NoR sponsored projects

ESA will organize a training course in Riga, Latvia, from 27 June – 1 July 2022, in collaboration with the Ministry of Education and Science of the Republic of Latvia, the Institute of Environmental Solutions, and Riga Technical University. Objective: The training is designed to promote and disseminate EO data and information-based solutions in various scientific and industrial fields. The program will provide theoretical information followed by practical computer exercises and feature the use of Copernicus Sentinel-1 data (SAR sensor) and Copernicus Sentinel-2 data (optical sensor). Audience: The course is intended for researchers, students, Ph.D. students, and young professionals who use EO technology within their research or work and want to improve their knowledge of remote sensing. Preference will be given to applicants from Latvia and other Baltic countries. Main topics: Introduction to ESA EO missions; SAR & optical data for land cover applications, including climate change impact; SAR & visual data for forestry, including climate change impact; SAR & optical data for agriculture, including climate change impact; InSAR data, including terrain motion due to gas; SAR for ship detection; Integrated applications.

This PhD research project aims to analyze land degradation in traditional and mechanically-constructed mountain terrace landscapes in Cyprus. Project objectives are:

1. To enhance existing models for evaluating the stability of terraced hillslopes.

2. To implement such models on large areas using GIS techniques.

3. To develop guidelines for the design and construction of new bench terraces. Those guidelines are expected to address the full spectrum of terrace design and construction (terrace geometry, land utilization, fill/cut percentages, dry-stone wall dimensions, vehicle access roads, field drainage, irrigation network, construction methodologies etc.) on a more practical view.

4. To develop models and methodologies for evaluating the stability of terraced hillslopes.

This study intends to estimate crustal deformation in the form of land subsidence from vertical displacement and velocity maps from SAR products and investigate sub-surface processes using gravity modelling techniques (2D modelling from GOCE products). Sub-surface structures are being modelled from gravity anomalies, and the displacement map from SAR differential interferograms will be obtained from the GEP platform. Although this study is limited by the absence of subsidence rates and uplifts, Bouguer anomaly data from the GOCE satellite data repository was acquired and separated into residual and regional anomalies. Forward modelling of sub-surface structures was achieved from residual Bouguer anomaly, while delineation of faults was done from the total horizontal derivatives of the gravity anomaly. The approach in this study will contribute to the knowledge base on remote sensing applications for crustal deformation studies in sedimentary basins within Nigeria and Africa as a whole.

Furthermore, the integration of GRACE and InSAR will improve the monitoring accuracy of crustal deformation detection as the separation of sub-surface densities modelled from the Bouguer anomaly and the faults delineated from the anomaly gradient corroborate the vertical displacement determined from the InSAR. This supports further probing of the sub-surface interactions that lead to deformation and subsidence in Nigeria and Africa. The objectives of this study will be summarised: To generate displacement and surface velocity maps for the last ten years using interferometric synthetic aperture radar (InSAR) datasets such as the Sentinel-1, ENVISAT, and RADARSAT. The deliverables will be further analysed with gravity anomaly distribution in the study area.

A major challenge facing African agriculture is the lack of field boundary (i.e. parcel) maps. Field boundary maps provide the foundations for understanding the characteristics and extents of agricultural systems and how these are changing. This information is essential to organizations that provide services that smallholder farmers need to improve their yields and access to markets, and to adapt to a rapidly changing climate. This project will develop a comprehensive, high-quality set of labels digitized on PlanetScope imagery over Africa intended for training generalizable, regionwide field boundary mapping models, and for refining and validating models for specific regions and years. The labels will be freely available under a Creative Commons license and hosted on Radiant MLHub, from where they will be easily ingested into machine learning pipelines. To date, no such labeled dataset exists, despite the growing interest across the public and private sectors in mapping field boundaries in Africa.

The primary objective of this project is to develop, implement, validate and showcase advanced AI-based methods to automatically map and characterize the spatial extent of slums from Earth Observation (EO) data. This objective is framed and informed by data needs of national and local governments, as well as the civil society, to monitor progress on SDG indicator 11.1.1 on the proportion of the urban population living in slums, informal settlements or inadequate housing. Furthermore, the objective is linked to the information needs of a diverse group of stakeholders that engage in understanding and improving local living conditions.

The primary objective of this project is to develop, implement, validate, and showcase advanced AI-based methods to automatically map and characterize the spatial extent of slums from Earth Observation (EO) data. This objective is framed and informed by the data needs of national and local governments and civil society to monitor progress on SDG indicator 11.1.1 on the proportion of the urban population living in slums, informal settlements, or inadequate housing. Furthermore, the objective is linked to the information needs of a diverse group of stakeholders that engage in understanding and improving local living conditions.

The primary objective in both zones is to identify and quantify tectonic ground deformations (and volcano-tectonic ground deformations in the case of El Salvador) by combining INSAR analysis and GNSS data from geodetic networks that are installed around the main active faults and volcanoes of the region. In the case of SE Spain, the study will focus on the active faults of the Eastern Betics Shear Zone (EBSZ) bounding the Guadalentin tectonic depression: the Alhama de Murcia and Carrascoy faults, both faults with high seismic hazard in the Iberian Peninsula. We have been establishing a continuous GNSS network around these faults since 2016 (Staller et al. 2018), and until now, four campaigns have been carried out. GNSS velocities will be used to validate the InSAR results. In the case of El Salvador, our main objective is to estimate the deformation field across El Salvador volcanic arc to determine spatiotemporal variations of the slip rate along the El Salvador Fault Zone (ESFZ). Geodetic estimates of slip rates along the ESFZ have been published by Staller et al. (2016) using 30 GPS campaign stations measured from 2007-2012

The objective of this project is to provide training and high-resolution NDVI and NDMI maps to facilitators of the Alito Training Center in Uganda (who manages the Alito Farm) in order to optimize farm inputs, maximize yields, and to promote sustainable agriculture practices.

The objectives of the project are:

• Continue to update the Web base Mapping Applications with improved accuracy and usefulness to the avocado industry and build on the solid progress.

• Yield forecasting model supporting other benchmarking project and crop forecasting and continue to allow investigation into the relationship between climate and yield to inform the remote sensing climate-based yield prediction model.

• Expand testing and structured feedback process to allow for improved decision-making and orchard management by producers using the CropCount Mobile Application as a means of improving productivity.

• The CropCount Moblie Application will also support avocado growers with new plantings or no historical data (that would be used in the time series analysis).

Investigate (propose a solution) how sustainable agriculture can be achieved in the African continent under drought conditions by co-developing innovative scientific EO-based and state-of-the-art modelling solutions with African experts. The project aims at increasing experts’ knowledge and capacity, developing an operative platform and database for results visualization and sharing with end­users.

The AFRI-SMART project will tackle this challenge in Morocco.

Objective 1. EO-based solutions addressing sustainable agriculture and drought monitoring for Morocco:

• Estimating water availability and crop irrigation water needs under a changing climate at multiple spatial scales at present and forecasted times;

• Improve water management at the national scale for present and forecasted water availability.

Objective 2. Agile development along all project phases to maximize the impacts and to guarantee that the final output is responsive to necessities. Sharing project approaches and activities, defining the output indicators, and customizing the web tool, so that uptake will be easily promoted in the stakeholders’ habits. Data repository and a web dashboard.

Objective 3. Integrated use of multiple earth observation data with hydrological – crop modelling schemes at different spatial and temporal resolutions (from SMOS soil moisture data to Landsat/ECOSTRESS land surface temperature and Sentinel-2 vegetation information).

As part of AGEO-INTERREG project, several Citizens’ Observatory pilots on geohazards (landslides, rockfalls, floods, peat movements, earthquakes, coastal hazards, geotechical risks) are being launched in France, Portugal, Spain, Ireland and UK. The use of EO processed products will be useful to analyze the hazard as a process through the estimation of deformation rates, flooded areas and geomorphological parameters, which will be extremely useful in the risk management tasks. Moreover, AGEO aims to encourage the local use of innovative EO products and services provided by European data infrastructures.

The project aims to face food insecurity using modern technologies, including remote sensing. The number of people worldwide affected by hunger increased in 2020 under the shadow of the COVID-19 pandemic. After remaining virtually unchanged from 2014 to 2019, the prevalence of undernourishment ascended to around 9.9 percent in 2020 from 8.4 percent a year earlier. In terms of population, taking into consideration the additional statistical uncertainty, it is estimated that between 720 and 811 million people in the world faced hunger in 2020. Considering the middle of the projected range (768 million), 118 million more people were facing hunger in 2020 than in 2019 – or as many as 161 million, considering the range’s upper bound.

Unless bold actions are taken to accelerate progress, primarily measures to address significant drivers of food insecurity and malnutrition and the inequalities affecting the access of millions to food, hunger will not be eradicated.

So we should use all of our tools to face this dilemma, and we, as gis and remote sensing and artificial intelligence experts, should use our techniques.

We believe we should develop applications that provide helpful information for the leading players in the field of agriculture who make food for us. In addition, we need to make our agricultural procedures more innovative and intelligent.

So the added value of our project would be providing valuable insights for agriculture sector players to help them make more efficient decisions.

Farms invest time and money to improve yields, but available solutions help by only 10-30% and are insufficient for rising food demand. Our company provides the first autonomous irrigation and fertigation system with recorded yield improvements of 200%+ and significant water and nutrient conservation enabled by a proprietary soil optimization algorithm. Our AI solution continuously collects real-time plant and soil data, determines precise irrigation and fertigation commands, and executes them on demand to maintain an ideal soil environment around the clock. Farms using Agrinoze can implement unique agrotechnical, and regenerative farming approaches incompatible with typical irrigation regimes, minimizing environmental and economic costs of food production. Agrinoze transforms farms into efficient and profitable local producers while paving a sustainable path to global food security. Access to satellite imagery will help us further optimize Agrinoze’s monitoring capabilities, leading to more significant yield improvements and resource-use efficiency. We are looking to implement satellite imagery to increase accuracy in two main areas:

1. Monitoring vegetation to improve our irrigation algorithm, which is currently based solely on in-field sensors.

2. Monitoring proper functioning of the irrigation system (lack of water, leakages etc.) to reduce waste of precious resources such as water and fertilizer and eliminate production loss.

Objectives: Agriculture is one of the few sectors that humanity can not live without, where the climate impact is large (20% of total emissions in Sweden) and furthermore assumed to be difficult to do something about it. However, increased agricultural productivity, i.e. more photosynthesis, results in positive climate effects which IPCC does not fully count. Huge amounts of CO2 are caught by crops that, in turn, generate huge amounts of O2. The yearly agriculture carbon dioxide binding capacity is approximately 15 tons per hectare (crop harvest, straw and roots) The operational understanding of what really happens in a field when crops are growing are clearly lacking in spite of the tremendous amounts of data that modern agriculture equipment is gathering. This results in suboptimal decisions for land use, crop selection, machine usage, fertilization and irrigation for both economic productivity and the climate. Remarkable is that the very detailed harvest data (measurements every fifth second) which have been collected by harvesters for many years is hardly used for operational feedback at least not in Sweden. The project will use AI to quantify limiting agriculture factors, to optimize crop growth in a climate beneficial way and long-term agriculture productivity. Hushållningssällskapets existing

platform markkartering.se will be used for operational usage of the results by the farmers as prescription files for the agriculture equipment.

Extensive field, soil, satellite and sensor data will be used as input to model algorithms driven by Al /ML and spatial analysis (GIS). These algorithms will over time be used to create prescription files to control inputs and other actions in the fields as efficiently and climate friendly as possible. The project objectives are:

#To create a climate positive prescription file model as a decision tool for the next level of precision agriculture. In the end, the result will be presented as prescription files for field equipment, including autonomous vehicles. The results can also be obtained in table form, text or as an interpolated map.

#Expand and enhance Hushållningssällskapet (the Rural Economy and Agricultural Societies) web- based decision support system Markkartering.se (that today has over 2000 active users, about 800 000 hectares) to support as climate friendly and operationally efficient production of food as possible. This will optimize the agricultural climate-affecting factors in multiple ways: #Enhance crop growth and thus bind more carbon into the soil that slowly heals high levels of greenhouse gases;

#Optimize agricultural machine usage and driving;

#Optimize inputs like fertilizers and pesticides;

#New governmental directives that focus on long-term sustainable climate improvements.

The result of our project will be gender neutral however we will try to develop a tool that will change the values of agriculture and in the long-term increase equality both in agriculture and in AI. The project potential of an increased gender balance for farmers is to offer an advanced decision basis that is easy to learn and use for all genders.

