Gathering and visualising data

We have covered a lot over the last five weeks! It can be hard to know where to start with these different techniques and datasets. So, this week we are going to present an example case study, demonstrating a real-life workflow. We will gather the most useful data and present it appropriately. We will also digitise the archaeological data contained within these datasets, teach you to make great maps, and how to properly credit your sources.

Lake Hamrin

Lake Hamrin is a reservoir in the Diyala Governorate of Iraq, approximately 100km northeast of Baghdad. It was formed following the construction of the Hamrin dam, built between 1976 and 1981 to generate hydroelectric power and store irrigation water from the Diyala river. Prior to the dam’s construction, archaeological surveys and excavations were carried out by the State Board of Antiquities and Heritage and teams from across the world, revealing a rich dataset then flooded by the reservoir. In more recent years, water levels have dropped in Lake Hamrin due to further damming of the river upstream and rising temperatures due to climate change. Once flooded archaeological sites are now emerging from the lake and are threatened by accelerated erosion from repeated cycles of wetting and drying.

As an example, we will accurately map the location of these sites and investigate the problems facing the archaeology of Lake Hamrin in more detail. You can follow along with the Lake Hamrin data if you like, but this case study is really intended to provide a complete workflow example that can be applied to your own areas of interest.

_{Lake Hamrin is located approximately 100km northeast of Baghdad in Iraq. Data courtesy of Natural Earth.}

Sentinel-2

Sentinel-2 imagery will provide a current overview of the Lake Hamrin area.

Downloading Sentinel-2 imagery

We searched the Copernicus Open Access Hub for Sentinel-2 imagery from 2022. Plenty of great, cloud-free imagery is available, but unluckily the lake sits at the boundary of four different Sentinel-2 scenes. To cover the whole area we have to download a lot of imagery!

_{The Lake Hamrin area lies at the boundary of four Sentinel-2 scenes.}

We selected four neighbouring, low-cloud images from April 2022. During the late spring water levels should be at their highest after winter. We downloaded these images, saving them in a Sentinel-2 folder within our project directory.

_{The four Lake Hamrin images have been downloaded and extracted.}

Using our Sentinel-2 imagery in QGIS

We could just add the four readymade true colour images (see Week 2.6).

_{Adding the four Sentinel-2 TCI images to our Lake Hamrin project.}

However, as we want to also use a near infrared band, we are also going to create our own composites as we did with the Landsat imagery (see Week 3.5) but merge the four Sentinel-2 scenes together. First, we create a virtual raster using the four scenes for each band we are interested in: blue (band 2); green (band 3); red (band 4); and near infrared (band 8A). This works in an almost identical way to that covered in Week 3.5, but this time we do not want to place the four images into separate bands.

_{Creating a virtual raster merging the blue bands from our four Sentinel-2 scenes.}

We can then create our final virtual raster using these four merged bands, generating a four band (red, green, blue, near infrared) composite for the entire area.

_{Creating the final composite as a virtual raster using our four merged band images.}

Once we have selected the right band combination (Week 3.5) and removed the black area Week 4.6 our Sentinel-2 image is starting to look good.

_{False colour infrared composite of the Lake Hamrin area.}

Improving the appearance of Sentinel-2 imagery

If your image looks a bit too bright or too dark, you can change how it is displayed by adjusting the minimum and maximum values for each band in “Properties” > “Symbology”. If your image is too dark you should make the minimum values higher, and if it is too bright you should make the maximum values lower.

_{Tweaking the maximum and minimum band display values can lead to much clearer imagery.}

You can also fine tune these values more precisely by using the “Histogram” tab.

_{Fine-tuning the minimum and maximum values for each band using the Histogram tab.}

You should see a considerable improvement in the clarity of the final image!

_{The final image is much clearer after fine tuning with the Histogram tab. Imagery courtesy of ESA.}

Landsat

We are going to use the Landsat archive to find snapshots from before the dam was built, and at some point from when the water was higher.

Landsat 1 – before construction

As the dam was built in 1976, we need imagery from before this date. Luckily Landsat 1 began collecting data in 1972 so we should be able to find some suitable data using EarthExplorer.

Satellite imagery was in its infancy then so there is much less to choose from in the archive. By searching the “Landsat 1-5 MSS C1 Level-1” dataset (“Landsat” > “Landsat Collection 1” > “Landsat Collection 1 Level 1”) we found a suitable image from 1973.

_{Searching for Landsat 1 imagery on EarthExplorer.}

After creating a virtual raster with the four Landsat 1 bands (Week 3.5) and tweaking the symbology, we end up with a good image from before the construction of the dam.

