Artificial Intelligence (AI) for Earth Monitoring MOOC

Topic outline

• Course summary

  Collapse all Expand all

  Discover the wealth of Copernicus data that is now available and explore new ways of working with this data using AI and machine learning techniques.
  

• Week 1
  • Topic 1a – Introduction to Copernicus
    • Copernicus is an EU programme for satellite Earth observation. The European Commission manages the programme and it is implemented in partnership with the member states, the European Space Agency (ESA), the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), the European Centre for Medium-Range Weather Forecasts (ECMWF), EU Agencies and Mercator Océan International.

      Copernicus provides a unified system through which vast amounts of data are fed into a series of thematic information services designed to benefit the environment, the way we live, humanitarian needs and to support effective policy making for a more sustainable future. The data comes from satellite Earth observation and in situ observations.

      The data they provide is freely available to everybody, so we can make choices about what we do and how we as a society look after the environment. It serves a range of users, from businesses and public services – to researchers and curious individuals.

      The name Copernicus pays homage to a European scientist and observer – Nicolaus Copernicus. Copernicus was a mathematician and astronomer who proposed that the sun was stationary in the centre of the universe and the earth revolved around it. This theory made a pioneering contribution to modern science.

      There are six Copernicus services:

      • Atmosphere
      • Marine
      • Land
      • Climate Change
      • Security
      • Emergency

      As there is a vast amount of data produced by Copernicus, artificial intelligence (AI) and machine learning (ML) allow us to manage and understand that data in a way that has simply not been possible before.

      
      

  • Topic 1b - Sentinel Satellites and Contributing Missions
    • Copernicus is built on a constellation of satellites that make a huge number of daily observations – taking advantage of a global network of thousands of land, air and marine-based sensors to create detailed pictures of Earth.

      Each of the Copernicus Sentinel Missions carries state-of-the-art technology to deliver a stream of imagery and data tailored to the needs of users. Most of the missions are based on constellations of two satellites to optimise global coverage and data delivery.

      Sentinel-1: Carrying an advanced radar, Sentinel-1 images Earth’s surface through rain and cloud regardless of whether it is day or night.

      Sentinel-2: Sentinel-2 carries a high-resolution multispectral optical imager to monitor changes in vegetation.

      Sentinel-3: Carrying a precision radar altimeter, an advanced imaging radiometer, and a wide-swath ocean and land imaging spectrometer, Sentinel-3 supplies a wealth of data related mainly to the marine environment.

      Sentinel-4: Sentinel-4 is an ultraviolet, visible and near-infrared spectrometer carried on the EUMETSAT’s Meteosat Third Generation satellites.

      Sentinel-5P: Sentinel-5 Precursor is the forerunner of Sentinel-5. It is a satellite carrying the Tropomi imaging spectrometer to provide timely information on a multitude of trace gases and aerosols affecting air quality and climate.

      Sentinel-5: Sentinel-5 is an ultraviolet to shortwave infrared spectrometer carried on EUMETSAT’s polar-orbiting Metop Second Generation A series.

      Sentinel-6: Sentinel-6 carries a radar altimeter to provide high-precision and timely observations of the topography of the global ocean.

      In addition to the data provided by the Sentinel satellites, the contributing missions play a crucial role in delivering complementary data. This ensures that the Copernicus programme covers the required range of observational requirements.

      There are around 30 existing or planned contributing missions. They fall into the following categories:

      • Synthetic Aperture Radar (SAR) to observe day and night the land and the ocean
      • Optical sensors to monitor land activities and ocean dynamics
      • Altimetry systems for sea-level measurement
      • Radiometers to monitor land and ocean temperature
      • Spectrometers for measurements of air quality

      
      

  • Topic 1c - The Role of Copernicus Services and Partners in Managing Big Data
    • This course focuses on four of the six Copernicus services.

      The Copernicus Atmosphere Monitoring Service (CAMS) provides continuous data and information on atmospheric composition. The service describes the current situation in the atmosphere, forecasts the situation a few days ahead and analyses consistently retrospective data records for recent years. CAMS supports many applications in a variety of domains including health, environmental monitoring, renewable energies, meteorology and climatology.

      The Copernicus Marine Environment Monitoring Service (CMEMS) provides regular and systematic reference information on the physical and biogeochemical state, variability and dynamics of the ocean and marine ecosystems for the global ocean and the European regional seas. The observations and forecasts produced by the service support all marine applications, including: marine safety, marine resources, coastal and marine environment, weather, seasonal forecasting and climate.

      The Copernicus Land Monitoring Service (CLMS) provides geographical information on land cover and its changes, land use, vegetation state, water cycle and Earth’s surface energy variables to a broad range of users in Europe and across the world in the field of environmental terrestrial applications. It supports applications in a variety of domains such as spatial and urban planning, forest management, water management, agriculture and food security, nature conservation and restoration, rural development, ecosystem accounting, and mitigation and adaptation to climate change.

      The Copernicus Climate Change Service (C3S) supports society by providing authoritative information about the past, present and future climate in Europe and the rest of the world. The C3S missionsupports the adaptation and mitigation policies of the European Union by providing consistent and authoritative information about climate change. They offer free and open access to climate data and tools based on the best available science and aim to help their users meet their goals in dealing with the impacts of climate change.

      Each of these services produces a vast amount of data on a daily basis. AI can be used to understand that data and make it manageable. For example: AI could be used to identify features in satellite images, or track changes in satellite images over a period of time.

      EUMETSAT is an intergovernmental organisation based in Darmstadt, Germany – currently with 30 member states. EUMETSAT provides data, products and support services to the Copernicus information services and user communities, with a focus on marine, atmosphere and climate. This involves delivering Earth observation data services to Copernicus from the Sentinel satellites, from its own Metop and Meteosat missions, from the ocean-monitoring Sentinel-3 satellites and Jason-3 and Sentinel-6, and from missions of its international partners (such as USA, China, India and Japan). EUMETSAT is responsible for operating the Sentinel-3 and Sentinel-6 satellites and will also operate and deliver products from the Sentinel-4, and Sentinel-5 instruments onboard Meteosat Third Generation and Metop Second Generations satellites, respectively.

      ECMWF is both a research institute and a 24/7 operational service, producing global numerical weather predictions. ECMWF has one of the largest supercomputer facilities and meteorological data archives in the world. They also operate two services from the EU’s Copernicus Earth observation programme, the Copernicus Atmosphere Monitoring Service (CAMS) and the Copernicus Climate Change Service (C3S), whilst also contributing to the Copernicus Emergency Management Service (CEMS).

      Mercator Océan International is a non-profit company providing ocean science-based services of general interest focused on the conservation and the sustainable use of the oceans, seas and marine resources. Mercator Océan implements the Copernicus Marine Service (CMEMS).

      The European Environment Agency (EEA) is an agency of the European Union. They provide sound, independent information on the environment and they are a major information source for those involved in developing, adopting, implementing and evaluating environmental policy, and also the general public. The EEA co-ordinates the Copernicus Land Monitoring Service (CLMS).

      Together, EUMETSAT, ECMWF, Mercator Océan and the European Environment Agency implement the WEkEO DIAS cloud computing platform. WEkEO is the EU’s Copernicus DIAS reference service for environmental data, virtual environments for data processing and skilled user support.

      The EU has various initiatives that support the implementation of AI throughout EO, and to provide information and guidance of how AI can add value to products across a broad range of EO applications.