The AI for the Arctic (AI4ARCTIC) project applies deep learning, in particular deep convolutional neural networks, for Earth observation applications within the cryosphere, focusing on sea ice and snow. The project trains deep-learning systems from relevant training data and tests and demonstrates the capability of deep learning by applying it to a large-scale inference of cryosphere-related variables. The project focuses on two use cases, one on snow mapping in Scandinavia and the other on sea ice charting in the waters around Greenland.

The AI4FOOD project investigates advanced Artificial Intelligence (AI) and Machine Learning (ML) techniques to develop new algorithms for the creation of fused (with a focus on Sentinel-1 SAR and Sentinel-2 optical) continuous data streams and evaluate aspects such as time series predictability over different land environments. A consortium of industry experts on data fusion and time series techniques, and open-source implementation and operational service provision to users do this. Within AI4FOOD, the consortium strives to create an open-source, modular, extensible, and reusable toolbox called fuseTS. To support the fusion of complementary EO data streams and time series analytics, relevant algorithms will be integrated into the toolbox as a service.

Next to the FuseTS Python library, which can be installed locally, the AI4FOOD toolbox will also be available as a cloud-based on-demand service. Existing operational services, such as the OpenEO platform and Euro Data Cube, will be used for the actual deployment. This cloud-based implementation will lead to a scalable software-as-a-service (SAAS) approach, ensuring that a wider group of users can implement data fusion and time series analytics techniques to develop EO-based services. The algorithms that are included in the toolbox are demonstrated through 3 real-life use cases:

• Subtle Land Cover Change Monitoring (FAO)

• Cropland Phenology Indicators (ITACyL)

• Agriculture and Land Management Activities Identification (Agency for Agricultural Markets and Rural Development, Slovenia)

As part of our Corporate Social Responsibility initiatives, we are working on a use case concerning whales. Overall, this use case would like to support remote detection of the whales and enable a lower rate of collision between whales and ships, to safeguard these whales. Whales are one of the most important species of and for our ecosystem. Yet, humans are primarily responsible for endangering this species, as many whales die from collisions with ships that cross their seasonal paths. We intend to automatically detect the whales’ locations by using AI algorithms within VHR satellite imagery and, depending on the training data quality, to detect the species of it. For training data, we have the logs of a few organizations that are also freely available. Manual detection of these animals in vast areas through satellite imagery is time-consuming and prone to error. By utilizing VHR satellite imagery of different areas, especially in the regions that whales usually cross, we want to identify the location of whales (in near-real time). After detecting their location, this information would be communicated to ship captains and respective organizations. Through this, we want to decrease the number of collisions and contribute to saving the species of whales and ensure their existence. Through our use case, we want to contribute indirectly to maintaining and improving our fragile ocean ecosystem.

ARGANS Ltd is sub-contracted to INDRA to deliver use case 3 in the GDA Disaster Resilience project. This use case will supply innovative products to support coastal change analysis and investment decisions. Use Case 3 will monitor coastal trends along the Volta Delta, including Keta & Songhor Lagoons in Ghana (reaching just across the border with Togo) to support the WACA program.

This use case will support Ghanaian institutions in assessing how natural events, including climate change and human intervention along the coast (such as coastal engineering and sand extraction), have affected coastal dynamics (or morphologies), so they can advise on their significance.

The main objective is to provide coastal change indicators derived from Earth Observations. One of the key user requirements is to optimise how many images can be made available, i.e. cloud free in the coastal area. In addition, one of the tasks will be to provide customer-ready co-registered waterlines and shorelines seasonally covering 25 years with enough examples to “bracket” any severe weather events or anthropogenic events such as harbour developments or sea defence construction.

The AIOPEN project will combine and extend the existing frameworks ASB (Automated Service Builder), EOPEN (Open Interoperable Platform for Unified Access & Analysis of EO Data), and EOEPCA (EO Exploitation Platform Common Architecture) with new and innovative services based on operationally mature AI /ML software capabilities to build a new platform that supports the end-to-end AI model development lifecycle.

Our Agriculture Innovation using Remote Sensing (AIRS) project intends to combine the areas of artificial intelligence and remote sensing to create a technological solution to monitor the grape leaves in vineyards using satellite images obtained by the European Space Agency. The correct assessment of these variables allows for sustained decisions to be made with an impact on the management of agricultural areas. Furthermore, implementing precision agriculture practices enables the reduction of pesticides and waste or irrigation water, resulting in a more sustainable agricultural system and the development of rural communities. The AIRS project’s innovation consists of using artificial intelligence to use high-resolution images acquired by Unmanned Aerial Vehicles to increase the resolution of images coming from the Sentinel-2 satellite. The project foresees the implementation in the vineyards of the members of Adega Cooperativa de Pinhel, with the results later made available to the agricultural community through an online platform.

Precise county-wide crop yield prediction provides valuable information for regional agriculture planning. However, it remains challenging for such an accurate forecast due to the effect of complicated weather and soil factors. The short-term weather variations, governed by the meteorological data during the growing season, and the long-term climate change, headed by historical aspects, are among the critical factors that dictate crop yields simultaneously. This project plans to develop deep learning-based solutions for predicting crop yields at the county level across the United States by using the visual Sentinel-2 satellite imagery data and the numerical data computed from the Weather Research and Forecasting with High-Resolution Rapid Refresh (WRF-HRRR) model. We first produce suitable datasets for any location of interest (e.g., a county/ parish) for model developments. Then, the transformer­based solutions will be developed to capture the direct impact of short­term weather variations on crop growth, learn the high-resolution spatial dependency among counties for precise crop tracking, and capture the effects of long-term climate change on crops. This project will result in a set of location-specific datasets available for public downloading to broadly impact the research community in data science, artificial intelligence, meteorology, and agriculture, among others. In addition, the developed deep-learning models will be packed into software toolkits or application portals for stakeholders (including farmers) to use.

The project aims to introduce the training process on specific platforms such as EO BROWSER, Google Earth, and QGIS. Calculations of the volume of salt harvest in the province of LA PAMPA during 2016-2021. Develop a theoretical, practical, and methodological approach to address a control between calculated and declared minerals. The work will include the location, recognition, and digitization of salt harvest beaches from 2016-2021. This will allow a calculation of the annual production volume that will be compared with the volumes declared by the production companies, thus developing a virtual control technique. The images used are those provided by the EO BROWSER platform, which will be digitized with QGIS, and from this, the calculation of harvest areas and volumes will be carried out.

The project is part of the broader research activity currently underway for the archaeological areas of Gabii and Villa Adriana, carried out by the University of Rome Tor Vergata. Although numerically abundant, the preservation of the archaeological sites on the outskirts of the town is often placed in the background compared to that of the “central” archaeological areas. However, natural phenomena linked to normal soil transformation processes are often accelerated by atmospheric phenomena caused by ongoing climate change. The risk of hydrogeological disruption of many ancient sites is one of these. The archaeological area of Gabii, active from the Iron Age to late antiquity, is an excellent example. It is mainly located along a ridge of tuff rock that bordered an ancient lake. It is currently in a precarious geomorphological situation, already witnessed by traces of visible lesions on the ground. Therefore, the study is necessary for three purposes:

• Understand whether a geological movement exists in the areas close to the archaeological remains.

• Quantify this movement over time.

• Identify a trend whereby information can be provided to the competent authorities so that action can be taken within the time required to preserve archaeological remains. In addition, results will be provided to local authorities for better decision-making on the risk associated with the ground motion of the archaeological features. The objective is to allow local authorities to familiarise with the usage of the available online Earth Observation Services and support them with the assimilation of these new services in the daily monitoring and forecasting routine.

As part of the understanding of the hydraulic behavior of the condo river and especially in the lower reach , as well as the protection of islands in this area, this study also aims environmental and economic aspects in the area , as it is already known, the pool Malebo is a strategic area concerning river transport (navigation), irrigation and agriculture, fishing, etc the protection of the morphological degradation of they is also one of the challenges in this work , for this purpose our study aims to master all the scientific information on the hydraulic and hydrological level in order to serve other economic activity and

environmental ,This prompts us to launch the main questions as follows: How can we develop an agent-based simulation model (ABMS) for navigation chart in the lower reach exactly in the stanely pool that includes (hydraulic, hydrological and socio-economic aspects), with this complexity, to find scenario optimal of navigation in the pool male boo , and if we can generate this approach in all the Congo river ? What are the assumptions, approaches and data needed to develop this model?

For my master thesis I am investigating the application of InSAR for glacial lakes in the Himalaya. In order to check the InSAR results I am using optical imagery – sentinel 2, which is why I would like access to the sentinel hub.

In this work, we have become increasingly convinced about the need to strengthen the quality and depth of meteorological data accessed to adequately capture crop primary production patterns. In particular, we identified (crop) evapotranspiration – ETc and radiation as variables of crucial importance. Unfortunately, we cannot publish an operationally viable service due to constraints and inadequate access to such meteorological data. The key result of our work is going to be a set of novel tools which enable monitoring of crop productivity at various spatial scales and are coupled with the already available SatAgro precision agriculture tools, in result enabling individual farms to map the profitability of particular crop fields and crop fields’ sections, also before the harvest. As with other SatAgro tools, the motivation to create these new functionalities is to optimize crop production and, in turn, increase the farm’s fitness and reduce its environmental impact simultaneously.

This project aims to study the application of transfer learning techniques on hyper-dimensional remote sensing data. Here is the list of specific objectives. Objective 1: Learn about the hardware in remote sensing data (e.g., aircraft types, sensors). Objective 2: Learn about image characteristics (e.g., wave bands, image distortions). Objective 2: Learn about different challenges that can be solved using remote sensing data. Objective 2: Learn about the structure of remote sensing data and the statistical methods used to interpret and correct it. Objective 3: Train convolutional neural networks on high dimensional remote sensing data. Objective 4: Test the viability of using the transfer learning technique to train prediction models on datasets with insufficient data to train prediction models from scratch. Objective 4: Document the work process and results in a graduation thesis. This project will be the final project of a bachelor’s degree program in Mathematics and Computer Science. Suppose the transfer learning technique will prove efficient in building prediction models on smaller datasets. In that case, it can be used for future applications (e.g., tracking invasive species in smaller areal surfaces, land cover changes, unsanctioned object building, etc.). This could be especially beneficial for countries like Slovenia, where datasets are significantly smaller due to the smaller land surface. Transfer learning would enable us to train prediction models using remote sensing data of other countries and use this knowledge (parts of the mode) as a starting point to build a highly accurate model using Slovenian data, which otherwise may not be possible due to insufficient data.

The research aims at the archaeological interpretation of the landscape by reading anomalies detected in the ground visible in optical data and radar data from satellites linked to climate-changing events like summer fires. The identified areas of interest are the archaeological areas of Gabii and Villa Adriana (UNESCO Cultural Heritage), located in Italy. The research aims to understand whether the archaeological interpretation of the landscape through the study of surface anomalies can change due to the occurrence of natural phenomena related to climate change. The archaeological area of Gabii, active from the Iron Age to late antiquity and currently consisting of about 70 hectares of countryside, could be an excellent place to test it: from a preliminary analysis already carried out using optical Google images dating back to August 2020, when much of the Archaeological Park was affected by a large fire that also involved the ancient remains, it emerged that the archaeological reading of vegetation anomalies might be different concerning the reading and archaeological interpretation of the vegetation anomalies visible in the same areas prior to the event. The study is necessary for the following purposes:

• understand how weather phenomena linked to climate change can change the perception of vegetation or ground moisture on optical and (possibly) radar data;

• starting from this study, satellite data will be used to perform pre- and post-event analysis on the area and set up a methodology for the forecasting monitoring to identify a Fire Risk Index through the integration of Artificial Intelligence technologies at a later stage;

• multi-temporal satellite data will be used to establish a vegetation index (NDVI time series) useful to understand lifetime changes in vegetation visibility linked to the average rise of temperatures.

Archaeological prospection in Saruq al hadid is of significant interest to find the complete story of prehistoric settlements lived in Dubai, United arab emirates. Located at 50 km in the southeast of Dubai at the north of al Rub’al khali desert, Saruq al Hadid (SA) archaeological site is discovered since 2002. More than 15 000 artifacts have been identified after more than 20 excavations. The location of this site in the middle of desert between the dunes is mysterious because there is no available close freshwater critical for human survival and raw material sources critical for metallurgical industry. The use of remote sensing satellite high resolution radar and multispectral images enhance widely the possibilities of archaeological prospection. This project aims to prescreen potential buried archaeological sites in that desert region. This work is the first attempt made until now in evaluating the detectability of archaeological remains using satellite images data in United Arab Emirates. The outcomes are important to guide and help the excavation missions and the archaeologist for the planning of future excavation campaigns.

