Deriving large-scale glacier velocities from a complete satellite archive: Application to the Pamir–Karakoram–Himalaya

TLDR: In this article, a semi-automated approach was proposed to derive robust and spatially complete glacier velocities and their uncertainties on a large spatial scale using complete satellite image feature tracking.

About: This article is published in Remote Sensing of Environment.The article was published on 2015-06-01 and is currently open access. It has received 165 citations till now. The article focuses on the topics: Glacier.

Figures

Figure 10: (a) Uncertainty of the final velocity for the Karakoram and period 1999-2001, (b) zoom over the Baltoro glacier (dash line), (c) histogram of the uncertainty on and off glacier

Figure 1: Processing strategy to derive glacier velocities from a complete multispectral satellite archive

Figure 7: Left : Success rate for each individual pair, in ascending order and for the result of the fusion (red). Right : MAD in stable areas for same pairs in same order.

Figure 8: Dispersion of the velocities estimated from all pairs for the Karakoram and period 1999-2001

Figure 11: Velocity vector coherence for the Karakoram region. A value of 1 means perfect alignment of all the vectors contributing to the median velocity, 0 means completely random directions.

Figure 4: MAD of each component of the velocity in stable areas (plain lines) and success rate (black dots) for different SNR thresholds

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "Deriving large-scale glacier velocities from a complete satellite archive: application to the pamir-karakoram-himalaya" ?

In this paper, the authors propose a processing strategy to exploit complete satellite archives in a semi-automated way in order to derive robust and spatially complete glacier velocities and their uncertainties on a large spatial scale. The authors then discuss the impact of coregistration errors and variability of glacier flow on the final velocity. 

Q2. What is the probability of the final velocity?

When few velocity estimates are available, i.e the measurement509 is spatially isolated or very few pairs allows for a measurement, the residuals510 reach over 20m/yr but as the number of merged velocity estimates increases, the511 confidence in the measurements reaches a few m/yr. 

Q3. What is the MAD for the optimum pair?

The MAD for the optimum pair is 5.5m/yr484 and the mean MAD for all single pairs 5.4m/yr, mainly due to orthorectification485 errors. 

Q4. What is the median uncertainty of the final velocity?

Over glaciers, the526 median uncertainty is 4.4m/yr, from a few m/yr on some glaciers tongues to527 10m/yr in some accumulation zones. 

Q5. How can the authors use this strategy to estimate the seasonal velocity variations?

More complex568 postprocessing strategy as for example time series inversion (Lanari et al., 2007)569 to select the coherent displacements along the time serie could be implemented,570 potentially allowing to derive the seasonal velocity variations. 

Q6. What is the way to calculate the SNR?

Once the feature-tracking parameters and the preprocessing steps are chosen,425 the authors can run the feature-tracking for each available pair to compute velocity426 fields and an associated SNR. 

Q7. What is the uncertainty of the final velocity?

The uncertainty map521 has a similar shape as σ (Figure 8), but is weighted by N ; in particular, on522 stable grounds where there are generally more measurements (less problems of523 saturation), the uncertainty is reduced whereas in snow covered areas, the low524 contrast reduces the number of measurements and uncertainty remains relatively525 high. 

Q8. How many data points are used to compute the median?

if the number of data points used449 to compute the median is lower than Nmin = 5, the authors discard the measurement450 because the median is not robust enough.