EO4GEO International Summer School

Overview

I got an opportunity to participate in EO4GEO International Summer School: Intelligent Earth Observation 2021 co-organized the University of Salzburg and UNEP/GRID-Warsaw Centre, in the framework of the EO4GEO project. Summer school was held from 6th June-9th July 2021 in online mode where experts working in the field of Geoinformatics with Artificial Intelligence present their findings. Participants also got chance to present their project work related to Earth Observation and its applications, as a team on July 6, 2021 in the frame of the ISDE conference hosted by University of Salzburg.

The summer school was expected to address recent advancements of artificial intelligence in EO*GI comprising topics such as analysis-ready data, data assimilation, machine learning, hybrid AI, data quality and reproducibility. The virtual summer school was divided into 3 different phases exhibiting different learning modes. The first phase was delivered on application cases in EO. Detailed information on several thematic domain including atmosphere, land and emergency response was focused in this phase. Second phase was titled as Concept2skills. During this phase as a participant, we received trainings in technical building blocks that helped us in identifying ideas that we wish to present or get involved as a summer school project. Third phase was dedicated in getting more familiar to the solutions that can be used for achieving targets.

EO4GEO summer school started with a simple but comprehensive overview on Copernicus followed by brief discussion in earth observation applications on climate change. EU actions related to CO2 emissions were underlined that lead to the discussion on application of Copernicus Climate Change Services. As an addition to this, an application of earth observation analyzing Urban Heat Island was presented. A session was held with an idea on different aspects of Garden Monitor/My Gardenlab. The idea behind the use of remote sensing in the creation of Garden Map was explained along with its advantages and disadvantages. They explained on how geospatial standards and open-source tools can be used to create an online platform for communication with the garden owners. Application Area Emergency and Radar Remote Sensing was discussed later in the session where focus was drawn on possible applications in the emergency application using active radar remote sensing. End of the first week gave an idea to all of us on several application of EO, that we could put forward for the project work during this summer school.

Second week started with a generic explanation on data and information sources without any specific focus on the platform. Preprocessing steps were introduced followed by classification and accuracy assessment procedures using machine learning techniques. More detailed information on Radar image processing was presented during a session in this week that helped in boosting ideas to the participants who chose their project based on Radar datasets. It was detailed presentation that included introduction to procedures followed in SNAP. For the first time I learned about MODTRAN (MODerate resolution atmospheric TRANsmission), a model is used for estimating atmospheric transmission of electromagnetic radiation. As a part of image progressing, later that day, we were insighted on Applications of sematic segmentation and LSTM in remote sensing image classification. Talks were dedicated to deep learning architecture and its application in remote sensing. Copernicus land monitoring service product introduction was highlighted during a presentation in this phase. This was followed by brief concept on semantic data cube. The week ended with an emphasis working on emerging remote sensing analysis platform, "Terrascope” and an interesting lecture on reproducibility research in geoscience. Terrascope try in making it easy by preventing data downloading same as in GEE and focus on data processing. It allows sharing of data in virtual computer and private computer. During the lecture on reproducibility research, a slide was presented that was highly motivation which said: "If you optimize for the outcome, you win one time and if you optimize for a process that leads to great outcomes, you can win again and again".

Groups were formed based on their interest in EO-related application cases presented in phase 1 and narrowed down by agreeing on a specific topic. These presentations were partly about satellite imagery-derived information products and partly about providing input for GI-based analyses. As a result, participants were given the chance to work together on an assignment related to the application discussed. This phase 3 consisted of a group work task to come up with a solution to a problem related to the topic of interest. 2 groups work together in climate change (Assessing Urban Heat Island Effect), 1 group was dedicated in thematic area of emergency (Mapping burn severity using sentinel 2 for the California wildfires) and 2 other team into land applications (1 group on Integration of sentinel 1 SAR and sentinel 2 MSI time series data for crop yield prediction over agricultural areas in Kenya and the other group on Monitoring the stability of Giza Plateau and its surroundings using SAR Interferometry).

Group Work

I was a member of a Climate Change group that focused on Urban Heat Island (UHI). But since two groups were interested in this topic, we collaborated as a single group with eight participants to complete the project with the goal of comparing Land Surface Temperatures (LST) in Lunel and Paris.

When we first began the assignment, we had different ideas as a big group, and each of us had a unique perspective on problems and solutions. To meet our goal, we narrowed down the project to a specific workflow as we progressed. For this project, I am delighted to team up with Vitoria Barbosa Ferreira, Scott Dearden, Yessica Gutierrez Quenta, Chukwuemeka Igwe, Tanya Singh, Hira Zafar, and Cesar Aybar.

During this assignment, I was able to grab in depth knowledge UHI and its effect. I learned that the primary cause of this phenomenon is the presence of large amounts of manmade materials such as concrete, tarmac, waste heat from buildings and vehicles is a smaller, secondary factor. Other factors that affect the UHI include greenspace, size of the urban area, shape and spatial pattern of the city. It’s important to note that while global warming and UHI’s are driven by different anthropogenic processes there are many interrelated issues at play. We chose Lunel as a town center dominated by traditional housing surrounded by agricultural areas whereas Paris has a mix of building styles and multiple large greenspaces.

Then first technique involved determining Land Surface Temperature in ArcGIS Pro where band 10, 5 and 4 were used to determine Top of atmospheric radiance and NDVI that were used in computing brightness temperature and emissivity respectively to ultimately get Land surface temperature. The outcome showed that the highest temperature experience by Lunel and Paris were 41.1819 °C and 45.0283 °C respectively in the month of Jube of 2019.

Second technique involved the use of Google Earth Engine to derive a time series based on the hot and cold spot detected from the outcome of ArcGIS Pro. After the creation of time series, the average temperature difference between hot spot and cold spot in Lunel is very less initially but it is increasing with time. Where as in case of Paris difference in hotspot and cold spot was observed to be more or less similar.

As an addition technique, our group tried in using Terrascope as a computing platform. We observed that with terrascope computational power increase and had access to large space. We were able to compute NDVI for Paris. But it was challenging when computing complex equation, it was difficult and encountered several errors so we stopped with the outcome of NDVI from these techniques.

You can find our presentation slide in this link.

Take away from summer school

The summer school was a great experience that help me widen my knowledge to apply preprocessing, The summer school was a wonderful opportunity for me to expand my knowledge of preprocessing, classification, accuracy assessment, and other aspects of the satellite data analysis workflow. I learned how to collaborate and work in a group. This school broadened my understanding of working with SAR data, terrascope, and the data cube concept. I found the summer school to be helpful to my studies. I believe that the knowledge and skills I gained during summer school will be useful in my future career or studies. I thank to all the organizer and the contributors in the summer school and also the selection committee for providing me with this opportunity.

Thank you for time and patient in reading this. Be good do good.