At ArchAI, we use ΑΙ to detect archaeological sites on LiDAR and satellite imagery automatically. We have shown the success of this technology with LiDAR data, detecting thousands of previously unknown sites, and our customers include the Forestry Commission and the National Trust. On LiDAR, we specifically look for earthworks (humbs and bumps in the landscape). However, most housing development occurs on farmland where ploughing has levelled earthworks, and satellite imagery is a more reliable source. In addition, archaeology is revealed on satellite imagery in agricultural fields through crop stress revealing sub-soil walls and ditches. Innovations in Satellite imagery have increased the frequency of high-resolution (<1m) information, meaning that a sufficiently trained ΑΙ can now derive archaeology from big data.

Our initial tests on aerial imagery have shown that the results are variable, depending on crops, seasonality, and weather. As such, we require to develop our technology further to account for this and create a higher volume of training data. Satellite imagery has higher temporal revisits, allowing for wider choice and availability of potential training data.

If our research is successful, we will be able to sell these assessments to our customers in the construction industry. It is a legal requirement to consider archaeology before development. However, archaeology is an unknown risk in projects, and it currently takes 6-24 months to reach high accuracy, which involves several stages of fieldwork and even costly excavations. Using satellite imagery and the SentinelHub, our proposed workflow could allow for instant assessment of construction project archaeological risks.

ARSET – Crop Mapping using Synthetic Aperture Radar (SAR) and Optical Remote Sensing is a collaboration between ARSET, Agriculture and Agri-Food Canada (AAFC), European Space Agency (ESA), University of Stirling, University of Ljubljana, and the CEOS Working Group on Capacity Building & Data Democracy (WGCapD). NASA’s Applied Remote Sensing Training Program (ARSET) has opened a new online advanced webinar series: Crop Mapping using Synthetic Aperture Radar (SAR) and Optical Remote Sensing. This three-part training is open to the public and builds on previous ARSET agricultural trainings. Here we present more advanced radar remote sensing techniques using polarimetry and a canopy structure dynamic model to monitor crop growth. The training will also cover methods that use machine learning methods to classify crop types using a time series of Sentinel-1 & Sentinel-2 imagery. This series will include practical exercises using the Sentinel Application Platform (SNAP) and Python code written in Python Jupyter Notebooks, a web-based interactive development environment for scientific computing and machine learning.

The broader research context of my thesis is represented by Short and Mid-Term Fire Danger prediction using EO Data, coming from both VHR and SR images like Pléiades, WorldView, and Sentinel-2 Imagery. The dissertation would explore novel and innovative applications of satellite imagery in the field of Short and Mid-Term Fire Dangers and Fire Forecasting. The proposal’s main objective is to provide development and management assistance for “Civil Protection” for deploying units, use controlled fires to cope with destructive fires, and aid States and local governments to cope with climate change to augment forest resilience.

The objective of the project is to focus on sustainable resources in Africa, assessing deforestation in countries like Gabon, Ivory Coast, Uganda etc. This will help in understanding the potential risk of deforestation and high risk areas, so that we can take necessary measures to manage the phenomenon. Moreover, this will help us as a company to attain our sustainable goals for the future.

Hot and dry compound events affect millions of people yearly and can potentially cause substantial damage to hazard-susceptible objects such as buildings, crops, or automobiles. Nevertheless, the knowledge about the quantification of their interactions evolving in cascade events remains limited (Tilloy et al., 2019). As a result, the total effects resulting from the interaction of multiple hazards can be underestimated as they lead to a more significant impact than the sum of single hazard effects (Terzi et al., 2019). In this study, we would like to focus mainly on hot and dry compound events, including droughts, fires, dust storms, and heat waves. The main objectives are: (i) To assess the risk regionally (by analyzing the vulnerability, exposure, and past hazards), (ii) to identify and quantify single natural hazards with the help of satellite data (For early warnings and needs for assistance during disaster events), (iii) to apply tracking algorithms on satellite data to reconstruct and to forecast individual natural hazards h trajectories, (iv) to combine (i),(ii) and (iii) to assess the total impacts from the hazards interactions. In this project, two principal datasets will be used: ERA-Interim fields and satellite data for detecting and tracking natural hazards.

The objective of our study is to use high-resolution satellite altimetry to assess wave renewable energy potential on the French coasts, with a particular focus on the coastal zone where the energy can be cropped. The novelty is to take advantage of the increased temporal and spatial coverage of high-resolution satellite altimetry data products from the Sentinel-3 mission and use the SAMOSA+ state-of-the-art retracker (Dinardo et al. 2018, Dinardo 2020). This retracker, differently from the SAMOSA2 retracker currently adopted for the generation of the official Sentinel-3 WAT products, allows obtaining more valid geophysical estimates near the coast where contaminated data are typically acquired. Moreover, the customisable processing options available at the SAR processing level and quality flags provided in SARvatore products can be efficiently used to refine the analysis and for filtering purposes to strengthen the analysis. The study period shall cover 1 December 2018 to 30 July 2022 to provide sufficient data to perform the study and indicate possible limitations. An assessment of the wave energy potential will be given for the coastal zone, which is characterised by high energy swell generated by remote westerly wind systems, which is also affected by the strong wave-current interactions that take place in the area where tidal currents are of the order of 2 m/s. The feasibility of high-resolution satellite altimetry-based assessment of wave renewable energy potential in the coastal zone is examined, taking advantage of the increased time and spatial coverage guaranteed by the Sentinel-3 high-resolution satellite altimetry data.

The project goal is to evaluate the accuracy of the atmospheric correction obtained with iCOR4S3 using in situ hyperspectral remote sensing reflectance and also a comparison with POLYMER reflectance data. The area of interest is focused on Lake Erie in North America. The results should contain the entire Sentinel-3 (S3A and S3B) OLCI data series for the Lake Erie area, processed with iCOR4S3. The results will benefit the academic community working on the Global Water Futures program and other researchers on similar topics.

Production of fertility maps, or “FMAPs”, which are, in essence, classified NDVI images, is essential to the functioning of our business. Soil sample locations are defined based on the field areas classified from low to high agricultural productivity. Based on soil sampling, we gauge the number of nutrients (Nitrogen, Phosphorus, Potassium) already in the soil. Our agronomists then provide variable rate fertiliser recommendations to grow a given volume of a particular crop. However, manual FMAP production is slow, labour-intensive and done field by field. Therefore, we need to automate and upscale the production of fertility maps, or “FMAPs”, to save time and money.

Along with the automation of the FMAP creation, we are also interested in synthesising a “peak green” image of a field for each of the five years. We rely on actual images taken close but not necessarily on the “peak green” day. Therefore, we are trialling using spatiotemporal interpolation to synthesise such an image. The interpolation process also relies on removing clouds and cloud shadows in the region of interest; that latter is a third objective and area of innovation. A fourth objective is automating geolocating the soil sampling sites within our client fields. Based on the produced FMAP, we are trialling a soil test point picking algorithm.

Agricultural field delineation is desirable for the operational monitoring of agricultural production and is essential to support food security; however, due to sizeable within-class variance of pixel values and small inter-class differences, automated field delineation remains challenging. Analyzing high spatial resolution Remote Sensing data permits the delineation of farm boundaries. Accurate delineation of farm boundaries is essential for planning and decision-making actions. First, it enables a better estimation of cropland area, which is important information for farmers and agricultural managers (e.g., ministries and private sector players). Farmers often use traditional measurement approaches to estimate the area of their farms, which sometimes leads to high under- or over-estimation. Accurate knowledge of farm boundaries (and, therefore, cropland area) will lead to efficient use of farm inputs such as seeds, fertilizers and pesticides. They may also help to optimize harvest logistics. Second, accurate information on farm boundaries can facilitate land registration and subsequent acquisition of land use rights for smallholder farmers (through a land tenure information system). Farmers, communities and the private sector are mostly deterred from investing in land resources due to unclear land use rights in rural areas. Developing an accurate parcel system through high spatial resolution remote sensing data is an essential first step towards creating a land tenure information system and, potentially, a land taxation scheme. Such a system will reduce land-related conflicts and encourage increased investment in agriculture. It can also improve farmer access to inputs and credits. Third, delineating farm field boundaries can improve crop type classification using object-based image analysis (OBIA) procedures.

Digital Surface Models (DSMs) are digital representations of the Earth’s surface that can be created using various technologies, such as aerial or satellite imagery, LiDAR (Light Detection and Ranging), or photogrammetry. Some typical digital surface model applications include urban planning, Agriculture, Natural resource management, Disaster response, Surveying and Mapping, Environmental monitoring, Archaeology and cultural heritage and Telecommunication. DSMs have a wide range of applications in various fields, and their usefulness is only expected to grow as technology advances. DSMs have the potential to benefit a wide range of stakeholders, including government agencies, businesses, farmers, environmental organizations, researchers, and educators. As an example Agricultural companies and farmers: DSMs can be used to monitor crop health and yield, as well as to plan irrigation and drainage systems, which can help increase efficiency and reduce costs. Satellite data can be a valuable source of information for generating DSMs, particularly for areas where ground-based data collection is difficult or impractical. Some potential benefits of using satellite data for generating DSMs are Wide coverage, Consistency, Timeliness, Cost-effective, and Remote areas. Unfortunately, some people in the above industries believe that only UAVs can answer their needs. Such a belief will limit satellite data usage, which will negatively affect the satellite data market. However, it is possible to produce higher-quality products using newer techniques like deep artificial networks. So this project’s initial goal is to make high-quality elevation models using high-resolution data (like 30 cm resolution and 50 cm resolution). Undoubtedly, high-quality products will change the attitude of different industries to satellite data and will positively affect the market.

This master’s degree project carried out by the University of Applied Sciences aims to improve the quality of cadastral data provided by governmental institutions. Exports of cadastral data provided by European countries provide accurate geospatial information about the location and geometry of buildings. This freely accessible data is used by researchers, companies and private individuals to perform analyses and evaluations that form the basis for decisions regarding the expansion of urban regions. While the data is of high quality in terms of geometric dimension, it is published at such long intervals that it reflects reality only to a limited extent, as there is a likelihood that changes in the building stock have taken place over time. The research aims to provide the missing component of temporal resolution using satellite data that has been consulted and to determine which buildings have been removed and which entries in the database are no longer representative. Applying the product to the dataset will maximize confidence in the data and provide end users with an approximation of the actual state. At the same time, users performing address-specific queries can be provided with an estimate of how far the answer can be trusted. Similarly, the models produced will be made available to government institutions so that, even before publication, there is an indication of where there have been demolitions of buildings and where construction work has taken place that may not yet be recorded in the database. The project serves as a support to the OpenData initiative of the EU, which enables a variety of different use cases regarding urban planning, energy-efficient construction and other areas in the building sector.

This research is one of the requirements to complete my studies at the State University of Semarang. The theme I took was the relationship between green open space and the level of thermal comfort, especially in the city centre of Semarang. Semarang City is one of the metropolitan cities in Indonesia with a high population density and, thus, a significant level of urban development. Continued development reduces green open spaces, even though these are crucial to improve the urban microclimate. This research aims at providing information to maintain the availability of green open spaces in Semarang City.

Thermal comfort level expresses the influence of microclimate on the human condition. The variables used in this study include the area of green open spaces, air temperature, humidity, vegetation density and the level of thermal comfort. This research will produce a map of the distribution of green open spaces, a map of vegetation density, a map of the distribution of the level of thermal comfort and how much green open spaces influence the surrounding temperature conditions. Besides that, from this research, it will be known which areas have a level of comfort that is classified as uncomfortable, a result that can be used as input to improve the local microclimate to provide comfort for the community in carrying out daily activities. It is hoped that this research can benefit the broader community regarding the importance of maintaining green open spaces in urban areas so that environmental conditions are maintained for comfortable living.

The workshop within the CCN 1 of the Agricultural Virtual Laboratory (AVL) aims to provide a good understanding and first hands-on training. The Sen4CAP project developed, validated and demonstrated an open-source toolbox (Sen4CAP system), which can automatically process Sentinel-1 SLC and Sentinel-2 L1C or L2A time series into a set of products relevant to the new Common Agricultural Policy. The primary users of this toolbox are national Agencies (and/or their sub-contractors specialized in EO), but also the private sector and researchers. The Sen4CAP project entirely relies on CREODIAS for the EO processing. The Panta Rhei conference aims to facilitate knowledge transfer between the agencies. This opportunity is unique to express the importance of the Sen4CAP system to its primary users. The workshop will focus on two main aspects:

1. communication of the main evolutions of the system up to now.

