheric noise in InSAR images. The reduction of noise has been achieved by using two techniques: 1) Non-linear In-Phase Power Law Transformation, and 2) Removing the phase delay caused by the total precipitable water vapour. Combining these two methods introduces the unique advantages of both techniques, which help in appropriate noise reduction. Noise can often inhibit accurate analysis of ground deformation, leading to further misinterpretations and incorrect forecast models. Therefore, noise reduction is an extremely important requirement to ensure accuracy and precision in data analysis and modelling.
Link to Final Project
Link to Project "Demo"
Detailed Project Description
Tools/Coding Languages/Software
- Python - Jupyter Notebook
- Main Libraries: Netcdf4, Matplotlib, OpenCV, Numpy
InSar data refers to Interferometric Synthetic Aperature Radar imagery, which shows the change in the deformation of the Earth’s surface over a period of time. InSar is usually created by taking two or more SAR images and calculating the phase change. This phase change is plotted on a map to show the dynamic properties of the area under observation.
InSar can be used for a wide variety of applications, such as sea ice monitoring, analyzing the ground before and after a volcanic eruption, and observing the changes in fault lines and areas near techtonic plates. Since SAR imagery is acquired using various radar satellites, the signals that are sent and received pass through the Earth’s atmosphere. Often times, due to the water vapour in the troposphere, noise is introduced into the signal and hence the data received. This noise can mask the true changes on the Earth’s surface and can alter our interpretations of the data. For this reason, it is important to be able to identify and/or remove this noise from our acquired images.
There are currently several methods out there to remove tropospheric noise from InSar images. Some of these include the following: * Averaging * Adaptive filtering * Smoothing
These methods are only partially effective, and can also sometimes effect the true ground signal by over-smoothing it, or masking it further. Therefore, we need to consider techniques that target the noise signal specifically.
I will be combining two techniques: 1. Correction using percipitable water vapour, and 2. Non-linear In-Phase Transformation
Both of these techniques provide unique advantages:
Correction using percipitable water vapour: Precipitable water vapor (PWV) is defined as the column height of liquid water equivalent to the total water vapor present between Earth’s surface and the top of the atmosphere. (Bevis et al., 1992). PWV can be related to the wet component of SAR phase delay, as follows. In order to remove this from our original InSar Image, we can subtract this wet-tropo phase delay from the phase of the original InSar image. Here Lambda is the wavelength at which the images were acquired, theta is the incidence angle of the satellite, and capital Pi is ~6.2. This method is effective in estimating individual tropospheric properties, and using atmospheric observations instead of broader models.
Non-linear Phase Transformation This is an empirical phase-based method to separate the tropospheric noise from the true ground signal, using the data acquired itself. There are no external sources of data used in this method, and the transformation required is measured from within the data. In this notebook, the non-linear Power Law is applied across the InSar image in question, in an attempt to reduce the effect of the tropospheric noise.
The data supplied by the Space Apps challenge was used. Three InSar images - Kilauea Volcano in Hawaii, Ridgecrest California, and Faultcreep (also in California) were used for this challenge.
Space Agency Data
Space Agency Data used:
- NASA/ESA Data for InSAR images
- MODIS data for precipitable water vapor
This data inspired our project with the breadth of its extent. While it was challenging to learn how to download this data, its usefulness was wide spread. There is still so much of the data that has not been explored in this project, and hopefully we continue to improve this project with the help of incorporating that additional data.
Hackathon Journey
The hackathon journey was challenging. Being the only member in the group resulted in a large bulk of work, including research, data download and transformation, coding, and presentation. Furthermore, this topic was quite tedious in application. While the concepts behind this topic, as well as the researched solutions, were fairly straightforward to understand, in application it was difficult to download the right data and work with the geospatial file formats.
However, despite the challenges, this was an amazing learning experience. My career goal is to enter this field - doing geospatial analysis with SAR imagery for environmental applications. Therefore, I was really able to learn a lot about the field, the current technologies and novel ideas, as well as about my own strengths and weaknesses.
References
- Python - Jupyter Notebook
- Main Libraries: Netcdf4, Matplotlib, OpenCV, Numpy
- Data:
- NASA/ESA Data for InSAR images
- MODIS data for precipitable water vapor
- Paper on noise reduction techniques: Murray, et.al.
- https://www.sciencedirect.com/science/article/abs/pii/S0034425719303451?via%3Dihub