_{Our 1973 Landsat imagery from before the dam was constructed.}

Landsat 5 – high-water mark

There was a lot more imagery to choose from for a high-water mark image. Landsat 5 was the obvious choice as it was online for much of the reservoirs early life – we downloaded an image from April 1990.

_{Our 1990 Landsat 5 image shows a more typical April water level.}

These three satellite images do a great job of telling the story of Lake Hamrin!

_{Our Landsat and Sentinel-2 imagery tells the story of Lake Hamrin. Courtesy of the USGS and ESA.}

Can you already see some of the sites we are interested in?

SRTM DEM

We downloaded SRTM elevation data from EarthExplorer (Week 4.4), we needed four tiles to cover our area of interest. The tiles have very different minimum and maximum values, due to the sharp contrast between the Mesopotamian plain and the Zagros Mountains, and so they would not look like a continuous surface if we simply add the data to QGIS.

_{The four SRTM tiles we need in EarthExplorer.}

So we created a virtual raster using these four tiles as for the Sentinel-2 images. This involves less work as the SRTM rasters only have one band. Once created, the merged DEM can be assigned a colour gradient (Week 4.6). We assigned minimum and maximum values based on the original tiles, but we dropped the higher end slightly to get a clearer view of the lake environs.

_{The SRTM virtual raster appears as a single, continuous surface.}

Slope and aspect (Week 4.7) are not hugely relevant to the archaeology of Lake Hamrin, but we did add a semi-transparent hillshade raster (Week 4.6) to highlight the topography.

_{SRTM data of Lake Hamrin including a semi-transparent hillshade.}

Although this creates an attractive DEM for a map, it can be more useful to select a colour ramp with more colours to see more detail. It can also be a good idea to have the minimum and maximum values change automatically as you move around the raster. Make sure “Min/max” is selected under “Min/Max Value Settings” and change “Statistics extent” to “Updated canvas”.

_{Selecting settings to update the colour of the min/max values as you move around the map.}

Now we can pick out some of the mounds around Lake Hamrin!

_{Some of mounded sites around Lake Hamrin are visible in the SRTM data.}

Historic maps

Next we will add several historic maps – a series with a grid, and an archaeological survey map.

Gridded topographic maps

A search of the Perry-Castañeda Library Map Collection (Week 5.5) yielded a topographic map series from the 1940s. Unfortunately, our Lake Hamrin area lies split between two of the maps! These were georectified using their grid (Week 5.7).

_{These maps cover the Lake Hamrin region, but frustratingly it is split between the two.}

To overcome this issue, we will clip away the borders of the two maps using the “Clip raster by extent” tool, found in the Menu in “Raster” > “Extraction”. Simply add in the minimum and maximum x and y values, save your new map as a TIFF or virtual raster and click run.

_{Clipping the overlapping maps by their extent.}

This allows both maps to be seen properly side by side.

_{The two clipped maps can now be seen properly.}

Including some information on the archaeological sites of Lake Hamrin.

_{A number of archaeological sites have been marked on the topographic maps.}

Ungridded archaeological survey map

Before the dam was constructed archaeologists carried out surveys and excavations in the area. Many published their results, including maps that can help us to locate their sites. The most detailed map comes from a book edited by MacGuire Gibson on excavations at Uch Tepe. It gives the names and locations of seventy sites, but it doesn’t include a grid! This had to be georeferenced manually, by finding shared features between the map and the Landsat and Sentinel-2 images.

_{We found eleven ground control points to georeference the archaeological survey map. Map is figure 2 from Gibson, M. 1981. Tepe Uch. Oriental Institute: Chicago.}

Corona imagery

Corona imagery mostly dates from before the construction of the dam and should be much higher resolution than the Landsat 1 imagery we’ve already found.

EarthExplorer

A search on EarthExplorer (Week 5.3) revealed that five strips would need to be downloaded and (partially) georeferenced, many hours of work!

_{Five strips of Corona would need to be downloaded and georeferenced from EarthExplorer to properly cover the Lake Hamrin area.}

CAST Corona Atlas

Luckily, lots of ready-georeferenced Corona imagery is available to download for free on the CAST Corona Atlas website. Although we don’t have time to cover this great resource in detail, it is definitely worth exploring as it contains over 2000 strips of Corona, mostly of the Middle East. Missions can be turned on and off, and then expanded to download individual scanned strips of film. The only slight downside with the CAST imagery is that the georeferencing is not always as accurate as when it is done manually.

_{All of the Lake Hamrin Corona is available to download for free on the CAST website.}

Once downloaded and added to our QGIS project, the last of our data has been gathered!

_{Adding our CAST Corona imagery into QGIS.}