2. Performing hands-on training with the system for the newcomers (from the download of the images from the suitable dataset up to the generation of more advanced products) and a question and answers session for the more advanced users.

Spatial planning is a process that aims to mitigate the impacts of human activities and eventual improvement of the state of the environment through the coordination and implementation of various practices and policies. Thus, a critical action for improving the state of the Baltic Sea is to strengthen the territorial and maritime spatial planning capabilities of the organizations operating in the area. We aim to develop an integrated data approach to obtain a full view of the essential processes of land and coastal water areas by combining currently available satellite data sources, in situ observations, and model predictions about dynamic land cover and water quality characteristics.

The BalticAIMS project will reach this goal through the following technical objectives:

• Identify suitable environmental data and GIS materials.

• Integrate, process, and store thematic information.

• Create the data access, visualization, and analysis systems and tools.

The main objective of the project is to use the DaaS service provided by the CREODIAS environment to run the Sen4CAP system to:

• Compare the accuracy of EOStat and Sen4CAP crop type classifications.

• Verify the quality of the crop type classification using only one Sentinel-1 sensor.

• Complement EOStat products with Sen4CAP and uptake in ARMA business processes.

• Evaluate a backup solution for the operational CAP monitoring.

Accurate information pertaining to the spatial distribution of various tree species in a forest stand is crucial for better monitoring and management of boreal forests. Such wall-to-wall information is lacking from field-based forest inventories. Meanwhile, remote sensing techniques based on satellite and

Unmanned Aerial Vehicle (UAV) data promises to highly reduce this information gap. The end result would be a publication and/or technical note that describes how ESA’s satellite date can be used along with UAV image data for better forest tree species mapping in boreal conditions. The benefit of userready data available from processing platforms (like F-TEP) will also be highlighted in the document.

The beneficiaries of this research project will be forest stakeholders such as forestry companies and government agencies.

The primary objectives of the project are to:

• Support definition and cooperative implementation of Danube and Black Sea regional priorities.

• Achieve measurable progress in embedding EO-derived information into the strategies and cooperation actions within the Danube and Black Sea region.

• Enhance and promote the use of the EO platforms capabilities for regional-scale processing, data fusion and information delivery.

These shall be achieved by developing and delivering a customised set of EO-based information services that utilise scalable cloud computing-based processing resources, high-throughput computing capabilities, and fusion of large volumes of EO data from the Sentinel missions and other European EO missions. The implementation shall be based on a regional approach to information collection and delivery. Integration with non-EO data is also foreseen as an essential component, including the opportunity to exploit the potential for innovative service delivery (or new data access, processing and management solutions).

Bark beetle outbreaks are a significant problem in the EU, causing more than 3 billion euros worth of damage to forests each year. Climate change is making this problem worse, as dry and warmer weather conditions are causing the beetles to multiply rapidly. Unfortunately, large forest owners and government bodies are struggling to spot outbreaks on time, and there are no effective prevention measures.

To preserve the value of our forests, we must establish a centralised prediction and alerting system. This system would help government bodies and forest owners react more quickly to outbreaks, which are likely to become more frequent.

Our proposed platform would provide new opportunities for forest owners to be notified of outbreaks using near real-time satellite imagery monitoring. This would streamline the entire forest management pipeline, from regulatory bodies to forest management, forest owners, and the wood processing industry. A centralised system can quickly alert all parties to take action to mitigate the outbreak.

In order to be successful, the system must be able to detect outbreaks as early as possible. This requires advanced technologies such as remote sensing and machine learning. The system must also be able to process large amounts of data quickly and accurately. Additionally, the system must be user-friendly and easy for forest owners and government bodies to access and use.

Implementing a centralised prediction and alerting system for bark beetle outbreaks would be a significant step forward in protecting our forests and preserving their value. It would also provide new opportunities for forest owners to manage their resources more effectively and efficiently. We believe that by working together, we can make a real difference in the fight against bark beetles.

The EU Copernicus programme has established itself globally as the predominant spatial data provider, through the provision of massive streams of high resolution earth observation (EO) data. These data are used in environmental monitoring and climate change applications supporting European policy initiatives, such as the Green Deal and others. To date, there is no single European processing back-end that serves all datasets of interest, and Europe is falling behind international developments in big data analytics and computing. This situation limits the integration of these data in science and monitoring applications, particularly when expanding the applications to regional, continental, and global scales.

The proposed C-SCALE (Copernicus – eoSC AnaLytics Engine) project aims to federate European EO infrastructure services, such as the Copernicus DIAS and others. The federation shall capitalise on the European Open Science Cloud’s (EOSC) capacity and capabilities to support Copernicus research and operations with large and easily accessible European computing environments.

“BACKGROUND: Canopy height is a fundamental geometric tree parameter in supporting sustainable forest management. Apart from the standard height measurement method using LiDAR instruments, other airborne measurement techniques, such as very high-resolution passive airborne imaging, have also shown to provide accurate estimations. However, both methods suffer from high cost and cannot be regularly repeated.

GOAL: In our study, we attempt to substitute airborne measurements with widely available satellite imagery. In addition to spatial and spectral correlations of a single-shot image, we seek to exploit temporal correlations of sequential lower resolution imagery. For this we use a convolutional variant of a recurrent neural network based model for estimating canopy height, based on a temporal sequence of Sentinel-2 images. Our model’s performance using sequential space borne imagery is shown in preliminary results to outperform the compared state-of-the-art methods based on costly airborne single-shot images as well as satellite images.

PREVIOUS WORK [1]: In our previous study, we adopted a neural network architecture to estimate pixel-wise canopy height from cost-effective spaceborne imagery. A deep convolutional encoderdecoder network, based on the SegNet architecture together with skip connections, was trained to embed the multi-spectral pixels of a Sentinel-2 input image to height values via end-to-end learned texture features. Experimental results in a study area of 942 km2 yielded similar or better estimation accuracy resolution in comparison with a method based on costly airborne images as well as with another state-of-the-art deep learning approach based on spaceborne images.”

The JESAC project aims to develop a monitoring system from very high-resolution (VHR) data for at-risk areas and reforestation activities to cover the information gap in semi-arid regions. The monitoring system will detect individual trees, monitor their growth, and determine their increase in biomass over time, which can be translated into their capture of carbon dioxide (CO2) from the atmosphere. This technology would allow for an accurate understanding of such under-monitored areas. The first expected result is automatically detecting tree crowns from VHR imagery. Being able to perform such a task in an automated fashion with a trained model can aid local, regional, or National Forest Inventories in transitioning to a more digitized, less time-consuming protocol. It can also increase the frequency of monitoring, as the sole availability of cloud-free multi-spectral satellite imagery would be sufficient for the model to detect the trees. This technology could also help monitor agroforestry parcels’ daily activities while accounting for their trees’ health and growth. Another expected result is the estimation of carbon stock from each tree. It can be achieved by determining the biomass stored in the tree in the form of leaves, trunk, and roots. Pairing VHR data with carefully designed in-situ measurement campaigns can provide the requirements to calibrate the models to perform the estimation. The technology will then be used to monitor existing reforestation activities, ensure their correct development, and produce carbon offsets based on observations. Finally, vegetation indexes, crown sizes, and evolution of tree growth can provide the health status of individual trees and whole agroforestry parcels or forests.

As a vital vegetation type, trees dramatically contribute to carbon sequestration and mitigating climate change. Australia’s rangelands cover about 80% of the country’s area. Trees in rangelands are essential for both the interannual variability of the carbon cycle and local livelihoods. Therefore, accurately estimating the tree cover in Australia’s rangelands is fundamental for detailed landscape pattern analysis to manage and conserve trees. However, most public interest in trees is devoted to forests, and trees outside of forests are not well-documented, especially in Australia’s rangelands. This project aims to develop and implement a machine learning model to accurately map tree cover in Australian Northern Territory and Queensland rangelands using high-resolution satellite imagery. The outcomes will improve the monitoring of rangeland trees and understanding of their role in mitigating degradation and climate change.

The overall project aim is to access karst caves in remote places in Greenland and sample the speleothems (mineral precipitates on cave walls). From these can be extracted geochemical signals that relates to climatic variations at that location. The image analysis is to be used for planning a field expedition in the summer 2023. The main purpose of the field expedition is to collect speleothems from caves in East Greenland to provide unique data about climate variations in Greenland prior to the time interval covered by data from the Greenland ice-cores (~ 130.000 years). Such data are valuable for calibrating and improving climate models, especially for the Arctic region and the Greenland Ice Sheet (GIS). Traditionally the climate models rely on calibration data from marine sediment cores and ice cores. Recent developments in a suite of techniques (stable isotopes, radiometric dating, etc.) have provided scientists unprecedented opportunities to advance the understanding of mineral deposits in caves (so-called speleothems), although nearly exclusively such studies focused on temperate and subtropical regions (Fairchild & Baker 2012; Wong & Breecker 2015; Comas-Bru et al. 2020). Data from the Arctic are rare and until now only one single example from northern Greenland has been published (Moseley et al. 2021), covering a time-window of ~50,000 years dated at around 550,000 years ago. Additional data from Greenland cave speleothems may provide a better understanding of climate changes through the very important period from 2.5 million to 130,000 years ago including Pleistocene glaciations and interglacials, and will thereby serve to narrow uncertainty in the future predictions of the Arctic climate and the fate of the GIS under the progressively warming world. In order to plan and optimise fieldwork, satellite images will be used for the initial mapping of potential locations for cave entrances. The geochemical analysis data will be made publicly available and the results will be published.

The Arctic+ team intends to develop a new regional Arctic SMOS SSS product (follow-up version, Arctic+ Salinity v4) to enhance two fundamental components for calculating freshwater content in the Arctic, namely:

1. Effective spatial resolution: Algorithms for reducing the incoherent noise in the brightness temperatures will be applied, namely the Nodal Sampling (Gonzalez-Gambau et al., 2016) and the multifractal fusion of the brightness temperatures(Olmedo et al., 2021a).

2. Better characterization of the sea surface salinity dynamics: We plan to mitigate the different errors affecting the SMOS measurements without using salinity values in-depth as a reference for the temporal biases’ correction.

With this approach, we aim at keeping the surface dynamics and not masking it with the sub-surface one.

The European Continental Crop Map is a machine learning-based crop map that contains six crop types (summer cereals, winter cereals, maize, potato, sugar beet and winter rapeseed). Vito developed it using the openEO Platform. The map was created for a year at a spatial resolution of 10x10m. The map is based on Sentinel-1 and Sentinel-2 data, more specifically on time steps and basic statistics (standard deviation, percentiles) for B6, B12, VV, VH, VV/VH ratio, and seven indices (NDVI, NDMI, NDGI, ANIR, NDRE1, NDRE2 and NDRE5). The map is created using a Catboost model trained using GridSearch, using the LPIS dataset for training and testing.

CECOES 1-1-2 is the Emergency and Security Coordination Center of the Autonomous Community of the Canary Islands. Manage urgency and emergency calls to 1-1-2 from citizens by activating firefighters, ambulances, or police. In addition, in a significant emergency, such as a forest fire or flood, it coordinates the response in these emergencies. The provision of satellite images in real-time is of vital importance for decision-making in emergencies. For example, this information was used in the volcanic eruption on La Palma island just one year ago. The CECOES 1-1-2 has a GIS viewer to collect all the georeferenced information for emergency management. The main emergencies that are managed from the CECOES 1-1-2 are:

• Forest fires

• Floods

• Marine contamination

• Chemical risk

• Transport of dangerous goods by road

• Volcanic and seismic risk

The Civil Protection authorities require up-to-date information for decision­making. Therefore having access to satellite information, fire risk index, etc., has been very useful in the last emergency on La Palma island. However, having this information available in an emergency is not easy. It is necessary to download and process it for decision-making, which is why access to the latest available information is required. In addition, this entire process requires trained personnel with a high degree of specialization. During an emergency, all the operational procedures must be in place so that access to GIS information is easy.