      For example:

      The AI4EU Consortium was established in 2019 to build the first European artificial intelligence on-demand platform and ecosystem with the support of the European Commission under the Horizon2020 programme. The activities of the project include:

      • The creation and support of a large European ecosystem spanning the 28 countries to facilitate collaboration people in Europe who use AI
      • The implementation of industry-led pilots through the AI4EU platform, which demonstrate the capabilities of the platform to enable real applications and foster innovation
      • Research activities in five key interconnected AI scientific areas (Explainable AI, Physical AI ,Verifiable AI, Collaborative AI, Integrative AI), which arise from the application of AI in real-world scenarios
      • The funding of SMEs and start-ups benefitting from AI resources available on the platform (cascade funding plan of €3m) to solve AI challenges and promote new solutions with AI
      • The creation of a European Ethical Observatory to ensure that European AI projects adhere to high ethical, legal, and socio-economical standards

      AI4Copernicus

      The EU-funded AI4Copernicus project will bridge the EU AI on-demand platform AI4EU with Copernicus Data and Information Access Services (DIAS) platforms, such as WEkEO to allow innovative players break new ground. The project will organise and fund four open calls for SMEs, realising eight small-scale experiments and nine larger-scale use-cases. Through its technological contribution and open-calls, AI4Copernicus will deepen the integration of AI4EU with multiple DIAS platforms. At the same time, it will enrich the European arsenal with sustainable, cutting edge applications in domains of high economic and societal impact, such as in Agriculture, Energy, Environment and Security.

      Deepcube DeepCube is a Horizon 2020 project funded by the European Union. Its aim is to unlock the potential of Copernicus data, using advances in the fields of Artificial Intelligence and Semantic Web - in an explainable way. The ultimate goal of DeepCube is to address new and ambitious problems that have a high environmental and societal impact, enhance our understanding of Earth’s processes that are correlated with climate change

      Supporting businesses using Copernicus data and services The Copernicus Accelerator is a 12-month acceleration programme designed explicitly for driven entrepreneurs and start-ups ready to turn their transformative business cases into impactful commercial solutions. It provides one-on-one guidance with expert mentors, a virtual training programme focused on start-up needs, and two boot camps held at leading industry events around Europe. This offers an opportunity for start-ups and entrepreneurs to develop their ideas on how to use Earth observation data into real companies.

      The Copernicus Masters is an international competition that awards prizes to innovative solutions, developments and concepts for business and society based on Earth observation data. EO and big data from space offer great potential for the creation of innovative products and applications for business and society. A wide variety of industries and areas of public interest stand to profit from the advances of the European Copernicus Programme.

      The ESA BIC initiative aims to inspire and work with entrepreneurs to turn space-connected business ideas into commercial start-ups companies. Since 2003, 700 start-ups have been fostered throughout Europe with thousands of new high tech jobs created thanks to the applications of space systems, the valorization of ESA intellectual properties and the space technologies transfers - and more than 180 new start-ups are taken in yearly at the ESA BICs.

      AI is a tool to help solve our grand societal challenges. AI will support Earth observation in big data analytics, help in cracking complex earth system models and in shaping new sensing technologies, such as onboard data compression and information retrieval. For the European citizen, AI will be the centrepiece in the EU industry’s transformation to be green, global and competitive. AI will be able to support downstream applications in industry that solve real world challenges and that can influence decision making.

      
      

  • Topic 1d – AI and Machine Learning in the Copernicus Programme
    • As one of the largest Earth observation data providers, the Copernicus archives contain a massive amount of data and information that can play a decisive role for policy support in many economic and industrial areas.

      Bringing in AI technology at the scale of Copernicus provides a unique opportunity to develop a new generation of products and services based on Copernicus data assets.

      Until now, the huge amount of data and the cost related to their analysis remain the major barriers to the widespread use of EO based services. AI can be the solution to quickly and efficiently respond to opportunities and challenges for the detection, identification and mapping of marine, atmosphere and land phenomena.

      When it comes to using AI for Earth monitoring, we face several specific challenges that make it different from other computer vision problems. The main issue is training data sets. For example, if you want to train a convolutional neural network (CNN) model to identify hurricanes, you need a training data set of labelled hurricanes or hurricane images, so that the model can learn from this to be able to identify hurricanes in new images it has not seen before. At the moment, these training data sets are few and far between and they take a long time to create.

      Additionally, remote sensing data is multimodal – for example, we may have both optical and SAR data, so combining different data sources is a challenge in itself.

      As well as this, in order to be fused with other data sources, the data needs to be geo-located, which adds a further level of complication. In addition, we have to consider the fact that EO is time-dependent, and we may have to consider time as part of the data. For example, we might have multiple images from the same location. AI helps us to deal with all these challenges in order to process a massive volume of multi-modal Earth observation data.

      
      
      

  • Topic 1e – Accessing Data and Using the WEkEO Platform
    • Within the field of EO, it is important to know what kind of data you are using and how to access it, because you are often dealing with big data.

      Big data is a large amount of data, which can be structured or unstructured. Big data is hard to access with traditional methods, for example – you would not be able to load a whole set of satellite data on a normal home computer because they take up huge amounts of memory. It is not only hard to access big data but it is also hard to work with.

      We need to find a way to digest all the information produced by satellites – because it is not possible to process manually the massive volume of data that is produced daily. AI and ML can be used to make big data more manageable and usable for a variety of applications.

      AI can help us to manage and process the electromagnetic quantities that we can see from satellite data, such as reflectance and backscattering, into physical quantities such as sea surface temperature and sea surface height.

      WEkEO

      The European Commission recently launched an initiative called DIAS (Data and Information Access Service) to provide single access points to all Copernicus data and information, alongside processing resources, tools and other relevant data.

      As key organisations in the Copernicus Programme, EUMETSAT, ECMWF, EEA and Mercator Océan International combined their long-standing experience to develop the WEkEO Copernicus DIAS service, one of a total of five DIAS platforms.

      WEkEO provides users with a single distributed tool for accessing, visualising and analysing all Copernicus data and services, including big-data analysis tools, to develop applications tailored to their specific needs.

      The platform also provides a specialised training and user support service using the experience of the key partners such as EUMETSAT, ECMWF, Mercator International, EEA.

      WEkEO's user support service is operated by experts who can handle all types of requests, from ICT problems to image processing library installation and code debugging, or detailed insight into the data structure, applications etc.

      WEkEO’s strength relies on its distributed infrastructure built thanks to the Copernicus organisation already in place in the four centres: infrastructure and data from the four centres are not duplicated but rather linked together, reducing costs for the European tax payer and avoiding the need for new energy-consuming infrastructure.

      This approach gives users instant, direct access to work with the most up-to-date Copernicus data instead of having to rely on copied, archive datasets.

      Finally, this federative approach can be extended to other selected external partners, so the system has the potential to continually grow and expand in the future.

      
      

  • Introduction to the WEkEO Jupyterhub Training Platform
    • Throughout the course, there will be hands-on tutorials where you can have a go yourself at using AI and ML algorithms to analyse Earth observation data. These practical elements of the course will take the form of Jupyter Notebooks, which work like documents containing both rich-text (such as paragraphs and links) and computer code, in this case Python, and the code can be run which will output results (such as graphs and tables).
      

