300km2 PhotoSat survey in Kurdistan showing existing lines of GPS survey points.

“It’s too dangerous to survey any of the ground targets”

“The survey crew says they can’t survey the ground targets because of land mines and unexploded shells. I think it’s because it’s 50 degrees Celsius (122 degrees Fahrenheit) over there and they don’t want to leave their air conditioned camp. Either way we just cannot get you surveys of any ground points,” said the Geophysical Operations Manager for a 3D seismic survey in Kurdistan in early 2012.

We needed ground survey control to ensure accuracy for the 3D seismic survey

What were we to do? The customer wanted a PhotoSat survey to plan a +$20M 3D seismic survey over a 300km2 area in Kurdistan. We already had the satellite photos but could not get survey coordinates for any ground features that we could identify on the photos.

To be of best use for the 3D seismic survey, the PhotoSat survey needed to be accurate to 30cm in elevation and 1m horizontal. However, without ground survey reference it would only be accurate to about 3m horizontally and vertically.

We had to find a way to use existing ground survey data

Our only option to produce an accurate PhotoSat survey was to figure out some way to use existing ground survey data.

The oil company already had seismic survey lines on a 5km by 5km grid covering the project area. Based on this data they had drilled an exploration oil well at a cost of more than $10M.

It was a discovery well. They had discovered a significant column of oil. Now they needed a 50m by 50m grid of seismic data covering the entire project to decide where to drill next.

Existing lines of GPS survey points every 25m

The oil company’s existing 5km by 5km grid of seismic lines had GPS survey points every 25m along the lines. These survey points were accurate to 10cm in x,y,z. The company provided us with this data.

Absolutely no sign of the seismic lines on the satellite photos

Initially we were hoping that we would be able to see the seismic lines on the satellite photos. We thought that we could use the survey points and some evidence of ground disturbance to accurately reference the satellite photos to the ground.

The satellite photos were already within 3m of their true position. We were sure that when we zoomed into the high resolution satellite photos we would see some ground disturbance along the seismic lines.

No luck there! Due to the arid, rocky conditions of the ground we could see absolutely no sign of the seismic lines on the satellite photos no matter how closely we looked.

Tried shifting the PhotoSat survey grid until it matched the ground survey points

We then attempted to match the elevation profiles along the seismic lines to the 1m PhotoSat survey grid. We chose areas where the seismic lines crossed each other at 90 degrees.

We tried shifting the PhotoSat survey grid in x,y,z until we got the best match between the seismic line elevation profiles and the PhotoSat survey grid at the line intersections.

Shifting and matching seemed to take forever

This shifting and matching seemed to take forever. We started by making an ArcGIS TIN from the PhotoSat survey grid. Next we intersected the TIN with a file of the survey points. Then we extracted the PhotoSat survey elevation at each survey point and compared the elevations.

We shifted the TIN and repeated the intersection and elevation extraction.  We did this over and over again. Finally we found the shifts with the best horizontal and vertical matches between the survey data sets where the seismic lines crossed.

This was a painfully tedious, slow process. That is, until we thought about the alternative – clearing land mines and unexploded shells to get new survey points.

Established ten ground control points at seismic line intersections

Eventually, at each of ten seismic line intersections we located the best horizontal and vertical matches between the PhotoSat survey and the seismic survey data. At all of these ten points the surveys seemed to match to within 25cm horizontally and 10cm vertically.

Adjusted the entire PhotoSat survey grid to match the ground control points

With these ten ground control points we were then able to adjust the entire PhotoSat survey grid. We were hoping we had produced a PhotoSat survey grid accurate to better than 50cm horizontally and 30cm vertically.

North South scatter plot of the elevation differences between +5,000 ground GPS survey points and the 1m PhotoSat survey grid. The standard deviation of the elevation differences is 28cm.

Achieved an overall match to the +5,000 GPS survey points of 28cm

When we compared elevations of the 1m PhotoSat survey grid to the elevations of all of the points along all of the seismic survey lines there was a match of 28cm RMSE.

After this first PhotoSat survey project in Kurdistan in 2012, we surveyed several other Kurdistan projects over the next three years. We encountered the same situation in each project, and solved them all the same way.

We eventually developed a process to find the best matches automatically.  This has now become our standard method on all advanced projects with existing ground survey data. We no longer need to incur the time and costs of surveying ground features that can be seen on the satellite photos.

More details of our current process are available by clicking here.

Looking for more information on PhotoSat surveys?

 Advanced projects with existing ground survey data

 Guidelines for Stereo Satellite Ground Control Targets

Satellite surveying accuracy studies

PhotoSat Technology