The urgency to develop methods capable of identifying specific drivers of forest disturbance events is highlighted in the UN REDD+ policy. Characterizing drivers is essential to understand the complex socioeconomic processes that cause forest loss. However, charcoal production across Sub-Saharan Africa is ineffectively monitored and regulated. This contributes to the uncertainties surrounding the ecological impact of the industry and makes it difficult to separate the drivers of forest degradation in the region. In addition, this limits our ability to grasp the effects on local processes and the shifting ecosystem dynamics. High spatiotemporal systematic observations of the Copernicus Sentinel-1 (S-1) synthetic aperture radar (SAR), with the intrinsic advantages of radar imagers, make it one of the most applicable sensors for detecting small-scale forest disturbances.

Furthermore, AI and cloud computing on EO Platforms (e.g., the Euro Data Cube) enable scalable exploration of deep stacks of SAR data at regional to continental scale. In this study, we demonstrate the potential for using S-1 SAR and other geospatial datasets with the help of AI to produce a methodology for scalable forest degradation monitoring of specific drivers in Sub-Saharan in an Open Science development framework. This will enable more effective management and protection of valuable woodlands and better inform policymakers on the extent of charcoal production across the region. Furthermore, it will allow one to understand better the shifting dynamics of these woodlands and their vulnerability to the changing hydrometeorological conditions within SSA. The second purpose the request will be used for is to prepare the tutorial notebooks in EDC for the RACE/EO Dashboard demo area at LPS 2022.

This project aims to study image segmentation and classification for pattern recognition of forests, non-forests, and agricultural areas in the State of Pará (Brazil), including implementing Artificial Intelligence tools to assist in describing existing areas. The Centro de Inteligência Territorial (CIT) is an independent, non-profit organization with researchers specializing in land use modeling and public policy evaluation. CIT is a hub for Science and technology projects, connecting researchers, policymakers, decision-makers, and experiences in territorial intelligence. Reconciling production needs, ecological restoration, and social progress in a territory are challenging. Working at the frontier of knowledge is essential to face this and other challenges.

CMIX II is the second edition of the joined ESA and NASA Cloud Mask Intercomparison eXercise activity in the frame of CEOS WGCV. It is an international collaborative effort aimed at intercomparing cloud detection algorithms for moderate-spatial resolution (10-30 m) spaceborne optical sensors. The focus of CMIX is on open and free imagery acquired by the Landsat 8 (NASA/USGS) and Sentinel-2 (ESA) missions.

Within the second addition of CMIX, dedicated reference datasets will be cerated to validate the participating cloud masking algorithms. One of these datasets is an expert pixel collection conducted on Sentinel-2 L1C and Landsat 8/9 Level1 data. Together with the participants, it was decided to provide information on cloud optical depth (COD) in addition to the expert classification, to have a numerical reference on different cloud transparency classes. To derive COD for any Sentinel-2 L1C or Landsat 8/9 L1 product, a surface reference is required. This reference can be a surface albedo or anything comparable.

In context of CMIX II an approach was developed using all S2 L2A data within an 18 day window of all years since the S2 launch, do derive a longterm average. The same is done for L8 L2 data. For this approach only the read and NIR band of the two sensors is needed, as well as cloud masking bands. The resulting product will called Land Surface Reflectance (LSR). The LSR can be used as a reference to estimate the COD. The reference dataset including the COD estimates will be published at the end of CMIX II. The reference dataset will comprise approx. 100 Sentinel-2 and 100 Landsat 8 products, the expert pixel collection and the COD estimates for all collected pixels.

The objectives of the project include the feasibility study on the use of VHR optical data for coastal erosion studies and the production of coastal erosion rates from VHR optical data for selected areas along the north Dublin coastline. The results will be shared over the GSI web mapping services for free as an example of the use of VHR to monitor coastal erosion.

The goal of the project is the measure coastal erosion/shoreline change rates along the County Wicklow coastline in Ireland. The results of the project will allow us to give an up to date, accurate, and relevant synopsis of how the soft sediment coastline of Wicklow in the east of Ireland has changed over the last

two decades and what its current state of erosion/accretion is. Using orthophoto data collected in Ireland since 2000 it has been possible to gain an understanding on shoreline change over the timeframe 2000-2012, however the quality of data has improved significantly in recent times, which is a strong opportunity to understand more recent changes to Irelands coastline.

As a results, we are trying to access VHR commercial satellite data to digitize shorelines ( e.g vegetation line) between 2012-2022 and compare the results with the orthophotos. The coastal area in question is composed of three main environments exposed bedrock, lowland beach or marshland areas, and soft sediment cliffs, with some areas classified as Special Areas of Conservation by Irelands National Parks and Wildlife Service. As climate change and rising sea levels begins to take effect in the coming decades it is important to have a good baseline understanding of the fluctuation of shorelines, especially those in low lying areas that are vulnerable to coastal erosion or habitat loss. This stretch of coastline contains several significant urban areas which represents a significant anthropogenic influence on the project, as shoreline change/coastal erosion rates can help influence informed decision making along the Irish coastline with respect to This project can feed into ongoing coastal vulnerability and coastal erosion projects occuring in Ireland and throughout Europe.

In this project, it is proposed to create a high resolution (<10m) coastal typology of the European coastline, which distinguishes land use / cover classes relevant to coastal flooding and erosion. During this sponsorship we will develop a methodology to classify the satellite imagery. Upon success we will scale this to the whole European coastline.

This project aims to research various implementations of machine learning algorithms in monitoring coastal waters and understand the potential implications of this research. The goal is to combine highly abundant remote sensing data with in-situ sensor data to monitor and predict water quality. Water quality measurements are used to determine the health of local ecosystems for wildlife preservation and food production, which are at risk due to harmful algae blooms (HABs). A trained machine learning solution can resist noise and incomplete data, often during a natural disaster event. The area of interest for this study is Biscayne Bay in South Florida due to ease of access to the site, the collected in-situ data, and the remote sensing data to be used from public online web services. A Python program is developed, and processes gathered in-situ data with remote sensing data from Sentinel Hub. The data is statistically analyzed, plotted, prepared, and used to train a machine-learning model. The model is cross-validated and performs to a certain degree. Recent literature investigation indicates several approaches for water quality measurement and estimation, many of which do not rely on a combined remote sensing and in-situ sensor data set. For example, certain developments use strictly in-situ sensor data or combine satellite remote sensing data with drone remote sensing data. Further investigation is necessary to improve the accuracy of the developed model; this includes a better selection of spectral satellite image source and bands, outlier and missing data handling, cross-validation parameters, and choice of machine learning modeling algorithms.

The coastal zone and its shorelines are potentially affected by sea level rise in the changing climate. However, shoreline changes are affected not only by absolute sea level rise but also by morphological changes and vertical land motion. So far, the individual contributions of these groups of shoreline changing processes are unclear. This thesis aims to separate these processes by quantifying their effects on shoreline changes. This project will use observations of retracted coastal radar altimetry, as applied here, and compare them with shoreline changes from optical remote sensing observations. Complementary data sets like tide gauges and GNSS observations will also be employed. The goal is to produce a time-variable shoreline attributed to sea level rise and morphological changes. This is initially done for a focus region (Terschelling, the Netherlands), but the methods will ultimately be applied worldwide.

Furthermore, it is planned to build an empirical model from these observations to predict future shoreline migration. It is expected that the results of this project will not only advance our observation-based understanding of coastal changes and the participating processes but will also serve as an essential observational input for coastal planning in light of climate change. Data and software will be publicly released on shared platforms such as Zenodo, GitHub and the Dutch Data Archiving and Networking Services (DANS) in line with the open science policies of the ITC Faculty of Geo-information Science and Earth Observation of the University of Twente.

More than half of the U.S. population resides on or within 50 miles of the coast, even though coastal zones constitute only 18% of the total U.S. land area. The combined effects of natural and anthropogenic activities/processes alter the morphology of these land surfaces, increasing the threat of steady inundation from SLR and the possibility of sudden and abrupt flooding and erosion emanating from storm surges/high tides. Even outside the coastal environments, largely anthropogenic activity-driven surface deformation processes are gravely endangering human lives and infrastructure. The proposed study area, Southern United States and portions of the (north and east) Gulf of Mexico coast, despite being largely tectonically stable, is experiencing subtle surface deformation and change mainly attributed to human activity-driven (anthropogenic) processes and a lesser degree due to glacial isostatic adjustment processes. With the documented increasing recurrence and intensity of natural disasters mainly due to anthropogenic-led alterations to the environment and climate change, an integrated research approach based on various datasets and novel techniques would be beneficial for monitoring the occurrences and impacts as inducing processes that initiated their circumstances. The proposed study aims to quantify surface deformation processes using fused satellite- and ground-based datasets and generate a complete deformation field of the study area. The temporal deformation patterns will be assessed to detect precursory hazard indicators crucial for developing hazard early warning systems. In addition, the factors and processes that directly or indirectly contribute to the occurrence of the hazards will be determined. Who will benefit from the project results: Communities, policymakers.

Previous research in a GIS Programming course requested teams of 2-3 students to develop a state-of-the-practice QGIS plugin (Anbaro#lu 2021). Consequently, students relied on Git to collaborate with each other while developing their plugins, did unit testing, provided language support and documented their plugins using Sphinx. Although, students learned valuable technical and practical skills, in order to have a critical spatial data science perspective, more theory should be integrated into teaching (Holler 2019, Kedron et al 2020). Therefore, the objectives of this experiment is to investigate how students utilise an open-source Python package, x2Polygons, to find the distance between georeferenced polygons. For this each student will digitise a number of polygons, with varying complexity – in terms of the number of edges each building possess and evaluate how different distance measures such as the Hausdorff distance; Chamfer distance, PoLiS distance (Avbelj et al 2021) and turn function distance correlate with each other. In this way, they will be able to assess the advantages and limitations of different distance measures.

The project’s main objective is to use the IaaS service provided by the CREODIAS environment to classify crops and predict yields based on satellite and meteorological data available in the EO data repository. The secondary objective is to provide the generated results to the ESA WorldCereal project and the EO4UA initiative.

This project aims to study crops and their impacts in the U.S. Midwest. The corn and soybean row crop system in the U.S. Midwest, contributing to one-third of the world’s production, faces grand environmental challenges related to excessive use of fertilization, soil carbon loss, and water quality degradation. Understanding the historical and present crops and their impacts on crop yields is crucial for global food security. The accurate estimates of current and historical crop acreages are essential for understanding crop adoption status, evaluating the outcomes of incentive programs, and designing effective agricultural management. Multiple sensor datasets, including ESA’s Sentinels and NASA’s Landsat, are used for high-resolution and long-term crop mapping in the U.S. Midwest. Detailed crop fields with crop information will be generated for the whole U.S. Midwest, which is essential for agricultural stakeholders. The spatial and temporal patterns and trends of crop maps provide crucial details for policy-makers and sustainable agriculture, which further secure crop yields in this region. High-resolution crop maps at the field level are lacking for the whole U.S. Midwest. Thus, this project can serve as the benchmark for future crop mapping. The developed crop detection algorithms are scalable to regions with similar settings and can be performed locally and globally. The science foundation integrates the knowledge of crop plant physiology and remote sensing of different targets, gaining insights into agricultural remote sensing and laying solid foundations for other research.

The project aims to provide a service based on satellite and weather data to satisfy farmers’ needs. Several segments of the market can benefit from this service. Mainly and firstly, the target is farmers. Farmers can use this service via web-based or mobile applications and a lot of helpful information about their farmland and make more optimized decisions that use pesticides and similar inputs. In this way, not only does the farmer benefit because of lower consumption of such inputs(which will pay off the cost of the service), but they also will experience a higher crop performance. The second segment interested in the product is the insurance service providers. We can provide precious information based on satellite image analysis to them. Furthermore, we can help them to handle claims because we know what has happened to the farmland.

The main objective of this project is to study the complementarity of spatial optical imaging, structural information from Synthetic Apertuge Radar (SAR) and environmental characterization data to model maize and sunflower seed production by aggregating these observations of different spatial and temporal resolutions. The thesis work will be based on Syngenta’s plot network in several parts of Europe and North America, where some varieties of maize and sunflower are evaluated under different environmental conditions.

The Feature Engineering of satellite observations used in the development of machine learning models will seek to estimate varietal parameters-functional extracts-which define, on the one hand, the phenology, but also the different varieties’ response to abiotic stress, and mainly to water stress. In the case of phenology, we will study the parameters that determine the development response to temperature and photoperiod, and that predict specific crop stages. The functional traits that characterize the response to abiotic stresses will make it possible to identify, from multi-environment observations, the most efficient varieties, and to predict their behavior and yield.