      Each Jupyter Notebook is hosted on the WEkEO platform and this allows you to access them in your internet browser.

      Before you can interact with each Jupyter Notebook you will need to sign up to WEkEO which you can do by following this link (If you would like this to open in a new tab you can right-click on the link and choose ‘open link in new tab’).

      In this video, Julia Wagemann will walk you through registering on WEkEO and how to access the Jupyter Hub, where you will find AI Notebook tasks for this course.

      How to Troubleshoot Jupyter Notebooks

      In this video, Julia Wagemann will show you how to troubleshoot the Jupyter Notebooks in case you have any issues while completing the activities.

      
      

• Week 2
  • Topic 2a – Overview of Types of AI
    • So, what is artificial intelligence? Artificial intelligence is the overarching name for a branch of computer science that deals with smart machines or robots that can think and perform tasks like a human. In artificial intelligence you can teach a computer to learn and make its own decisions. AI is the broad concept of machines being able to carry out tasks in a way that we would consider ‘smart’, and we encounter these technologies on a daily basis: from voice assistants in our home, to Netflix recommendations or autonomous cars.

      AI relies on data and on training algorithms. In this context, an algorithm is a set of rules for a computer to follow. In AI, data and training algorithms are what makes the computer seem intelligent, because they allow it to make decisions based on the rules that it has learned to follow.

      For a computer to learn it needs to find links between things, develop understanding and make decisions based on the training data that they are given. Sometimes, people can think of AI as a ‘black box’ – where you put everything into a machine and a magic formula will give you the exact results that you want. However, it is very important to remember that you only get out of AI what you put into it. So, if you enter poor quality data into a training algorithm, you will get poor results. Later in this week and throughout the course, we will go into more detail about training data and why it is so important in Earth observation.

      Machine Learning is a subset of artificial intelligence. It is the art of teaching a computer to learn and improve from experience rather than being explicitly programmed. Machine learning is an ‘empirical approach’, which means it learns from what is seen or experienced rather than theoretical models. We encounter machine learning in our day-to-day lives: for example, using face-detection to unlock our phones, or getting real-time traffic updates on Google Maps.

      The next subset of machine learning is deep learning. This is a kind of machine learning that aims to solve problems by mimicking the biological structure of the brain. To do this, they use a type of model called a neural network, which follows the way that the brain makes connections between different pieces of information. Later in this course, you will learn more about different kinds of neural networks.

      Today, deep learning is reaching such a high level of accuracy that it has the ability to replace many difficult, expensive and time-consuming tasks, which opens up huge opportunities for businesses and science, including in Earth observation.

      Furthermore, another subset of machine learning is reinforcement learning. Reinforcement learning is the training of machine learning models to make a sequence of decisions, taking into account interaction with the environment. In reinforcement learning, the AI faces a game-like situation: the computer employs a trial and error system to come up with a solution to the problem that it has been given. To get the machine to do what the programmer wants, the artificial intelligence will either be rewarded or penalised based on the actions it performs. Its goal is to maximise the total reward and solve the problem.

      An example of the use of reinforcement learning is the AlphaGo programme which was able to beat a professional human Go player for the first time.

      In this week, we will go into more detail about the kinds of problems that can be worked on using AI, and the different kinds of algorithms that can be used to solve these problems. Later in the course, you will see how these concepts are applied in the field of Earth system monitoring to solve problems that we face in the real world.

      
      
      

  • Topic 2b Part 1 – Types of Machine Learning Problem: Supervised Learning
    • In machine learning there are three kinds of learning problems (or tasks): these are supervised learning, unsupervised learning and reinforcement learning. The next two videos will cover supervised and unsupervised learning, and you will have a look at reinforcement learning in the first Jupyter Notebook task.

      Supervised learning is a method of training an algorithm where a data set consists of two parts: the data itself and a label. For example, a picture of a strawberry and a label which describes it as a strawberry.

      The algorithm is rewarded when it predicts the correct label, and it can rearrange itself if it predicts the wrong label.

      Similarly, the data can be associated with a real number, for example representing a geophysical parameter, rather than with a label. This is an example of a regression problem, where the task is to predict a continuous quantity.

      This can be used to do two things:

      • Classification: sorting data into classes
      • Regression: parameter estimation

      
      

  • Topic 2b Part 2 – Types of Machine Learning Problem: Unsupervised Learning
    • Next up is unsupervised learning: this is where a ML is given a training data set that does not have any labels.

      The ML is left to its own devices – and its goal is to find structure in the input that it was given.

      This can be useful in Earth observation, because scientists are often dealing with very large datasets.

      Like supervised data, unsupervised data can also be used for classification, clustering and feature identification.

      To make things more complicated, you can also perform semi-supervised learning. This is a combination of supervised and unsupervised learning. When you don’t have enough labelled data to produce an accurate model and you don’t have the ability or resources to get more data, you can use semi-supervised techniques to increase the size of your training data. Unlabelled data, when used in conjunction with a small amount of labelled data, can produce considerable improvement in learning accuracy.

      Jupyter Notebook Task - Reinforcement Learning

      Reinforcement learning is the training of machine learning models to make a sequence of decisions. For this first Jupyter Notebook task, you will have a look at an example of a reinforcement learning workflow.

      For this Jupyter Notebook workflow tutorial, please follow TBC to go to WEkEO. You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      You may need to log in if you are not already. If you need to register with WEkEO you can do so here.

      
      
      

  • Topic 2c - Understanding Machine Learning Workflows
    • So, what’s the process of setting up a machine learning workflow?
      

      In simple terms: you firstly need to understand what your problem is.

      After that, you need to choose appropriate data and the correct learning algorithm.

      Then comes the most crucial part of machine learning: preparing your training data set.

      Once that’s done you can train your model. It is very important to evaluate the performance of the model. You do this by testing it on a selection of the training data that you held back from the ML, and didn’t use before starting the learning process.

      Then if everything has gone to plan, your problem is solved! But if necessary, you can re-run your model with optimised parameters.

      Throughout this course, you will have the opportunity to look at and interact with many ML workflows, like the one you have already seen in the step before this!

      Jupyter Notebook Task - Understanding Machine Learning Workflows

      In this Jupyter Notebook task, you will learn about machine learning workflows.

      For this Jupyter Notebook tutorial, please follow TBC to go to WEkEO. You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      You may need to log in if you are not already. If you need to register with WEkEO you can do so here.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below.

      
      

  • Topic 2d – Common Machine Learning Algorithms
    • An algorithm is essentially a set of instructions that a computer follows to solve a problem. There are many kinds of machine learning (ML) algorithms, but these are some common examples that you will encounter during this course:

      • K-Nearest Neighbour (KNN) - This is a simple classification and regression method. It assumes that similar things exist in proximity and therefore similar things are near to each other. The principle behind nearest neighbour methods is to find a predefined number of training samples closest in distance to the new point, and predict the classification label from these. In general, the distance can be any metric measure: standard Euclidean distance is the most common choice.
      • Decision trees - These algorithms have a flowchart-like structure, which looks like an inverted tree. Each internal node or condition in the tree represents a “test” on an attribute (such as a yes or no question). The branches then split based on this test. . The end of the branch that is the decision, or leaf. In the world of AI, decision trees are used to develop learning machines by teaching them how to determine success and failure.
      • Random Forest - This technique is a collection of decision trees, known as forests. For classification tasks, the output of the random forest is the class, or decision selected by the most trees.
      • Support Vector Machines (SVM) - This is a linear model for classification and regression problems. It takes in data points and creates a line or a hyperplane which separates the data points into classes.
      • Gradient boosting - This is another ensemble technique (like Random Forest), it uses a ‘boosting’ method to combine decision trees, which means that it takes in many weaker trees to create stronger trees. It does this by each tree attempting to minimise the errors of the previous tree. Since trees are added sequentially, boosting algorithms learn slowly, but slower learning algorithms are known to perform better.

