<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<style type="text/css" style="display:none;"><!-- P {margin-top:0;margin-bottom:0;} --></style>
</head>
<body dir="ltr">
<div id="divtagdefaultwrapper" style="font-size:12pt;color:#000000;font-family:Calibri,Helvetica,sans-serif;" dir="ltr">
<p style="margin-top:0;margin-bottom:0">Dear GLIMS colleagues,</p>
<p style="margin-top:0;margin-bottom:0">I am pretty sure that most of our science community is not giving adequate attention to the important and useful differences between accuracy and precision of measurements when it comes to glacier dynamics. I want to
prove the valid relationships we should be using when we digitize a glacier. I have thus devised a simple test. I attach nine small images, 20 x 35 pixels, corresponding to a dark object that is shrinking from one time stamp (image file) to the next. There
is a dark object immersed in a light background. The dark object is black-ish but has some small variation in brightness. The boundary is known precisely (by me) and is discrete (very narrow in width compared to the pixel size). </p>
<p style="margin-top:0;margin-bottom:0">Your challenge is to tell me what the area is of the dark object in each time stamp-- let's say years (area in square pixels), and then also tell me what the average shrinkage rate is (using any method you want that uses
your analysius output). You can use any method you want for measurement. You can decide whether there are outliers (in which case you may exclude the outliers), or if it is one continuous dark object without outliers. You can either:</p>
<p style="margin-top:0;margin-bottom:0">(1) Do the analysis in quanta of 1 pixel (so it's a binary classification-- dark or light for each pixel).</p>
<p style="margin-top:0;margin-bottom:0">(2) Or resample more finely or otherwise do subpixel analysis, but then revert back to the original pixel resolution as you see in these attached images when you report areas; the pertinent scale is the pixels you see
in these 35 x 20 pixel images.</p>
<p style="margin-top:0;margin-bottom:0">Then tell me your results and how you did it-- manual subpixel, manual pixel binary approach, an automated method, or whatever; and whether the shrinkage rate is a linear least squares or hand-drawn visual regression
curve or an arithmetic mean of the eight sequential image pairs, or a difference of area of the zeroth and ninth time stamps divided by the number of years-- whatever you want to do.</p>
<p style="margin-top:0;margin-bottom:0">I am hoping for ten or more analysts.</p>
<p style="margin-top:0;margin-bottom:0">Feel free to offer your response when considering it a Rohrschach test, but that won't be analyzed or reported
<span>😉 </span></p>
<p style="margin-top:0;margin-bottom:0">Cheers,</p>
<p style="margin-top:0;margin-bottom:0">Jeff Kargel </p>
<p style="margin-top:0;margin-bottom:0"><br>
</p>
</div>
</body>
</html>