The objective of the project is to leverage the high-resolution and multi-spectral data provided by Sentinel-2 to create detailed maps of crop distribution, which can help farmers make informed decisions about land use, crop management, and food security, which will enhance and lift the lives for many poor farmers. Furthermore, we want to unite the farmers and educate them regarding crop management and food security. We currently have a community of 50+ farmers, and we plan to make decisions on sessional crops, land decisions, soil management, and food security. Also, this will help us in the future in many ways.

Accurately determining crop growth progress and crop yields at the field scale can help farmers estimate their net profit and enable insurance companies to ascertain payouts, ultimately bolstering food security. At field scales, the trifecta of management practices, soil health, and weather conditions combine to impact crop growth progress, and this progress can be monitored in-season using satellite data. The project aims at creating field-scale results that will be made available to farmers and regional-scale results that will be available to policymakers via the NASA Harvest website and relevant peer-reviewed publications.

This project aims to develop methods for satellite-, model- and AI-based yield forecasting of crops in the context of Austrian agriculture. To achieve the project goal, the following technologies and methods will be combined or further developed:

• Sentinel 2 spectral data (satellite imagery) will be used as input for the AI component, the PROSAIL reflectance model and the iCrop growth model.

• The AI component will be trained to detect crops using manually labelled and existing training data from the Austrian and US regions. The trained model should be able to correctly classify the crops visible on the satellite images by 90%.

• With the PROSAIL reflectance model, possible reflectance values are calculated by discrete variations of the input values and stored in a database. The reflectance values seen in the satellite images are then matched with the values in the database to infer possible input value combinations and crops.

• The results of the AI component and the reflectance model are passed to the fuzzy logic classifier to determine the crop finally.

• The detected crop type and growth parameters (e.g. Leaf Area Index) derivable from the spectral data will be used to calibrate the iCrop growth model for yield prediction (as well as harvest timing, phenology, fertilizer and water requirements).

The developed methods should be able to determine the crop species on overgrown Austrian field pieces based on satellite data with about 90% accuracy and generate yield forecasts based on these data. For the Austrian area, such applications do not exist at present. In the international context, there are applications for detecting agricultural crops based on satellite imagery, but no approaches use this as a basis for calculating yield forecasts with growth models. The added value for the addressed user groups is the scalable generation of yield forecasts based on daily updated satellite data.

Many strong active faults have developed within the Tibet Plateau, Tienshan and its adjacent regions, forming multiple seismic zones due to the collision and continuous extrusion wedging between the Indian and Eurasia plates. Those faults directly control the spatial distribution of severe disaster zones in mega-seismic areas. But the lack of quantified descriptions of geology and geodesy in this area makes it very limited to understand its geophysical environment and rupture process of active faults. This study focuses on strong active fault zones in Western China, especially in the central Qinghai-Tibet Plateau and Tienshan region. GPS and InSAR will be used to monitor the crustal deformation and to derive an accurate 3D velocity map of the area. We expect to densify the existing GPS network, form several profiles across those active faults, and then integrate the GPS and InSAR measurements to derive the velocity maps and geometry of different segments of the faults, cooperating with geology and geophysics data. It can elaborate the advantages of two means and get the fine fault monitoring and structure analysis to reveal the graben deformation characteristics, tectonics, deformation pattern and evolution mechanism of the studied regions. Finally, we will inverse the lock depth of different segments and investigate the relationship between crustal deformation features and strong earthquakes and the relationship between the lock depth and deep structure. It will improve the ability of earthquake forecast by revealing the kinematics pattern and dynamics and dynamics background of the continent deformation in China and its adjacent regions.

The Cryosphere Virtual Laboratory (CVL) project will develop, test and demonstrate a prototype community open science tool where ΕΟ satellite data and derived products can be accessed, visualised, processed, shared and validated. In addition, the tool will provide access and facilitate sharing of relevant space and nonspace data (aerial, UAV, coastal radar, in-situ etc.). Following an Open Science approach, the CVL will mainly be designed to support scientists in accessing and sharing ΕΟ data, high-level products, in-situ data, and open­ source code (algorithms, models) to carry out scientific studies and projects, sharing results, knowledge and resources.

The project’s main objective will be to demonstrate that it is possible to precisely monitor forest biomass to validate and certify La Belle Foret’s methodology to generate carbon credit thanks to high-resolution satellite data. More specifically, we want to carry out some ground measurements (LiDAR or manual forestry tree counting process) and combine them with satellite imagery to precisely estimate the aboveground biomass with only a few ground samplings using the allometric models that we developed in-house.

We are currently working with the french ministry of ecological transition to generate carbon credits certified by the French Label Bas Carbone, and we are laureate of the ΒΡΙ France Deep Tech sponsorship, which rewards our effort to set up deep tech solutions to finance the protection of the french forest.

We will mainly use Pleiade and Pleiade Neo Airbus data for this study (50 cm and 30 cm resolution over six spectral bands).

Since every methodology certified by Label Bas Carbone is made publicly available, the main results of this project may be publicly disclosed if it appears that satellite certification can take part in the overall certification process by the French Label Bas Carbone.

The project aims to monitor the cultural heritage sites in Libya, particularly the UNESCO WH sites that are facing many human and natural hazards and threats nowadays.

Danube Data Cube (DDC) is a regional data exploitation platform built on and follows the logic of the Euro Data Cube (EDC) infrastructure, a computational environment reflecting the Digital Twin Earth concept of the European Space Agency to support sustainable development. DDC is a cloud-based platform with data and analysis tools focusing on the Danube Basin. As a regional platform service, it demonstrates the data cube technology’s data storage and analysis capabilities, maximizing the benefit of the synergy of satellite and ancillary data with dedicated analysis tools. The DDC concept includes extensive Machine Learning capabilities, including analytical tasks and decision support algorithms. One of the key themes of the platform is water management, from regional strategy and public information to field-level irrigation management.

Currently, DDC works on a regional and a local (field-level) showcase. Both are related to water management.

This project is the second phase of the Danube Data Cube.

DDC is a regional data exploitation platform built on and follows the logic of the Euro Data Cube (EDC) infrastructure, a computational environment reflecting the Digital Twin Earth concept of the European Space Agency to support sustainable development. DDC is a cloud-based platform with data and analysis tools focusing on the Danube Basin. As a regional platform service, it demonstrates the data cube technology’s data storage and analysis capabilities, maximizing the benefit of the synergy of satellite and ancillary data with dedicated analysis tools. The DDC concept includes extensive Machine Learning capabilities, including analytical tasks and decision support algorithms. One of the key themes of the platform is water management, from regional strategy and public information to field-level irrigation management. Currently, DDC works on a regional and a local (field-level) showcase. Both are related to water management.

Among the renewable energy sources, solar and wind are rapidly becoming popular for being inexhaustible, clean, and dependable. Meanwhile, power conversion efficiency for renewable energy has improved with great technological leaps. Following these trends, solar and wind will become more affordable in years to come and considerable investments are to be expected. As solar and wind plants are characterized by their high site flexibility, the site selection procedure is a crucial factor for their efficiency and financial viability. Many aspects affect site selection, amongst them: legal, environmental, technical, and financial. Today, information gathering for site selection assessments is a manual and time-consuming process. The main objective of this project is to develop a dataset of existing solar power plants* by applying computer vision on satellite imagery.

Objective 1 (O1): Achieve 90 % accuracy for a specific data layer* using SOTA deep learning models.

Research Question 1: Can the required accuracy be achieved with publicly available 10×10 meter

image resolution, or must higher resolution imagery be used?

Research Question 2: How can the training-data creation process be made for the specific data layer to

achieve the required accuracy?

Research Question 3: How can the training of the model be made for the specific data layer to achieve

the required accuracy?

Since the project is researching the process of site selection and the utilization of data for renewable

energy projects, the idea contributes to positively influence the UN’s Sustainable Development Goal 7;

Clean energy for everyone. For the region to be able to produce enough clean energy, it is necessary to

accelerate the development of renewable energy projects.

The project aims to develop innovative extensions for Jupyter and Visual Studio Code to allow the launch of remote Jupyter Kernels in the cloud. Furthermore, as part of our application, we want to demonstrate Proof of Concept of our offering.

We will also look at the security aspects (how to authenticate an external Jupyter server toward the EODC and how the fine-grained access rights to the datasets are implemented).

The objectives of the project are:

1. Assessment of decadal Mass Balance and Ice Thickness of glaciers of Sikkim Himalaya (Study

Area 1200 Sq. Km)

2. Estimation of Glacier Facies using GLCM and Random Forest Classification.

3. Movement analysis of glacier in Sikkim using PSI.

Statement on availability of results.

The results shall be available online through any suitable data sharing portal on request for research purpose only. It shall also be shared with nodal agency for sub-regional level policy response. Subregional policy requires concrete evidence especially in developing Nations like India.

The project objective is to create the decade data set for the land use map for India based on the NRSC classification scheme for 1995,2005,2015. Under this study, a methodology based on the CNN technique, which uses the yearly seasonal pattern to identify the LULC, will be integrated with the spectral response. The multi-temporal classification will lead to the level 3 classification data by applying the hierarchal classification technique. The decadal LULC will be analyzed for the temporal variability observed in different classes. The classified data will be used with the LULC prediction models to provide future integrated scenarios. The different LULCs will be integrated to give the weather models to analyze climatic variability due to LULC changes.

The objective of this project is two-fold and the requested data can be used for both tasks while testing processing capabilities on The Food Security Platform (TEP). Firstly, we will demonstrate that it’s possible to do bottom-of-atmosphere (BoA) correction on Sentinel-2 Reflectance at Bottom of Atmosphere/VISTA Algorithm (available on TEP as ‘Vista_S2-L2A) using deep neural networks. We estimate that this can improve processing speeds by x100 to x500 while keeping accuracy high. This is inspired by an existing algorithm, developed for another project, that in production as a side effect efficiently fixed incorrect Sen2Cor bottom of atmosphere correction. This is in particular interesting on, important and frequently used algorithms with long processing times like BoA processing algorithms. The results will be avg. pixel error measured against ground truth imagery. We will also present the relevant processing speeds improvements and requirements to run said algorithm (eg. GPU accelerated processing). The results will be made available on TEP as ClearSky Vision demo data, and if possible produced on TEP. It will, furthermore, be measured against data in-sample and out-of-sample, and the project will be finished off by producing a tile unavailable on the platform. This project has the potential capability of greatly reducing required resources for BoA correction on Sentinel-2 imagery by doing it in

a fraction of the time (leaving data storage as the final limitation). Not only making it a fast and efficient process but it also makes near-real time monitoring more achievable.

ClearSky Vision has already developed an algorithm for cloudless Sen2Cor imagery (using deep learning and multiple satellites for data fusion). This approach won ClearSky Vision the Copernicus Masters Bay Wa competition in 2020. It combines Sentinel-1, Sentinel-2, Sentinel-3, and Landsat 8 into daily cloudless Sentinel-2 imagery. This project will further prove, to what degree cloudless results on Vista S2-L2A will match the accuracy from prior cloudless Sen2Cor imagery tests. The objective is to determine whether this (more complex) processing method will make the cloudless process more difficult or what’s more likely, improve the consistency of the output. The results will be made available on The Food Security Platform as ClearSky Vision demo data (10 spectral bands).

Urban expansion is giving rise to new challenges globally, especially in African countries badly affected by climate change, population, and, most importantly, economic growth. Government agencies must estimate cities’ growth, thus enabling better urban planning to meet challenges. Machine learning and Computer Vision techniques can allow government agencies to generate models which can control Urban expansion beforehand. So, this research focuses on using satellite images to tackle the Urban expansion of certain areas using deep learning techniques. For Urban expansion, I selected the area of Dakar, Senegal, one of the Seaports on the Western Coast of Africa. Dakar region also suffers from various development issues associated with environmental deterioration, such as the decrease of green areas, farmlands, and wetlands. Therefore, economic activities suffer from these problems. This research aims to provide a deep learning model which can predict Dakar’s urban expansion so the state can plan the land transformation and economic growth accordingly. Moreover, this project will also help all the sentinel hub users who want to work on satellite images or multi-temporal data to solve Urban expansion-related problems. This research will help them create a pipeline for using satellite images to develop a deep-learning model to predict the urban expansion of their desired area.

DeepESDL users or teams will be provided with individual subscriptions for external services to ensure that dedicated resources are available to them. The requested subscription is required for the first set of Early Adopters, which are currently onboarded, and their associated use cases as well as for the DeepESDL consortium to integrate the Sentinel Hub service, demonstrate and validate ts functionality and for using it in training sessions for new users.