      As well as these ML algorithms, you will also come across deep learning algorithms or methods, which are a modern update to Artificial Neural Networks (ANN) that exploit abundant cheap computation. They are concerned with building much larger and more complex neural networks. A neural network, or ANN, is an artificial system that mimics the way a human brain works, with neurons connected to synapses that transmit information. Neural networks can be trained so that they can recognise images, numbers and patterns, and they learn and improve their accuracy over time.

      Deep learning algorithms include:

      • Convolutional Neural Network (CNN or ConvNet) - These are used for image recognition, pattern recognition, and computer vision.
      • Recurrent Neural Network (RNN) - These are mostly used for time-series data to make predictions about future outcomes.
      • Multi-Layer Perceptron (MLP) - These are used for classification prediction problems where inputs are assigned a class or label and regression prediction problems where a real-valued quantity is predicted given a set of inputs.
      • Generative Adversarial Network (GAN) - These can be utilised for translating data from images, such as text to image translation or image to photo translation. They can also generate new images from a given set of images that are similar to the given dataset, yet individually different.

      
      
      

  • Topic 2e - Python Libraries for Machine Learning
    • More general frameworks:

      scikit-learn - Open-source ML framework in Python

      Tensorflow - Comprehensive ecosystem of ML tools, libraries and community resources. Allows to easily build and deploy ML powered applications. Keras is a Deep Learning Framework built on top of Tensorflow 2.0

      PyTorch - Open-source ML library based on the Torch library

      More specialised frameworks / libraries:

      xgboost - Open-source software library which provides a gradient boosting framework. Integrated into scikit-learn

      Jupyter Notebook Task - Python Libraries for ML

      This Jupyter Notebook task will provide you with an overview of the four most recurrent Python Libraries and frameworks.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below.

      
      

  • Topic 2f – Ethical Considerations
    • The use of AI often sparks debates about ethics and bias. AI can represent most human insight, but some human insight is inherently biased. It is important that we have a high degree of confidence that the data being used for AI is correct and accurate.

      EO research makes use of various types of data, such as satellite images and social media data. However, improved image resolution raises questions under the new European General Data Protection Regulation (GDPR) which came into force in 2018. The collection, storing, processing and transfer of such data needs to be done with much greater attention to legal and ethical requirements by adopting supporting technologies that give data contributors greater control over their personal data. Therefore, technologies such as blockchain are also likely to be researched and deployed, including from an ethical perspective, for fundamental AI4EM research.

      There are other ethical issues, such as algorithmic bias, resulting in discrimination, which need to be checked and addressed at early stages of technological development. Building responsible AI4EM that embodies social norms and values while ensuring sustainable and inclusive development is very important. Accordingly, we will address issues of data protection, privacy, data portability, and equality at the level not just of data collection, but also at the level of data use and dissemination within the sphere of ethics in AI4EM

      Data Insights: In Conversation with Joana Miguens

      In this first data insights video, we hear from Joana Miguens, who is Copernicus Data Services System Architect at EUMETSAT.

      
      

  • End of week summary

• Week 3
  • Topic 3a – Introduction to Monitoring the Land
    • 
      

      The Copernicus Sentinel satellites provide us with a wealth of data that can be used for monitoring Earth’s land.

      Sentinel-1 is a two-satellite constellation. Each satellite carries an advanced radar instrument that provides an all-weather, day-and-night supply of imagery of the Earth’s surface.

      Sentinel-2 is a constellation of two polar-orbiting satellites placed in the same sun-synchronous orbit, phased at 180° to each other. Sentinel-2 systematically acquires optical imagery at high spatial resolution (10 m to 60 m) over land and coastal waters. Its purpose is to monitor variability in land surface conditions, and its wide swath width (290 km) and high revisit time - it supports monitoring of Earth’s surface changes.

      Sentinel-3’s main objective is to measure sea surface topography, sea surface temperature and ocean colour but it also monitors land surface temperature, and land surface colour with high accuracy and reliability to support environmental monitoring and climate monitoring.

      The Copernicus Land Monitoring Service (CLMS), which is part of the Copernicus programme, uses data from the Sentinel satellites - and other satellites - to provide geographical information on land cover to a broad range of users in the field of environmental terrestrial applications. This includes land use, land cover characteristics and changes, the state of vegetation, and water cycle and Earth surface energy variables. It is free to access by any citizen or organisation in the world.

      The service is divided into four main sections:

      1. The systematic monitoring of biophysical parameters produces a series of qualified bio-geophysical products on the status and evolution of the land surface. This is produced at a global scale every ten days with a mid-spatial resolution and is complemented by a long-term time series. The products are used to monitor vegetation, crops, water cycle, energy budget and terrestrial cryosphere variables.
      2. Land cover and land use mapping produces land cover classifications at various levels of detail, both within a pan-European and global context. At the pan-European level, these are complemented by detailed layers on land cover characteristics, such as imperviousness, forests, grassland, water and wetness and small woody features. At global level, a reference for the land cover mapping is represented by the FAO’s modular-hierarchical Land Cover Classification System.
      3. Thematic hot-spot mapping aims to provide tailored and more detailed information on specific areas of interest, known as hotspots. Hotspots in the context of CLMS are prone to specific environmental challenges.
      4. Imagery and reference data provide satellite image mosaic in high and very high resolutions and reference datasets. This includes satellite image mosaics from contributing missions covering the territory of Europe as well as Sentinel-2 image mosaic production at global level. It also consists of reference datasets providing homogeneous pan-European coverage of some key geospatial themes, such as hydrography and elevation.

      CLMS is jointly implemented by the European Environment Agency and the European Commission DG Joint Research Centre (JRC) and has been operational since 2012.

      
      
      

  • Topic 3b – Land Cover Classification
    • Most metropolitan areas face increasing problems connected to irregular urban expansion, loss of natural vegetation, decreasing open space and general reduction of spatial extension. Towns have turned from small, isolated centres to big settlements.

      The importance of AI in monitoring land cover changes by means of EO data at various scales – continental, national and local – is crucial from at least two points of view because it reduces the photo interpretation loads and assures the objectivity of results.

      Jupyter Notebook Task: Tile-based classification based on EuroSAT data - Training

      In this Jupyter Notebook task you will use EuroSAT data to complete land cover classification of Rome, Italy, using deep Convolutional Neural Network (CNNs). The EuroSAT dataset is based on Sentinel-2 satellite images that cover 13 spectral bands and consisting of 10 classes with 27000 labeled and geo-referenced samples.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below.

      Jupyter Notebook Task 2 - Tile-based classification based on EuroSAT data - Inference

      In this Jupyter Notebook task you will use EuroSAT data to complete land cover classification of Rome, Italy, using deep Convolutional Neural Network (CNNs). The EuroSAT dataset is based on Sentinel-2 satellite images that cover 13 spectral bands and consisting of 10 classes with 27000 labeled and geo-referenced samples.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      
      

  • Topic 3c - Land Cover Usage Change
    • Land cover usage changes through time, due to various factors such as crop rotation or due to the seasons. Because of this, you need to classify images with a temporal (relating to time) resolution by adding recurrent connections into the model.

