Distance in SRTM image vs. Google Earth/etc

Question

I am producing DEMs and various maps for own book project; just learned to use GIS software for this project. I am using QGIS (2.14.21) and SRTM data (CRS:WGS84 EPSG:4326). I want square maps that extend a specific distance in four main directions (E,W,N,S) from the center (38.504421, 43.331467 or 38° 30' 15.9156'' N 43° 19' 53.2812'' E).This was my procedure: in Bing Maps, measured 100km to west from center & 100km to north from there to find coordinate1. did same thing to east&south to find coordinate2. checked coordinates&distances in Google earth. then used the coordinates to clip the raster image in QGIS. did this for 100km, 200km and 400km.

Now I have "rectangular" maps instead of "square" maps. I measured 400km radius images in QGIS, they are correct in N & S, but QGIS shows a little above 500km in E & W. Measurements are same for DEM and raster images. In G.Earth when I measure 400km from my center point to east there is (for ex.) a mountain; that "same mountain" is in correct position in my clipped images in QGIS (correct, according to my procedure, 400km radius), but when I measure the distance from center to that mountain in QGIS, I get 500km. This measurement difference occurs only in east and west.

I searched for an explanation and concluded that this shape difference is the result of map projection. But I am biology teacher in profession and I have very little knowledge about GIS, cartography, etc. But I need to know whether "map projection" is the answer or I must correct all of my maps. Because, in a chapter of the book I will argue the effect of map projection (the distortion) and I will use the shape of my maps as an example. Also I must give the distance information correctly in all chapters to explain the limits of the area in the maps. I could only use QGIS measurements to clip original rasters; I didn't do this, because I wanted to pinpoint a center location in satellite photos and G.Earth seems to measure distances more accurately (like length of a street).

Is rectangular maps OK? Are all these the result of "map projection"? If I must correct the maps, how can I do it?

Answer 1

Yes, this is most likely a projection issue.

Google/Bing imagery are presented at a Spatial Reference System known as Pseudo Mercator (also known as Web Mercator). This is a projected SRS based on the Mercator projection, but optimized for fast response time in web services (it uses spherical Earth calculations instead of ellipsoidal Earth, which simplifies things a lot).

Your QGIS project, however, is in a geographic SRS (World Geodetic System 1984). This means that its coordinates are tied to the planet's surface, not a projected plane. One consequence of projecting a rounded surface into a plane is that certain geometrical attributes will not map 1:1. The Mercator projection is a conformal projection, meaning that it retains values for angles as they'd be measured on the planet's surface, but other things like distance and area get distorted. The distortion varies with distance, and is also affected by where in the planet you are (the closer to the poles, the more a Mercator map distorts in the horizontal axel).

This means that, anything measured in Google Earth will have some distortion, so your 400km are actually 500km. You can read more about different map projections and how they affect geometry here.

WHAT TO DO ABOUT IT?

Since your source projection is conformal, it'll always give you errors in measurement. So, what you need to do is change sources, to an image source that is either in an equidistant projection (very unusual) or in geographic coordinates. LandSat images come in WGS84, so they'll give you accurate measurements. Sentinel-2 images are another option.

You'll have to download these images (the scenes that span the approximate area you desire, you can use the visual cues from Google Earth to estimate that), merge them together in QGIS (Raster -> Misc -> Mosaic), then do what you did before, but this time measuring the distances directly in QGIS (make sure the project CRS is set to WGS84 - 4326).