The project aims at creating a Landcover semantic segmentation model to identify the changes in resources of Sri Lanka, such as forest cover.

The project will use differential interferometric Synthetic Aperture Radar techniques (DInSAR) to measure land deformation caused by earthquakes and land subsidence. Geological instabilities could cause the differential movement of ground at different depths. This phenomenon is a gradual settlement of soil that causes inundation of land, expansion of flooding areas, disturbance of drainage systems, changes in slopes, and damages to infrastructure foundations in urban areas. This study will also analyze the deformation’s behavior and cause of subsidence within the research area. Moreover, the study will try to understand the necessary connections and interactions between people and natural events to prevent or lessen the extensive social, economic, environmental, and infrastructure effects. Supporting information requires accurate and timely change detection on Earth’s surface to make better decisions about land deformation and the event’s temporal ramifications. To enhance the measurement of small-scale surface deformation using SAR Interferogram. The information, including detailed coseismic deformation based on the study’s interferometric results, will be helpful in the community’s disaster management and mitigation activities. The results will include the land deformation map and the Subsidence map of the study area. The aim is to provide information to urban planning and management authorities. Finally, the project will be used to gain expertise in using the Geohazard platform.

The access to Earth Console is made in the frame of Proposal DETECT-B01, which is part of the Collaborative Research (CRC) 1502 of DFG (https://www.lf.uni-bonn.de/en/research/crc-detect). The main goal of DEECT-B01 is to estimate river discharge and water storage change from space using satellite altimetry. The central hypothesis of DETECT-B01 is that the new generation of space-borne altimeters, including Delay Doppler(DD), laser and bistatic SAR altimeter techniques, outperform conventional altimetry(CA) and in-situ measurements. They provide surface water levels and discharge of higher accuracy and spatial and temporal resolution thanks to the new river slope and width parameters. The better sampling will improve flood event detection and long-term evolution estimation, providing valuable further information to the overall CRC. In the first four years of its 12 years possible duration, two research questions have been addressed in the CRC:

1. How can we fully exploit the new missions to derive water level, discharge, and hydrodynamic river processes?

2. Can we separate natural variability from human water use?

Over the last decade, with SAR altimetry data, merging innovative space observations with in-situ data provides a denser and more accurate two-dimensional observational field in space and time compared to the previous two decades. This process allows better monitoring of water use’s impact and characterising climate change. River discharge and water storage change innovatively derived from the nadir and swath-altimetric measurements of river slope, height, and width will enable to validate the modelling (e.g. through budget studies) and will be used for assimilation in the IMS. A multi-sensor database will be built starting from 1993 and used in the Integrated Monitoring System (IMS) from other partners in the CRC. B01 will also monitor the exchange of water between rivers, lakes and reservoirs and the impact of natural and human disturbances, like water use. The project will contribute to the CRC’s key objectives in that it addresses the surface water compartments by improving new observation types and including them in the modelling.

Object detection is one of the mandatory steps in transferring imagery data into land cover information. Deep machine learning networks have shown automatic object detection capabilities and generated promising results. The patch-based Deep Neural Network (DNN) is one of the architectures designed for pixel-based object detection in aerial images.

Road extraction from remote sensing images is significant to urban planning, navigation, disaster assessment, and other applications. Although deep neural networks have shown a strong ability in road extraction, it remains challenging due to complex circumstances and factors such as occlusion.

Road extraction from remote sensing images is significant for updating geographic information systems (GIS), urban planning, navigation, and disaster assessment. In the past, the most widely used way to extract roads was through manual vision interpretation, which takes a lot of time and has a high labour cost, and the extracted results may vary due to the differences of interpreters. Automatic road extraction technology can improve the efficiency of road extraction, so it has become a hot issue in this field. This deep learning model is used to extract roads from high-resolution satellite imagery.

Road layers are useful in preparing base maps and analysis for urban planning and development, change detection, infrastructure planning, and various other applications.

Digitizing roads from imagery is a time-consuming task and is commonly done by digitizing features manually. However, deep learning models are highly capable of learning these complex semantics and can produce superior results. Furthermore, deep learning models can automate this process and reduce the time and effort required for acquiring road layers. My project is to train a model that can detect and build road networks anywhere in Egypt.

In the first stage as an early adopter user of GEP, several previously unknown landslides have been identified in the escarpments of the main faults of the Córdoba ranges, coincidentally in sectors where there is clear evidence of neotectonic activity. These landslides have been recognised by remote sensing techniques, geomorphometric analysis and field surveys, but they have not yet been characterised nor quantified in their rate of movement and speed. Results were exposed in the ARGENCON 2020 workshop, held in December 2020 in Argentina. This proposal aims to continue analysing gravitational processes in Sierras of Cordoba based on their geomorphometric parameterisation, with the estimation of its displacement obtained with DinSAR techniques. Quantifying the local relief through geomorphometric parameters has been done in combination with measures obtained after using the P-SBAS (Parallel Small BAseline Subset) algorithm through the services of the Geohazards Exploitation Platform (GEP). Displacement maps generated by this technique allowed the detection of active processes not previously registered. The sections of Sierra Grande and Sierra Chica fault scarps, which limit the San Alberto and Punilla valleys, respectively, and the Cerro Uritorco slopes, are the areas with the most significant evidence of displacement. Creeping, debris flow, collapses, and rock avalanches were recognised there. Results obtained via Early Adopter Program demonstrate that methods based on DinSAR can reveal morphologic features that otherwise could not be disclosed. In addition, it verifies that platforms based on cloud services that can process large volumes of data are beneficial for identifying and monitoring dynamic geomorphological processes and obtaining predictive information on areas with the potential to slide. The in-phase information provided by SAR images through a multitemporal analysis efficiently detects and evaluates possible new mass removal processes that are taking place or have taken place in recent years.

Agriculture is a vital sector in the West African economy, providing sustenance and income to millions of people. The timing of crop sowing is crucial in determining crop yield and quality. It is influenced by various factors such as weather conditions, soil moisture, and land preparation practices. Farmers’ practices determine the sowing date, and social constructs, such as traditions and beliefs, influence these practices. The project aims to produce country-wide maps for various phenological metrics using remote-sensed crop vegetation dynamics in West Africa. Notably, this study seeks to create multi-year sowing date estimation maps that will be valuable resources for understanding the spatial variability in sowing date strategies among different regions in West Africa. This approach will enable researchers to examine how environmental and social factors influence farmers’ sowing date decisions, leading to improved crop yield and quality and better management of West African agricultural systems. These maps will also be used as input layers in spatialized crop simulation models, contributing to the analysis of the impact of different factors, such as changing climate, genotypes, and agricultural practices, on crop productivity. As such, the study will provide valuable insights into how farmers can optimize their crop-sowing practices to achieve maximum yield. Time-series analysis of medium resolution optical remote sensing products will be performed to conduct this study. This analysis will target croplands detected from land cover/land use (LULC) products generated annually by stakeholders, such as ESA WorldCover. The project’s outcomes will be helpful for policymakers, agricultural extension workers, and farmers alike. By understanding the spatial variability in sowing date strategies among different regions in West Africa, stakeholders can tailor agricultural interventions and policies to the specific needs of different regions. By examining the impact of changing sowing dates on crop productivity, stakeholders can develop targeted strategies to enhance crop yield and quality.

This proposal intends to exploit the automated and unsupervised IREA-CNR N-SBAS processing tool integrated within the (GEP), to generate an up-to-date crustal deformation map of the country of Algeria by the mean of Sentinel-1 SAR data. The velocity maps will be generated for both ascending and descending passes so it will be possible to get the 2-D velocities (east-west and up-down) and resampled to 200 meters. The final results we will propose will be in the InSAR reference frame and ITRF.

The project, under the supervision of Prof. Jurgen Kusche, the head of APMG Institute of Geodesy and Geoinformation, University of Bonn, aims to determine the marine geoids of the West African coast using sentinel-3 data. To achieve the above aim, the following objectives will be used:

• To compute the Mean sea surface (MSS) in the West Africa region using sentinel-3, with SAR closer to the coast than before. This will aid the marine geoid in the region, where existing geoid models in West Africa are decades old.

• To compute Mean Sea Level (MSL) from the tide gauge.

• To determine an existing geoid model from the ICEGEM webpage.

• To calculate Mean Dynamic Topography (MDT) from Existing solutions based on stages 1 to 3.

• To compute Marine geoid by subtracting MDT from MSS.

Our project aims at developing a website that provides information about the impact of the Coronavirus and economic and social factors. First, starting from diverse data, we will investigate the global implications and the effects of the virus on economic and social life. After that, we will talk about rice and how it was affected by the weather conditions, investigating the Mekong River Basin.

The USeEO project ‘Development and verification of custom EO tools for resilience management in Poland’ addresses the need for resilience building by providing value-added satellite-based crisis information and establishing efficient and operational data flow lines between EO tool providers and decision-makers. It aims to develop and validate a set of customized EO-derived information products to support different stakeholders working in the resilience sector in Poland and verify the utility and benefits resulting from using these products. The customized EO-based products will be prepared based on High-Resolution Sentinel-1 and Sentinel-2 data. Also, it is planned to use Very-High-Resolution data to answer crises in which HR satellite data are not enough to fulfill the specific end-users’ needs. The solution is dedicated to the Government Centre for Security and the Regional (voivodeship) Management Centre in Rzeszów.

GEO Mountains (https://www.geomountains.org/) is an initiative of the Group on Earth Observations (GEO). Mountainous regions provide numerous goods and services to both highland and lowland populations globally. However, climatic and environmental changes, large-scale political and socio-economic transformations, and the unsustainable management of natural resources threaten this increasingly. Decisions on policy and investment, from the level of local governments to international agencies, must be based on knowledge that reflects both the generalities and specificities of mountainous regions. The paucity of observations from highelevation regions and associated major gaps in the understanding of mountainous systems thus represent key challenges that must be overcome. In October, GEO mountains released amajor iteration (v2) of the Inventory of In Situ Observational Infrastructure. This update includes many more researchoriented mountain observatories, operational stations, and locations where longterm monitoring is being undertaken. Looking ahead, GEO Mountains will consider providing data storage and linking for those sites that are not able to make their data available in an open repository otherwise. Also capturing extensive metadata for each site to facilitate a comprehensive, interdisciplinary “gap analysis” of in situ mountain observations (i.e. for many variables and with respect to geography, time, and elevation). The project will use the GEP services to develop improved inventories of past avalanches and landslides in remote mountain regions of the world, including the Andes, HKH, Central Asia, and East Africa.

ARTEMIS aims to develop a multi-modal service for processing satellite, terrestrial and available spatial data and the generation of products related to the quality, health and sustainable development of economic forests, with emphasis on chestnut forests. These products will be distributed through a dynamic and user-friendly online platform, which will support operations to facilitate monitoring and improvement of chestnut production and enhance actions for biodiversity protection against climate change. It is known that the Mediterranean chestnut forests in the region of Thessaly have been “degraded” despite being considered productive forests. Moreover, the long-term lack of planning for alternative crops and the insufficient policies for supporting mountain populations’ economic growth has hindered the production of chestnuts, especially in the forests of Mouzaki. Therefore, there is a need to develop modern practices and technologies that will support the continuous monitoring of natural and managed ecosystems and promote, in the long term, the growth of primary production while preserving biodiversity. The project will mainly address the forest health threats in selected areas, mainly caused by biotic factors (insects, diseases, etc.), thus resulting in gradual degradation and destruction of production. As many studies focus primarily on assessing damage driven by abiotic agents (fires, droughts) in forests, it is worth investigating and proposing solutions for the timely evaluation and management of early symptoms of decline, as well as the mitigation of further damage.

Objectives of this project are: • Develop and validate methods for determining the stage and advancement of desertification via diffuse reflectance spectroscopy in the MIR aiming at obtaining prediction models for chemical and physical attributes in soils under intense degradation process. • Build a spectral library using the wavelengths observed in soil samples from the region, highlighting the distinction between the spectra observed in desertification area soil samples; • Understand the link between spectral attributes and chemical and physical attributes of the studied soils; • Prepare maps of the spatial variability of soil attributes, using the results obtained from analyzes carried out in the laboratory (measured values) and obtained by sensors (predicted values) in the study area. • Create land use and land cover maps using high-definition satellite imagery data provided by Sentinelhub.

