Internship Report 2023

 Spatial Services  is an Austrian GIS and Remote Sensing company with significant experience in Big Data Analysis, Artificial intelligence, Drone Imagery, emergency management, and more. For almost 10 years and in partnership with the Department of Geoinformatics at the University of Salzburg, Spatial Services has provided consultancy services and developed GIS projects for ONGs and the private sector. Some of those projects include:

• SELINA – Science for evidence-based and sustainable decision about natural capital Providing robust practical information and recommendations to stakeholders from both the public and private sectors, SELINA will pave the way towards the transformative societal change required to achieve the ambitious goals of the European Biodiversity Strategy 2030 and the Green Deal.
• AWS Great Britain – Market study for smart city solutions Together with geolocated social-media posts, such as tweets from Twitter, can detect hotspots of target information. This information can be used to identify areas of the city where certain issues or events are taking place, such as traffic congestion, crime, or community gatherings.

In addition to the previously mentioned projects, Spatial Services works hand in hand with Medecins Sans Frontieres (MSF), providing a wide variety of analyses and map products like Population assessments, flood analyses, and destroyed buildings detection. This is crucial especially for emergency management, as this improves the coordination and effectiveness of emergency response activities on the ground.

During the Internship period, I contributed mainly to the MSF requests, which included Dwelling Extraction, Maps Delivery, and Flood Detection in areas with a high presence of refugee camps, in countries like Nigeria, Somalia, Kongo, Turkey, and Sudan. The motivation behind this was to provide information to NGOs on the affected population after an emergency (e.g. Earthquake and Floods) by carrying out a building, huts, or tents census before and after the disaster. In this way, the NGOs could coordinate and prioritize assistance on the ground based on the number of affected people.

With respect to the Flood Detection tasks, they were crucial to determining not only the extent of the flooding but also as a preliminary step for calculating flooding speed rates. Finally, I contributed to the automation of documenting processes, including table filling and the creation of feature class schemas based on Excel files in ArcGIS Pro.

The dwelling extraction was done in ArcGIS Pro software. Initially, this process was carried out through manual and visual inspection of the dwellings. This task consisted of digitalizing the dwellings observed in the related raster images representing the same area but in different time periods. As a result, it could be possible to estimate the population of a refugee camp, before and after a disaster, or just its population growth.

After a couple of those tasks were completed, Spatial Services hosted a Deep Learning Workshop, in which we learned how to use the Deep Learning Packages and Libraries in ArcGIS Pro to create, train, and apply a model to detect dwellings. In this way, the extraction of dwellings was automatized and the time required was reduced considerably.

The workshop hosted was based on the " Use deep learning to assess palm tree health " Esri Tutorial. Finally, we were also taught how to use the Segment Anything Model "SAM" on ArcGIS Pro.

The flood detection was carried out through Google Earth Engine (GEE), a cloud computing platform that provides access to high amounts of satellite imagery and processes large-capabilities to analyze thousands of images simultaneously in the cloud, without requiring to download the data locally.

In this case, flood detection was applied to South Pakistan, between August and October 2022. The code implemented can be accessed through the following link:  https://code.earthengine.google.com/d2ea54907eb8fb2eabbfb99613693855 

The algorithm developed uses the "COPERNICUS/S2_SR_HARMONIZED" collection. Using the "S2cloudless" probability layer, cloud and shadow masking was applied. Then, with a modified Normalized Difference Water Index (NDWI), using band 11 (SWIR 1) instead of band 8 (NIR), and a threshold of 0.20, the water pixels were detected.

As a result, a time-series chart was created in which the flood extent can be observed.

Finally, the interns were asked to improve the internal processes of documentation, database, feature class creation, and interoperability between ArcGIS's existing data and incoming Excel data. For this purpose, using ArcGIS Pro Model Builder and the Arcpy Python package, different geoprocessing tools were created and tested. The main functionality of the tools was to populate automatically all the fields of an existing feature class with the information provided in just one row of an Excel spreadsheet. This was accomplished through the use of  cursors .

Working with the Spatial Services team was a great experience. They were supportive and willing to answer any question. The learning outcomes which include deep learning, raster analysis, Google Earth Engine, and Python programming through ArcGIS Pro, were perfect for clarifying and consolidating concepts that had not been made clear in class. Additionally, they were aware of the limited time disponibility during the ongoing semester and were flexible with the schedule.

I would like to take the opportunity to thanks Spatial Services for the opportunity , and encourage Salzburg students looking for an intern to consider this company.