Map elevation at edges of map taken on same day not matching

Hi there,

We have recently signed up to a pro package while we try and figure out the nuances of the product, trying to fit in analysis of the data with other time commitments. We are using DD to look for signs of movement in the land.

We thought we were on to something with a couple of maps showing changes to the extent even the slope was reversed on someScreenshot_135

I have done some reading and comprehend the limitations of the way the pictures are calculated, it’s unfortunate but we are not able to use GCP’s even if we were able to afford them. I was still expecting the data to be accurate relative to the local topography.

In the earlier days we would collect data on a field by field basis, but have since moved on to larger areas to reduce workload and increase accuracy.

While looking over some of our data from March and attempting to line up the edges of the maps and adjust the elevation colours to look similar I noticed that even when calibrating both maps to a common point there is a difference of up to 7 meters in surrounding common points on the maps.

This makes importing the data into Google Earth in order to visually work out on a large scale where things are moving quite difficult. We have several fault lines in this area and are concerned one or more of them are moving.

We are a small community group, so are unable to afford the business package in order to download the contour data which might help figure out what is happening where with the data.

It is a shame because this method of measuring the land would have made our amature study more robust. As we are dealing with authorities they will be scrutinising the data closely and any inconsistencies will only harm our case.

It appears that we have reached an impasse, as the data cannot be proven to be reliable even on maps that were created from the same takeoff location on the same day, of fields right next to each other.

I can only hope that the changes in elevation and slope we see in some maps are not over exaggerated, I am putting my reputation on the line using this data to highlight genuine concerns. If we manage to get someone to investigate the situation further and there is no change in the land elevation it will make me look quite the fool.

From reading your material on the process I felt confident that each map would be accurate to the tolerance of the Z axis on every map, and changes such as the ones we are seeing [images attached] would be an accurate representation of what the land is doing, even if the exact values were not obtainable without further investigation and equipment.

In short I guess I am asking, given the situation is as it is, and we cannot go back in time and gather the data again, how much can the local elevation changes represented on these maps, relative to the previous maps taken in the same locations, be relied upon, if at all.


We are using a DJI Mavic Pro, I know there is a barometer drift issue that can cause some elevation errors, but 7 meters between two points is a bit much. We were quite hopeful for what we might be able to achieve with this service. The account is registered under another email address, I am working with the account holder.

Thanks in advance,


So many questions. Besides the barometer issue it is also known by other posts in this forum that there is also an issue in global accuracy caused by rounding off of the geotag data on the Mavics. Can you provide an outline of your complete workflow? Are those elevation maps or cut/fill maps? The colors of elevation maps can be manipulated and will depend upon the entire range of the elevations in the map. Obviously 7m is way out of whack so I think something in the drone data or process is awry. With known issues a standard deviation up to 3m can be expected without GCP’s and everyone thinks ground control points are expensive, but it’s a matter of purpose. GCP’s for real global accuracy can be, but there are ways to get fairly inexpensive submeter repeatable accuracy which I assume is what you need for a study such as this. Set targets, use an auto-level, process the map without GCP’s as a baseline, open QGIS to get x,y coordinates and add the elevation from the auto-level loop. You then have a base network to create repeatably relative maps.
This may sound like a lot of work, but it’s really not as a setup stage of the project. Surveying is work and a good understanding of best practices and techniques used is a must for any project of this nature to be proven and successful. I would be happy to work through some analysis with you if you like? Just message me.

@MichaelL We collect the data using DJI GS Pro, 80% overlap on all sides, then uploaded as a complete batch [less than 1000 pictures]. These are the elevation maps form the toolbox. We are simply unable to create GCP’s due to lack of access to the locations.

We don’t have any softare to process the data other than Google Earth.

I am new to all of this.

Makes sense. If you can’t get on the land then you can’t physically place targets. You might try the site below. Any permanent/semi-permanent object that you can see on the map can be a GCP. The point is to use the same network every time. Your real world elevations will be off, but your maps will be relative to each other.
I would also encourage you to learn a piece of GIS software. QGIS is probably the most widely used and has great support online. Like most other topics there is plenty of information on YouTube to get you going. The best part is being able to pull your maps in and get real coordinates in your coordinate projection. You can locate relative GCP’s in this manner as well. Again elevations will not relate to something like a survey benchmark, but your first map will be the base and the rest will be relative to that.

I would suggest running the processes again through DD. Like you I am using a Mavic Pro and albeit I too am not using GCPs, some of the results have been very very good when comparing with a measured topographical survey, but others have been disastrous. I’ve uploaded the same set of images in the past and received four very different results and like you had to question which way the ground was sloping. Given it was the same set of photos I uploaded, I could only put it down to inconsistencies in how DD had processed the information. I have added GCPs in the past and compared and again with a measured survey and still had inconsistent results that cannot be explained.

All I can advise is that you reprocess them again and see what comes out the other end but like you, I do question how reliable the results can be. In my mind and putting aside the fact that you can manipulate the elevational spectrum, something’s very wrong if the fall of the ground is changing from left to right.

You can see a bit more infom on the situation here;

To monitor change or movement over time you can fly the subject area first and then create relative GCPs from that map to be used in future maps so that they are in relation to each other. The only requirement is to use permanent objects that will not move.