`measuresann_interp` documentation

The `measuresann_interp` function provides annual surface velocities from Mouginot et al's MEaSUREs dataset.

## Contents

## Installation

This function requires all 11 of the .nc data files that make up Mouginot's MEaSUREs Annual Antarctic Ice Velocity Maps 2005-2016, Version 1. Download them at http://nsidc.org/data/NSIDC-0720 and put them somewhere that Matlab can find them. Then you should be good to go.

## Syntax

[vx,vy,t] = measuresann_interp('velocity',lati,loni) [v,t] = measuresann_interp('speed',lati,loni) [errx,erry,t] = measuresann_interp('error',lati,loni) [count,t] = measuresann_interp('count',lati,loni) [...] = measuresann_interp(variable,xi,yi) [...] = measuresann_interp(...,'method',InterpolationMethod) [...] = measuresann_interp(...,'inpaintnans')

## Description

`[vx,vy,t] = measuresann_interp('velocity',lati,loni)` gives the polar stereographic x and y components of ice surface velocity at the geographic location(s) given by `lati,loni`. The output `t` vector is just `2006:2016`. Each year corresponds to roughly January 1 of that year. In other words, the first value of `t` is `2006`, indicating the velocity measurements were acquired from mid 2005 to mid 2006.

`[v,t] = measuresann_interp('speed',lati,loni)` gives the scalar value of ice speed at the geographic location(s) given by `lati,loni`. Speed is calculated as `v = hypot(vx,vy)`.

`[errx,erry,t] = measuresann_interp('error',lati,loni)` gives the x and y components of error estimates.

`[count,t] = measuresann_interp('count',lati,loni)` gives the number of scenes used per pixel.

`[...] = measuresann_interp(variable,xi,yi)` performs any of the above, but uses polar stereographic coordinates `xi,yi` in meters instead of `lati,loni`. Coordinates are automatically parsed by the `islatlon` function.

`[...] = measuresann_interp(...,'method',InterpolationMethod)` specifies an interpolation method. Default is `'linear'`. Differences between interpolation methods will probably never amount to a hill of beans compared to measurement uncertainties.

`[...] = measuresann_interp(...,'inpaintnans')` attempts to fill in missing data using John D'Errico's `inpaint_nans` function. (The `inpaint_nans` function is already included in the `measuresann_interp` function.) I cannot guarantee the accuracy of this approach, but if you just have a few missing pixels it can reasonably fill in a little bit of missing data. For large domains, the `'inpaintnans'` option might be slow.

## Usage note

This function will likely be slow for large geographic domains. If you're working on a specific glacier it will probably be pretty quick, but the whole continent will be slower.

## Example 1: Time series at a single location

Here's a time series of Pine Island Glacier surface velocity. I'm using the AMT function `scaloc` to get the location of Pine Island Glacier.

[lat,lon] = scarloc('pine island glacier'); [sp,t] = measuresann_interp('speed',lat,lon); plot(t,sp,'ro-') axis tight box off

## Example 2: Data grids

If you enter more than one location, outputs will be 3D matrices, where the third dimension of the matrix corresponds to time. Output will be MxNx11 because there are 11 years of data in this dataset.

Let's define a grid centered on PIG and make it 250x250, at 0.5 km resolution:

```
[lati,loni] = psgrid('pine island glacier',250,0.5);
```

## Mean velocity

Now we can get the speed and corresponding time vectors easily with `measuresann_interp`:

```
[sp,t] = measuresann_interp('speed',lati,loni);
```

Now we have `sp`, which is a 501x501x11 matrix. Plot the mean velocity from 2005 to 2016 like this:

figure pcolorps(lati,loni,mean(sp,3)) axis tight measuresps('gl','color','k') cb = colorbar; ylabel(cb,'mean surface velocity (m/a)')

There's quite a bit of missing data there. Actually, there's missing data wherever there are any less than 11 years of data. An easy workaround is to plot the `nanmean` of the velocity, but just be aware this approach is effectively making a mosaic of velocity measurements collected at different times.

figure pcolorps(lati,loni,nanmean(sp,3)) axis tight measuresps('gl','color','k') cb = colorbar; ylabel(cb,'mean surface velocity (m/a)')

## Velocity trends

To assess velocity trends I'm going to employ another one of my functions called `trend`, which simply gives the linear least squares trend for 3D datasets. And we'll plot with a `cmocean` (Thyng et al., 2016) balanced colormap:

figure pcolorps(lati,loni,trend(sp,t,3)); colorbar axis tight caxis([-100 100]) cmocean balance cb = colorbar; ylabel(cb,'velocity trend m/a') measuresps('gl','color','k')

## Inpainting missing data

Unfortunately the `trend` function does not work anywhere there's a single missing data point. We could loop through each grid cell and calculate the trend of whatever datapoints are available, but then we'd be making a mosaic of trends obtained at different times. That's not an too much of an issue if you're making a mean velocity map, but trends can be mighty sensitive to the time over which they're calculated.

As a workaround the `measuresann_interp` function includes an option to fill missing data. I cannot guarantee this is a good idea everywhere, but it can provide some meaningful insights in some circumstances. Try recalculating with the `'inpaintnans'` option and then plot the trend just like we did above:

[sp,t] = measuresann_interp('speed',lati,loni,'inpaintnans'); figure pcolorps(lati,loni,trend(sp,t,3)); colorbar axis tight caxis([-100 100]) cmocean balance cb = colorbar; ylabel(cb,'velocity trend m/a') measuresps('gl','color','k')

Warning: I will attempt to fill in missing data with inpaint_nans. This might be slow and the numbers it gives you could be totally bogus. Nonetheless, if you're inpainting a few pixels of missing data that are surrounded by high-quality data, maybe it's okay. Maybe.

## Citing this data

If this function is useful for you, please cite the dataset, the peer reviewed article, and Antarctic Mapping Tools for Matlab. It can feel like overkill, but different individuals and entities get credit for the different roles they've played in making the data freely available.

Dataset Citation: Mouginot, J., B. Scheuchl, and E. Rignot. 2017. MEaSUREs Annual Antarctic Ice Velocity Maps 2005-2016, Version 1. Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. http://dx.doi.org/10.5067/9T4EPQXTJYW9.

Literature Citation: Mouginot, J., E. Rignot, B. Scheuchl, and R. Millan. 2017. Comprehensive Annual Ice Sheet Velocity Mapping Using Landsat-8, Sentinel-1, and RADARSAT-2 Data, Remote Sensing. 9. Art. #364. http://dx.doi.org/10.3390/rs9040364

Antarctic Mapping Tools for Matlab: Greene, C.A., Gwyther, D.E. and Blankenship, D.D. Antarctic Mapping Tools for Matlab. Computers & Geosciences. http://dx.doi.org/10.1016/j.cageo.2016.08.003

## Author Info

This function and supporting documentation were written by Chad A. Greene of the University of Texas Institute for Geophysics (UTIG), May 2017.