      For example, a tree changes colour throughout the seasons, but it doesn’t matter if it’s green or yellow – it’s always a tree.

      Land cover usage is frequently used to address various issues, such as food security, urban sprawl as well as to address ecosystem health.

      
      

  • Topic 3d Part 1 – AI For Agriculture: Precision Farming
    • Agriculture provides the most basic needs of humankind. New agricultural techniques should be able to meet future demands while maintaining or reducing the environmental footprint of agriculture.

      Emerging technologies such as AI and EO could be utilized to make informed management decisions aimed to increase crop production. Precision agriculture (PA) uses these technologies to optimise agricultural inputs, in order to increase production and reduce input losses.

      The use of remote sensing technologies for PA has increased rapidly during the past few decades. The unprecedented availability of high resolution (spatial, spectral and temporal) satellite images has promoted the use of remote sensing in many PA applications including crop monitoring, irrigation management, nutrient application, disease and pest management, and yield protection.

      The EFTAS CropAnalyzer application is used to annotate field photos of crops automatically to the growing crop type. Therefore, it could support field-based data collection for training and validation of crop type products. The network is trained on 80,000 field photos from 2012-2019 collected from ~40,000 locations. It provides a model accuracy of 90%.

       

      Jupyter Notebook Task: How to perform soil moisture inversion

      This Jupyter Notebook workflow will guide you through the different steps to train a sequential neural network model in order to predict soil moisture based on SAR backscatter and incidence angle information.

      Soil moisture is a crucial parameter in precision agriculture and in crop condition monitoring, so it is important to be able to measure it in an accurate and precise way.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below.

      
      

  • Topic 3d Part 2 – AI For Agriculture: Food Security
    • Crop mapping is important for many applications, especially those supporting environmental, economic and policy issues. PhiLab partnered with the World Food Programme, because they need accurate crop type information to help manage food supplies and children’s welfare in developing countries. Availability of new sensor platforms and the latest developments in transferring models between different image domains create an opportunity to improve existing crop type mapping approaches.

      This project was an unprecedented attempt to fuse information extracted from images captured by satellites, drones and from cameras and smartphones.

      Using three datasets of drone images acquired in Malawi and Mozambique, they found that transfer learning from computer vision could be successfully applied to crop type classification. It means that low-level visual information can also be learned from millions of images addressing a different applicative scenario (for example the distinction between cats and dogs). Moreover, the classified drone data can be used to increase the number and diversity of input samples needed by Sentinel-2 based classification systems (such as Sen2Agri) which make use of multi-temporal information to increase the accuracy of crop type mapping.

      Jupyter Notebook Task: Land cover classification with Sentinel-2 data - Supervised

      For this Jupyter Notebook, you will have a look at both supervised and unsupervised land cover classification with Sentinel-2 data. The aim is to classify a Sentinel-2 image into four different land cover classes: water surfaces, artificial surfaces, low vegetation and tree cover.

      This type of land classification allows for assessment and monitoring of land use changes, which is important for the sustainable management of natural resources, environmental protection, food security, and humanitarian programs.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below!

      Jupyter Notebook Task - Land Cover Classification with Sentinel-2 Data - Unsupervised

      This Jupyter Notebook workflow introduces you to an example of an unsupervised land cover classification with Sentinel-2 data.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      
      

  • Topic 3e - Mapping Deforestation in Real Time
    • Forest degradation is a global phenomenon contributing to climate change because it contributes to CO2 emissions and it also leads to the loss of natural ecosystems and declining biodiversity.

      Beyond a critical point of species removal or diminishment, the ecosystem can become destabilized and collapse. Although several methods have been developed to detect and map forest degradation, for example by using optical, synthetic aperture radar (SAR) and LiDAR data, there is no single method that can be applied to monitor forest degradation.

      This is largely due to the specific nature of the degradation type or process and the timeframe over which it is observed. Typical indicators of deforestation are decrease of vegetation indices and biomass, which can be derived from high resolution optical (such as METimage on Post-EPS) and scatterometer data (such as Metop Advanced Scatterometer (ASCAT)), and increase in surface albedo (sunlight reflected from the surface of the Earth).

      Furthermore, deforestation leads to short-term and long-term changes in the emissions of the three most abundant greenhouse gases (CO2, CHG, and NOx). AI may help to relate the existing indicators and develop a more smart deforestation monitoring system.

      IIASA and SAS have collaborated to create a citizen science project. They launched a crowdsource-driven app to gather the collective intelligence of the crowd. They asked volunteers to review and judge images of the rainforest and tag areas that have been affected by human development.

      The crowdsourcing helps to improve AI algorithms, speeding up what used to take years to analyse, and therefore help to drive vital policy responses to protect the Earth’s rainforests more quickly. The project covers 807,000 square kilometres of the Amazon. To date [8th Feb 2021] – they have evaluated 731,547 square kilometres.

      
      

  • Topic 3f - Mapping the Extent of Forest Fires
    • Large-scale forest fires are one of the most harmful natural hazards affecting climate change and life around the world.

      During recent decades, the number of fires have been increasing dramatically, in terms of both frequency and magnitude. To mitigate their impact on climate and nature, effective early warning systems are required that are able to identify areas sensitive to forest fire, as well as to detect the early onset of forest fires. The forest fire hazard depends on the availability of living biomass (live fuel) and dead biomass (dead fuel), which can be monitored with Earth observation systems. The forest fire risk depends on the evolution of various indicators, such as precipitation, wind speed, relative humidity or air temperature. Finally, the forest fire behaviour depends on dead fuels’ moisture content, wind, and terrain conditions.

      Forest fires are increasing in frequency and severity on a global scale, causing serious human, environmental and economic damage. Information about burn scars and vegetation loss can be obscured from optical and infrared sensors by smoke, haze and clouds in the aftermath of fires. SAR sensors have the ability to image the surface through clouds, smoke and haze and can provide information about structural changes in vegetation as well as changes in surface water content. This is particularly useful in areas of persistent cloud cover where a clear view from medium resolution optical images can take months to get after an incident and vegetation regrowth may make detection difficult. As well as the necessity to monitor fire outbreaks and predict spread in real time to coordinate fire-fighting efforts, measuring the contribution of emissions from fire events to atmospheric carbon dioxide is vital for climate modelling. A burned area labelled dataset from Copernicus Sentinel-1 constitutes a vital step towards the use of deep learning methods for rapid reporting of medium to high resolution burned areas in all weather conditions.

      The aim of PhiLab’s research is to develop a novel method for generation of burned area measurements, using rapidly generated AI ready data, maximising the information available from a time series of Sentinel-1 SAR SLC products. This will generate labelled data from archive S-1 SLC data. The selection of training areas will be automated and based on global stratified sampling by fire incidence (using high confidence active fire hotspot from Sentinel-3 and MODIS), land cover type (using ESA Land Cover CCI), season, and Sentinel-1 availability. Pre-processing of Sentinel-1 SLC images will be fully automated and cloud based.

