on December 30, 2022

How to destripe the thermal band in SLC_off LandSat 7 ETM+ data?

Hiking & Activities

Taming the Stripes: Getting the Most Out of Landsat 7’s Thermal Band After the SLC Failure

Landsat 7. It’s been orbiting our planet, diligently collecting data, since 1999. A real workhorse, right? But then, back in May 2003, things got a little…stripy. The Scan Line Corrector (SLC), the gizmo responsible for keeping the images nice and clean, well, it gave up the ghost i. Suddenly, about 22% of each image was missing, leaving these annoying wedge-shaped gaps i. If you’ve ever worked with Landsat 7 data, especially the thermal band (Band 6), you know exactly what I’m talking about.

Now, why is this a big deal? The thermal band gives us crucial information about land surface temperature, vital for everything from tracking urban heat islands to monitoring volcanic activity. But here’s the kicker: Band 6 has a coarser resolution (60m) than the visible bands (30m) i. So, those stripes? They really mess with the data, making accurate analysis a serious challenge.

Think of it like trying to paint a picture with a brush that keeps skipping parts of the canvas. Frustrating, to say the least.

So, how do we fix this? How do we “destripe” the thermal band and get back to extracting meaningful information? Let’s dive in.

First, a Little Background: Why the Stripes?

The SLC was designed to compensate for the satellite’s movement, ensuring a continuous, seamless image i. When it failed, the ETM+ started scanning in a zig-zag pattern, creating those data gaps and overlaps i. The USGS, bless their hearts, removes the duplicated data, which is why we end up with those distinctive stripes in the first place i.

Before You Destripe: Getting Your Data Ready

Before you jump into destriping, it’s essential to prep your data. Think of it as laying the groundwork before building a house. This usually involves a few key steps:

  • Radiometric Calibration: This is basically converting the raw data (digital numbers) into actual radiance values i. The USGS provides the necessary calibration parameters. Pro-tip: pay attention to the thermal band calibration updates; they’ve tweaked it a few times to correct for bias errors i.
  • Atmospheric Correction: The atmosphere can play tricks on the data, so you need to remove those effects to get accurate surface reflectance or land surface temperature values i. There are several atmospheric correction models out there; pick the one that best suits your area and data.
  • Geometric Correction: Making sure everything lines up spatially. Luckily, Landsat 7 data usually comes in Level-1T format, which is already pretty good in terms of geometric accuracy i.

The Destriping Toolkit: Methods to Fill the Gaps

Alright, now for the fun part: actually getting rid of those stripes! There are several techniques you can use, each with its own strengths and weaknesses.

  • Gap Filling/Interpolation: These methods try to fill in the missing pixels by guessing their values based on the surrounding data. Think of it as a digital “fill-in-the-blanks.”
    • Local Interpolation: Using the values of nearby pixels to estimate the missing ones i. Simple, right? Techniques like nearest neighbor or bilinear interpolation can work, but be careful. They can smooth out the image, which isn’t ideal if you’re working with a complex landscape i.
    • Histogram Matching: Using data from another Landsat scene to fill the gaps i. The trick is to find a “fill” scene that’s similar to your striped scene and then adjust its radiometric properties to match.
    • Multi-scene Gap Filling: This is where you use multiple SLC-off scenes to fill the gaps i. Since the gaps are in different locations in each scene, combining them can give you complete coverage. It’s a bit more work, as you need to carefully align and adjust the scenes.
    • Spatial-Spectral Methods: These are the fancy techniques that use both spatial and spectral information to fill the gaps i. Methods like Area-to-Point Regression Kriging (ATPRK) sound intimidating, but they can give you better results in some cases.
  • Filtering Techniques: Instead of filling the gaps, these methods try to smooth out the stripes.
    • Morphological Filters: These filters can “close” the gaps, but they can also make the image look blurry i.
    • Wavelet Transform: This technique breaks down the image into different frequency components, allowing you to target and remove the striping by tweaking specific coefficients i. It’s a bit more advanced, but can be effective.
  • Multi-sensor Data Fusion: Using data from other sensors, like Sentinel-2, to fill the gaps i. Sentinel-2 has similar spectral characteristics to Landsat 7, making it a good option. The challenge is dealing with the different spatial resolutions.

    • Getting Your Hands Dirty: Software and Implementation

    So, where do you actually do all this? Luckily, there are several software packages that offer tools for destriping Landsat 7 data.

    • ENVI: A popular choice with a dedicated “Landsat ETM+ Destriping” tool i.
    • ArcGIS: Another common platform with tools for image interpolation and raster analysis i.
    • QGIS: A fantastic open-source option with various plugins for raster processing i.
    • Other Software: ERDAS IMAGINE and PCI Geomatica are also worth checking out i.

    The exact steps will vary depending on the software and technique you choose, but the general idea is the same:

  • Import your Landsat 7 data.
  • Identify the gap pixels using the gap mask file (provided with the data).
  • Apply your chosen destriping algorithm.
  • Evaluate the results and tweak parameters as needed.
  • Export your destriped image.

    • A Few Words of Caution

    • Garbage In, Garbage Out: Destriping can’t magically create data. If the surrounding data is poor, the interpolated values will be too.
    • Smoothing Alert: Be aware that many destriping techniques can smooth the image, reducing detail.
    • Artifacts Happen: Some methods can introduce artificial patterns or color distortions. Always carefully inspect your results.
    • Computation Can Be Costly: Advanced techniques can be computationally intensive, especially for large datasets.

    The Bottom Line

    The SLC failure in Landsat 7 was definitely a setback, but it’s not the end of the world. By understanding the nature of the stripes and using the right destriping techniques, you can still extract valuable information from the thermal band. Just remember to be mindful of the limitations of each method and always validate your results. Happy destriping!

    You may also like

    Field Gear Repair: Your Ultimate Guide to Fixing Tears On The Go

    Outdoor Knife Sharpening: Your Ultimate Guide to a Razor-Sharp Edge

    Don’t Get Lost: How to Care for Your Compass & Test its Accuracy