The vision of the DE Africa is to provide a routine, reliable and operational service, using Earth observation to deliver decision-ready products enabling policy makers, scientists, the private sector and civil society to address social, environmental and economic changes on the continent and develop an ecosystem for innovation across sectors. The mission of the DE Africa is to process openly accessible and freely available data to produce quality products. Working closely with the AfriGEO community, DE Africa will be responsive to the information needs, challenges and priorities of the African continent. DE Africa will leverage and build on existing capacity to enable the use of Earth observations to address key challenges across the continent.

DE Africa portal: https://www.digitalearthafrica.org/

Africa has many agriculture challenges so monitoring and management was identified as high priority. Addressing these challenges requires better technology and policies to improve the management of small farms. Airbus Pleiades and SPOT have an impressive acquisition capacity that offers better earth observations of the area of interest and it is useful for monitoring of environmental practices and agri-environmental measures, which is very useful for the primary focus of the project: developing prototype crop delineation method as part of a Food and Agriculture Organization of the United Nations (FAO). The method is developed for country government users. The exact test area will likely be in Rwanda. For the small farms, higher resolution is desired.

Airbus Pleiades and SPOT high-res images were recognized as very useful for delineation of smaller crop fields.

Digital Gaia is an open intelligence and analytics platform that maximizes the environmental impact of regenerative projects and investments. Digitally­enabled, the Digital Gaia technology platform decentralizes impact assessment through a verification platform designed to enable high-integrity initiatives targeting regenerative effects in critical ecosystems. The open platform hosts a decentralized Natural Intelligence Network (ΝIΝ), which combines human expertise in ecosystem health and Artificial Intelligence to generate algorithmic impact assessments across all dimensions of natural climate solutions and associated investments. We will incorporate data layers from the farmers’ inputs, satellite imaging, remote sensing, climate modeling, and much more to create robust estimates for this project. This data aggregation into our active inference engine will result in tailored insights for the farmers, scientists, modellers and investors, ultimately increasing interoperability, transparency, and agility across the regenerative economy. The specific objective of this project is to create interactive, collaborative digital twins of 50 regenerative agriculture and agroforestry projects across Europe, Brazil, and the US. These digital twins will provide farmers, impact investors, scientists, and stakeholders with insights into these projects’ impact and actions for tangible improvement. The results will be free of charge to farmers and their nonprofit stakeholders through an interactive dashboard tracking the life of their project. These solutions scale up from the last mile of impact, where we focus on helping nature stewards and other innovators with the capacity to take action to optimize, demonstrate, and monetize their projects’ impact, creating clarity, trust and accountability for investors.

The project’s main objective is to build a nationwide dataset of slow-moving landslide deformation as the backbone of a national Early Warning System for Slow Moving Landslides. In many temperate countries, like Romania, the majority of active landslides are slowly moving, and the reactivations of inactive landslides are pretty frequent. Certain thresholds of precipitation represent the triggering. Therefore, building a Landslide Early Warning System (LEWS) requires at least a landslide inventory or a landslide hazard model besides the real-time rainfall data. We propose to use temporal landslide deformation from radar data obtained using DinSAR techniques and for training an AI model to predict landslide reactivation based on the trend of deformation and rainfall. Such a dataset and model can be used to implement the first LEWS for Romania, and the tested methodology could also be used for other areas.

The use of numerical weather prediction (NWP) models to drive snow cover models has recently become more and more investigated, thanks to the improved computer performances allowing to increase the spatial resolution and decrease the computational time. But still, some processes cannot be explicitly treated in the models because they are caused by phenomena happening at a fine scale. Thus the simulation of the snow cover is affected by the uncertainties of both atmospheric and snow cover models. Furthermore, the errors may increase if the simulations cover long periods; thus, the assimilation of observations in the snow models can help to reduce the simulation biases and make models converge to the observations. However, in situ observations of the snow conditions are usually done with automatic weather stations (AWS) and manual measurements. Thus they are sparse and insufficient to force a spatially distributed snow cover model. Instead, satellite data cover large areas at different resolutions and are the perfect candidates to correct snow cover models using gridded data from coarse to satisfactory resolutions. Optical data, for example, can give information on snow cover extent and albedo. At the same time, with DinSAR techniques, it is possible to estimate the snow height variation between different dates or even the snowpack liquid water content. Our project aims to develop an assimilation algorithm that will improve the snow cover model simulation quality using high resolution remote sensing data, to provide helpful information for avalanche warning services, hydrology services and even climates studies.

The main objective of my work is to assess different domain adaptation techniques regarding geographical domain shifts in land cover classification. First, different deep-learning segmentation models will be trained on Sentinel-2 data with CORINE land cover maps as reference data. The Sentinel-2 input will probably be multi-spectral (but not multi-temporal), and the CLC map from 2018. The initial dataset, called the source domain, will only contain samples from a specific geographic region (like Germany or a federal state of Germany). After an architecture (probably U-Net) which shows acceptable performance on the source dataset is found, the model will be applied to different geographic regions (the target data set) in Europe. Due to the domain shift across different areas, the model’s performance is expected to drop. This domain shift arises from different class distributions and other spectral and spatial properties of the classes. Then, different domain adaptation techniques will be applied and compared to mitigate the performance decrease. The key idea behind domain adaptation is that there are only labels for the source domain (e.g. Germany) but not for the target domain (e.g. Greece). But this technique will still be possible to improve the performance on the target domain. Especially in remote sensing, where labels are rare and expensive to acquire, domain adaptation can help achieve valuable results even with fewer labels. So far, research on domain adaptation in remote sensing has focused mostly on high-resolution aerial imagery (ISPRS Potsdam and Vaihingen) with 3-channel inputs. Only a few works deal with medium-resolution satellite imagery. Still, in these cases, they primarily classify pixels based on their spectral and temporal properties without considering spatial information (the surrounding pixels with fully convolutional networks).

The pilot aims to develop a pan-European scale drought impact monitoring platform using the new CLMS service High-Resolution Vegetation Phenology and Productivity (HR-VPP) derived from Sentinel 2 images.

DEMs (including DSMs and DTMs) are critical in land-use planning, infrastructural project management, soil science, hydrology and flow-direction studies. Because DSMs characterize the bare Earth and its above-ground features, their use is widely applied in fields such as urban planning (i.e., in investigating how a proposed building would affect the views of residents and businesses, power line corridor inspections and aviation planning). DEMS and DSMs are powerful and efficient tools for applications in various sectors. There are different ways to generate DEMs, including satellite data processing. In this project, we want to apply deep learning-based algorithms to make high-quality digital surface models. There are several proposed added values if one can implement such a workflow.

One of the advantages will be that the amount of manual work will decrease significantly. This is important for processing steps like DTM and BHM (Building Height Model) generation. Previously, if someone wanted to have a high quality BHM, it would need manual work to filter the DSM and then finally make BHM. The main problem here is that the derived satellite-based DSM doesn’t include details compared to UAV-based DSMs. So the expert operator must manually examine the DSM and detect where the buildings are. Then, after the expert detects the feature is a building, he can filter that. As it is clear, the process is time-consuming. Now we want to use very efficient matching algorithms to provide a better initial point and then apply deep learning algorithms.

Earth Observation Advanced Science Tools for Sea Level Extreme Events (EOatSEE) is a project funded by ESA and proposed by a consortium of institutions and companies internationally recognized for their work in the Marine, Coastal, and Earth Observation topics. It aims to provide an advanced reconstruction of the relevant processes included in extreme sea level (ESL) events and their related coastal hazards by taking advantage of the novel capabilities and synergies offered by the latest advances in EO technology. Therefore, the solid scientific knowledge arising from EOatSEE shall enhance the fundamental scientific understanding and predictive capacity of such events and our potential better to assess the related risk and vulnerability of coastal zones.

CentraleSupélec – a French high school of engineering – organizes a course on Satellite Earth Observation dedicated to around 110 first year students, from the 22nd November to the end of January. This course is an introduction to optical and SAR remote sensing. It is based on the use and processing of ESA’s Sentinel images. The support of EOCARE is requested to allow the students to carry out mini-projects on 3 topics at the end of the course.

The action focuses on integrating Earth Observation (EO) data from the European Space Agency (ESA) Copernicus Programme into the statistical production process for further analyses and projects within the fields of agriculture, forestry and environment. Critical aspects of the action are:

• the preparation and provision of required EO data and infrastructure;

• the methodological development for land cover classification using the EO data to gain spatially explicit data on different categories, focused on the needs of the statistical production process – namely woodland, grassland and cropland and the evaluation of a possible higher granularity within those classes;

• the dissemination of results and data and their further use for evaluating these integrations into statistical products and applications.

The ESA Earth Observation for Sustainable Development Lab (EO4SD Lab) project’s goal is to procure an Earth Observation (EO) processing and e-collaboration environment (the Exploitation Platform) dedicated to Development Assistance. The primary objective is to help the Sustainable Development community make better use of satellite capabilities to improve the delivery of projects on the ground. The Exploitation Platform is based on online processing and provides a new solution for using satellite imagery complementary to conventional service provision methods. End users can connect to the EP to retrieve EO based information products. In addition, expert users can directly generate products on the EP and integrate and share their service chain. The project’s technical activities will initially focus on designing and deploying the pre-operational Exploitation Platform, a cloud-based portal enabling users to find and use EO-derived information, products, and services relevant to their needs. This portal will follow the concept of the Thematic Exploitation Platforms (TEPs) in that it brings together large EO data archives and processing and analysis capabilities within a cloud environment – hence removing the computing or technical barrier regarding the user’s systems and infrastructure. After the initial platform release, and in parallel with functional improvements, activities will focus on ensuring the Exploitation Platform can showcase the potential of EO through the execution of several service pilots – these will be individual projects run on the platform that create information or product that can help meet the specific needs of engaged end users.

The ESA Earth Observation for Sustainable Development Lab (EO4SD Lab) project’s goal is to procure an Earth Observation (EO) processing and e-collaboration environment (the Exploitation Platform) dedicated to Development Assistance. The primary objective is to help the Sustainable Development community make better use of satellite capabilities to improve the delivery of projects on the ground. The Exploitation Platform is based on online processing and provides a new solution for using satellite imagery complementary to conventional service provision methods. End users can connect to the EP to retrieve EO based information products. In addition, expert users can directly generate products on the EP and integrate and share their service chain. The project’s technical activities will initially focus on designing and deploying the pre-operational Exploitation Platform, a cloud-based portal enabling users to find and use EO-derived information, products, and services relevant to their needs. This portal will follow the concept of the Thematic Exploitation Platforms (TEPs) in that it brings together large EO data archives and processing and analysis capabilities within a cloud environment – hence removing the computing or technical barrier regarding the user’s systems and infrastructure. After the initial platform release, and in parallel with functional improvements, activities will focus on ensuring the Exploitation Platform can showcase the potential of EO through the execution of several service pilots – these will be individual projects run on the platform that create information or product that can help meet the specific needs of engaged end users.

Artificial Intelligence (AI) is the transformational technology of our era. Earth Observation (EO) will significantly benefit, as in other areas, from its application by lowering the cost of adoption and accelerating market uptake. The Earth Observation Training Data Lab (EOTDL) aims to develop open-source tools to create, curate, analyze and use AI-ready EO datasets. A European cloud-based repository of datasets and AI models will be created, maintained and improved. Training capabilities will also enable researchers, engineers, and non-expert users alike to efficiently train AI models in the cloud with the available datasets and keep track of state of the art. Many areas will benefit from this platform: having a repository of AI-ready EO datasets will strengthen industry capabilities for exploiting EO data as a whole and help accelerate EO market penetration. Furthermore, to enable Digital Twin Earth simulations, it is needed that quality datasets exist for researchers and engineers to use and build quality models and applications.

Agricompas is developing a data analytics platform in an IPP (International Partnership Program) project funded by the UK Space Agency. EcoProMIS or Ecological Production Management Information System aims to provide all stakeholders involved in rice production with valuable insights in farmer and crop performance. Information is provided free to farmers (who are sharing in-situ data) to access information on crop management, soil and environmental conditions during the crop production cycle to improve decision making. Paid Analytics as a Service is provided to various stakeholders. A matchmaking platform will nurture stakeholder relations with safe and sound information of technical, economic, social and environmental processes. Farmers can place product and service requests with providers that can develop and tailor their offerings based on better farmer and field information.

EcoProMIS uses Earth Observation technology, especially from Sentinel satellites, to monitor maximal 500,000 hectares from up to 16,000 Colombian rice farmers in the four main ri