      Instruments on Copernicus satellites are used to produce images of smoke, flames and burn scars (Sentinel-2), monitor fire hotspots (Sentinel-3) and provide analysis of air pollution, such as carbon monoxide (Sentinel-5P). The C-band SAR sensor from the Sentinel-1 satellites may provide unique insight into the severity of burns, as well as complementing the other Copernicus mission datasets. Initial priority is being given to generation of Analysis Ready Data (ARD) for gathering labelled training data from Sentinel-1 archives. The ARD generation aims to address the limitations of both datasets (e.g. flattening and masking geometric distortions in SAR, gap filling of clouds and shadow in optical), and will build on and contribute to existing efforts externally and in the PhiLab.

      
      

  • Topic 3g - Informal Settlement Mapping
    • Informal settlements are home to the most socially and economically vulnerable people on the planet. However, there is very little data available on them, and where there is, it is often incomplete.

      Informal settlement detection is very important for census, as it has an important human health benefit. If informal settlements do not appear on maps, they can miss out on vaccinations and other healthcare, for example during Ebola, as well as economic and social aid.

      In this audio lecture, Bradley Gram-Hansen explains how machine learning and low resolution multi-spectral data from Sentinel-2, can be used to help map informal settlements in developing countries.

      
      

  • Topic 3h - The Ethics of Mapping Poverty
    • The Munich Technical University (TUM) is aiming to build a map of poverty. This involves a regression problem, rather than a classification problem – which means that the AI is used to predict a number rather than a category.

      Data about poverty, for example from a census, is mapped onto satellite imagery. The AI can then make predictions about how this will change over time, or how this might extend into different locations.

      This brings in lots of ethical implications which they are looking into at TUM. For example – are you putting refugees at risk by mapping where their settlements are? There are positives to this but also it could be problematic if the data is used by adversaries.

      There is also the question of bias. For example, if an electricity company commissions a company to use AI to find out where to provide electricity, they will be looking for a return on their investment – so they may be biased towards places where there are people who will be able to pay for the electricity. In this video, Professor Xiaoxiang Zhu from TUM will discuss this complex issue.

      
      

  • Week 3 - End of Week Summary DC

• Week 4
  • Topic 4a - Introduction To Monitoring The Oceans
    • Whether it is following rising sea levels, mapping ocean surface winds and currents, monitoring water quality, or helping track sea turtles or jellyfish blooms, satellites have made a huge difference to our ability to monitor the world’s oceans. And because the ocean, land and atmosphere are inextricably linked, what happens in the oceans makes a big difference for life on Earth.

      Key ocean monitoring satellites are as follows:

      Copernicus Sentinel-1 satellites carry a synthetic aperture radar instrument (SAR) that provides radar data to monitor sea ice, marine wind, oil spills and marine navigation.
      
      Copernicus Sentinel-2 satellites collect high resolution ocean colour data in coastal areas.
      
      The Copernicus Sentinel-3 satellites collect observations of global ocean colour, sea surface temperature and sea surface height.
      
      Copernicus Sentinel-6 and Jason-3 satellites provide global sea surface height observations for climate monitoring and ocean and seasonal forecasts.
      
      In addition, EUMETSAT’s Meteosat satellites provide hourly measurements of sea surface temperature, while the Metop satellites provide global observations of ocean surface wind, sea ice, and sea surface temperature, plus collect in situ observations of the ocean through the ARGOS system.

      While satellites play a key role in providing the big picture of what is happening in the oceans it is also important to collect in situ data in the oceans to ground truth satellite measurements. Unfortunately, this is more difficult as they are so vast, but this is one area where AI can really fill in the gaps by predicting measurements where data are sparse. However, this can also pose problems, because we need the ground truth from the oceans in order to train the models in the first place. This is why in some cases, AI is used for data fusion between satellite data and in situ networks. AI and machine learning can also help to provide new insights on marine ecosystems, manage human activities on the oceans and monitor the impact of climate change.

      The Copernicus Marine Service (or Copernicus Marine Environment Monitoring Service) is the marine component of the Copernicus Programme of the European Union.

      CMEMS provides free, regular and systematic information on the state of the Blue (physical), White (sea ice) and Green (biogeochemical) ocean, on a global and regional scale. It is funded by the European Commission (EC) and implemented by Mercator Ocean International.
      
      It is designed to serve EU policies and international legal commitments related to ocean governance, to cater for the needs of society at large for global ocean knowledge and to boost the Blue Economy across all maritime sectors by providing free ocean data and information.

      CMEMS provides key inputs that contribute to combating pollution, marine protection, maritime safety and routing, sustainable use of ocean resources, developing renewable marine energy resources, supporting blue growth, climate monitoring, forecasting, and more.

      It also aims to increase awareness amongst the general public by providing European and global citizens with information about ocean-related issues.

      
      

  • Topic 4b – Tracking Ships
    • Monitoring the movement of ships and their activity at seaports, as well as current and forecasted weather conditions, is critical for the shipping industry, from trade to tourism.

      Deep Blue Globe is a start-up that came out of a Copernicus Hackathon at EUMETSAT in Darmstadt and then developed with the support of the ESA Business Incubation Centre in  Hessen and Baden-Württemberg.

      They have developed an AI-based autonomous ship navigation system that  considers  real time maritime traffic and weather conditions based on satellite data for any vessels, in particular in the shipping industry.

      Deep Blue Globe  uses data from the Copernicus Programme and also benefited from WEkEO as a data access and development platform.

      The company is using  AI/machine learning to develop a number of different products and services, including:

      Poseidon - a maritime autonomous navigation system powered by artificial intelligence considering real time maritime traffic and weather conditions based on satellite data for any vessels, in particular in the shipping industry. The benefits of smart ship-routing include: reducing CO₂ and SOx emissions, fuel cost saving and an increase in safety.

      OneWave - a 24/7 service of Wave Front identification in the open sea that uses Copernicus satellite  data (mostly altimetry) to complement the existing tsunami early warning systems.

      Jupyter Notebook Task: AI for ship classification

      For this Jupyter Notebook task you will complete ship classification using Sentinel-1 and OpenSAR ship datasets with a convolutional neural network (CNN) used for large-scale image recognition.

      Ship detection and classification from satellite imagery is useful for marine surveillance, so that governments can be aware of what type of ships there are and what they are doing - from fishing, drilling and exploration to cargo transport and tourism, or if they will have any impact on the security, safety, economy, or environment.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below.

      
      

  • Topic 4c - Marine Safety
    • Marine debris is any persistent solid material that is manufactured or processed and directly or indirectly, intentionally or unintentionally, disposed of or abandoned into the marine environment. 

      Marine debris is a hazard to navigation safety, endangers wildlife and is a threat to human health.

      With optical satellite images and a machine learning approach, debris can be detected and characterised, and we can monitor their drift with ocean forecast products, which helps to  make marine navigation more safe.

      
      

  • Topic 4d - Monitoring Marine Life: Part 1 - Fish
    • Machine learning is becoming increasingly important for monitoring marine life.

      Being able to monitor the distribution of key areas for fish in the ocean is very important to improve our  understanding of  marine ecosystems. Better knowledge means we can manage marine resources in a more sustainable way and also better understand the effects of climate change in the future.

      Fish are also a vital source of food for billions of people worldwide and they also provide a livelihood to millions. However, fish stocks are at threat due to overfishing and climate change.

      In this video, Georgios Sylaios explains how satellite data and ML can be used to investigate the distribution of pelagic fish in the Mediterranean Sea.

      
      

  • Topic 4d - Monitoring Marine Life: Part 2 - Jellyfish
    • Large swarms of jellyfish can cover hundreds of square miles of ocean and can cause both economic and environmental damage. Examples of this include reducing tourism due to the closing down of beaches and damage to fisheries due to the clogging of fishing nets and preying on eggs and larvae of fish.

      With satellite images, jellyfish swarms are detected with ML and we can monitor their drift with ocean forecast products.

      JellyX is an ocean-wide monitoring tool for the detection of jellyfish swarms and the prediction of their drift, based on ocean data. Fish farmers, whose activity is severely impacted by this phenomenon, can detect jellyfish swarms in advance, preventing important profit losses. With high estimation accuracy of jellyfish presence or absence, JellyX provides fish farmers with the solution for predicting and preventing the economic impacts of jellyfish outbreaks.

      JellyX combines:

      • Copernicus Marine Service products
      • Copernicus Sentinel 3 satellite
      • In-situ observations and ancillary data sources
      • automatic classification algorithms

      The Copernicus Marine Service products used in JellyX are currently focused on the Mediterranean Sea but concern satellite datasets and biogeochemical models that will be applied to other oceans and seas.

      Supervised machine learning techniques are used to estimate the abundance of jellyfish and classify the risk of proliferation as low, medium and high risk.

      
      

  • Topic 4d - Monitoring Marine Life: Part 3 - Turtles
    • All marine turtle species are experiencing serious threats to their survival. These threats include pollution, accidental killing in fishing nets or by fishing hooks, and changes to their habitats such as coral reefs, mangrove forests, seagrass beds, as well as their nesting beaches.

      In order to minimise the death of turtles from fishery equipment, it is important to know where the turtles are located, so that they can be avoided. Adult turtles can be easily equipped with beacons that allow tracking them by satellite and determining areas they occupy, but this is not the case for juveniles, who are generally too small to be equipped with the tracking beacons. So, in order to understand where juvenile turtles are and when (known as ‘lost years’), Philippe Gaspar has been using the Sea Turtle Active Movement Model (STAMM), which was designed by Philippe and Maxime Lalire at CLS (Collecte Localisation Satellites).

      STAMM makes extensive use of Copernicus Marine Service products, especially global current velocity and primary production, which is used to simulate where juvenile turtles may go by the currents and speed the turtle is swimming, and the most favorable habitats. By using AI, improvements can be made to STAMM, in order to generate more accurate models of turtle movements across the oceans.

      
      

  • Topic 4e - Expanding Our Vision Of The Sea Surface
    • Remote sensing data provides a huge number of sea surface observations, but it cannot give direct information on deeper ocean layers, which can only be provided by sparse in situ data. The combination of measurements collected by satellite and in situ sensors represents one of the most effective strategies to improve our knowledge of the interior structure of the ocean ecosystems.

      Neural networks allow us to expand this vision of the sea surface that we get from satellites and in situ observations. This means we can obtain 3D reconstructions of the ocean by using sea surface temperature (SST) and ocean colour surface fields with a multi-layer-perceptron (MLP) network.

      These two variables are of primary importance for many upper-ocean biophysical processes. Artificial neural networks can efficiently model eventual non-linear relationships among input variables, and the choice of the predictors is thus crucial to build an accurate model.

      
      
      

  • Topic 4f – Using ML To Differentiate Between Sediment and Chlorophyll
    • Being able to tell the difference between chlorophyll and sediment is fundamentally important for ocean researchers using EO data. It is very difficult to be able to tell the difference between sediment and chlorophyll from satellite images - particularly in coastal areas – yet the difference can be crucial for monitoring the marine environment.

      Chlorophyll is a pigment in marine phytoplankton (microscopic marine plants) and being able to identify it is important because it plays the same role in the ocean that grass plays on land, as a key part of the food chain and removing CO2 from the ocean.

      Using ML to solve this problem makes marine data as a whole more robust and researchers can be more certain about the results they are getting. For example, in coastal water around the UK there is a lot of sediment, so researchers need to be careful that what they think they are seeing is what they are actually seeing. Using ML to make this differentiation becomes a key part of their workflow.

      Jupyter Notebook Task 1 - ML for Retrieval of Water Quality Indicators: Naive Bayesian

      This Jupyter Notebook workflow guides you through classification of ocean colour data for the derivation of water quality relevant information. It will guide you through the process of training of a naive bayesian classifier on Sentinel-3 OLCI data in order to identify complex waters.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this notebook with your fellow learners, please write in the discussion below.

      Jupyter Notebook Task 2 - ML for Retrieval of Water Quality Indicators: Neural Network

      In this workflow example, we build on the basis that was introduced in the previous step, and introduce an auto-associative neural network approach that can be applied anywhere in the ocean.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      
      

  • Topic 4g - Using ML To Combine Water Quality Data from Different Satellites
    • Measuring chlorophyll concentration is the primary goal of most ocean colour (OC) studies. Sentinel-3 OLCI is the workhorse for this task, as it is the best OC radiometer we have in space. However, OLCI has a 300 meter resolution. Though not designed for marine work, the Sentinel-2 MSI radiometer has 10 meter resolution.

      In this video, Ben Loveday explains how ML can be used to seamlessly combine the two, delivering results that are much better than classical approaches. ML approaches like this are likely to become the mainstay of any coastal OC work. The technique is highly relevant to any researcher or agency that is investigating or monitoring the coastal or inland water domain.

      Jupyter Notebook Task: Chlorophyll Retrievals from Sentinel-2 and Sentinel-3 OLCI data

      For this Jupyter Notebook task, you will have a go at using machine learning for chlorophyll-a retrievals. The ML algorithm used for this task is the Mixture Density Network (MDN) which is trained and validated using a sizeable database of co-located chlorophyll-a measurements and in situ hyperspectral radiometric data resampled to simulate the Multispectral Instrument (MSI) and the Ocean and Land Color Imager (OLCI) onboard Sentinel-2A/B and Sentinel-3A/B, respectively.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below.

      
      

  • Topic 4h - Using Copernicus Data and Machine Learning To Monitor Marine Protected Areas
    • Marine Protected Areas (MPAs) are areas of the ocean that are set aside for long-term conservation aims. They can come in many forms, ranging from wildlife refuges to research areas. Examples of MPAs include Cook Islands Marine Park and Papahānaumokuākea Marine National Monument in the US, and there are around 3150 MPAs in the EU Member States under the Birds and Habitats Directives (SPAs and pSCIs) covering a surface of over 550,000 km², which is about 10% of the total marine area of the Member States.

      It is vital that these areas can be monitored so that we can continue to protect and preserve their marine ecosystem and biodiversity, and one way we can do this is with satellite and in situ data, and machine learning.

      An example of this in action is the Marine Analyst, which is an online marine data access platform to help people access detailed data on marine parameters, particularly in MPAs. It includes Copernicus satellite and in situ data and was developed during a Copernicus hackathon in Tallinn, Estonia, and it is running on WEkEO. It uses machine learning to provide some of the parameters for people to use to monitor the health of marine protected areas.

      Data Insights: In Conversation with Dr Hayley Evers-King

      In this second Data Insight, we hear from Dr Hayley Evers-King, Marine Applications Expert at EUMETSAT.

      
      

  • End of Week 4 DC

• Week 5
  • Topic 5a: Part 1 - Introduction to Monitoring the Atmosphere
    • 
      

      Some of today’s most important environmental concerns relate to the composition of the atmosphere. The increasing concentration of greenhouse gases and the cooling effect of aerosols are prominent drivers of a changing climate, but the extent of their impact is often still uncertain.

      At the Earth’s surface, aerosols, ozone and other reactive gases such as nitrogen dioxide determine the quality of the air around us, affecting human health and life expectancy, the health of ecosystems and the fabric of the built environment. Ozone distributions in the stratosphere influence the amount of ultraviolet radiation reaching the surface. Dust, sand, smoke and volcanic aerosols affect the safe operation of transport systems and the availability of power from solar generation, the formation of clouds and rainfall, and the remote sensing by satellite of land, ocean and atmosphere.

      To address these environmental concerns we need data and processed information. The Copernicus Atmosphere Monitoring Service (CAMS) has been developed to meet these needs, to support policymakers, business and citizens with enhanced atmospheric environmental information.

      CAMS provides consistent and quality-controlled information related to air pollution, and heath, solar energy, greenhouse gases and climate forcing – everywhere in the world.

      Over the last few decades, research by the World Health Organization has established that air pollution is responsible for a broad range of adverse health effects. This evidence, together with environmental protection policies, has stimulated the development of air quality monitoring and forecast services in Europe.

      Like weather forecasts, air quality forecasts are based on specialized Numerical Weather Prediction models, ingesting observations of trace gases and particles, whilst also taking into account natural and anthropogenic emissions. CAMS delivers air quality forecasts every day.

      Atmosphere monitoring from space

      There are various key satellites which play a role in providing the data that we use for monitoring the atmosphere.

      In Europe, EUMETSAT operates the Metop-A, -B and -C satellites, which circle the globe via the poles and continuously collect data from an altitude of 817km. Aboard the Metop satellites the Global Ozone Monitoring Experiment-2 (GOME-2) measures vertical profiles of ozone and other constituents of the atmosphere, such as Nitrogen Dioxide (NO2) and SO2. Also, the IASI instrument on the Metop satellite observes ozone, carbon monoxide (CO) and methane (CH4).

      The next generation of Metop satellites – known as Metop-Second Generation will bring Metop’s valuable contributions into the 2040s, by operating in three successive pairs and carrying enhanced and new instruments, as well as the Copernicus Sentinel-5 instrument. The Sentinel-5 instrument will be exploited in combination with the IASI-NG infrared sounder on board the Metop-SG satellites, to observe the signature of trace gases across the spectrum, from the UV to the thermal infrared.
      
      In geostationary orbit, the Meteosat satellites provide imagery for the early detection of fast developing severe weather and can be used to characterise aerosols in the atmosphere including volcanic ash and dust.The next generation of Meteosat satellites, Meteosat Third Generation (MTG), will be active from the end of 2022 until the 2040s and carry the Copernicus Sentinel-4 instrument to monitor air quality over Europe every hour at high spatial resolution.Sentinel-4 will work with the new Infrared Sounder (IRS) and imager (FCI) on MTG.

      Sentinel-5P is the first Copernicus satellite dedicated to monitoring the atmosphere. The satellite carries the state-of-the-art Tropomi instrument to map a multitude of trace gases such as nitrogen dioxide, ozone, formaldehyde, sulphur dioxide, methane, carbon monoxide and aerosols – all of which affect the air, health, climate.

      
      

  • Topic 5a: Part 2 - Introduction to Monitoring the Atmosphere - The Added Value of Satellites
    • Dr Federico Fierli from EUMETSAT introduces how we can used satellites to monitor the Atmosphere and why this is important for citizens.

      VIDEO

      
      

  • Topic 5a: Part 3 - Introduction to Monitoring the Atmosphere - ML for Weather Modelling
    • Numerical weather prediction (NWP) was first implemented in the 1920s, but it was not until the 1950s when it produced accurate results, and this was because of the introduction of computer simulations.

      There are now many state-of-the art NWP models that are used for weather forecasting. However, these models have the challenging job of representing our complex Earth system. In this video, Peter Duben explains how machine learning can help with this challenge and improve NWP.

      
      

  • Topic 5b: Part 1 - Effects of Air Quality on Human Health and the Environment
    • Poor air quality is a very serious environmental risk to public health around the world. According to the World Health Organization (WHO), there are 4.2 million deaths every year as a result of exposure to ambient air pollution and 91% of the world’s population live in places where air quality exceeds WHO guidelines.

      As well as air pollution, the depletion of the ozone layer has also posed a risk to human health due to increasing the UV levels leading to more cases of skin cancer, cataracts and impaired immune systems.

      Bad air quality also has negative effects on our environment, for example, acid rain can travel hundreds of kilometres causing damage to trees and the pollution of soils and water bodies such as lakes and rivers, harming wildlife and ecosystems.

      Sentinel-5P is the first Copernicus mission dedicated to monitoring our atmosphere. The satellite carries the state-of-the-art TROPOMI instrument to map a multitude of trace gases such as nitrogen dioxide, ozone, formaldehyde, sulphur dioxide, methane, carbon monoxide and aerosols – all of which affect the air we breathe and therefore our health, and our climate. Information from the mission is used through CAMS for air quality forecast and for decision-making. The TROPOMI instrument is a spectrometer that maps the global atmosphere with a resolution as high as 7km x 3.5km. At this resolution, air pollution over cities can be detected.

      Satellite remote sensing, such as from Sentinel-5p, provides information with the required spatial coverage to monitor global air quality. Together with the application of ML algorithms, scientists are able to extract implicit relationships among variables in training datasets to address atmospheric inversion problems.

      In this video, Fabio Del Frate explains about the damaging effects of air pollution, and how we can effectively use satellite remote sensing and artificial intelligence to set up services that can, for example, inform citizens of risks to their health from ambient air pollution and UV.

      
      

  • Topic 5b: Part 2 - Tracking Air Quality
    • Air quality has attracted considerable scientific interest over the last few decades because it both directly and indirectly affects many social and economic activities, but above all it adversely affects human health and wellbeing.

      Tools have been developed to track air quality so that citizens and decision makers have access to real-time air quality data.

      An example of this is Plume Labs, who have created products to help citizens reduce their air pollution exposure with air pollution forecasts down to the street level and Flow, a smart personal air quality tracker. Their air quality tracking technology works by gathering the latest pollution levels from over 12,000 monitoring stations and 80 public agencies around the world, and local specifics such as geographical features, land use, population density, and human activities. They then use AI algorithms and atmospheric models to turn this massive amount of data into hourly forecasts, maps, and more. Generating truly real-time data, and predictions up to seven days in advance.

      Jupyter Notebook Task: AI for Ozone Concentration Estimation

      For this Jupyter Notebook task, you will have a look at the process of training a neural network based on Sentinel-5P/TROPOMI data in order to predict total column ozone concentration.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      
      

  • Topic 5c - Methane Retrievals Using Sentinel-5P
    • Methane is a powerful greenhouse gas and is one of the most significant drivers of global warming. It has an impact on climate change 84 times greater than CO2 (per unit mass, over 20 years). It is also a precursor to the formation of tropospheric ozone, which can have a number of indirect effects on agriculture productivity, crop yields and human health.

      Methane is one of the key targets of Sentinel-5P’s Tropospheric Monitoring Instrument (TROPOMI).

      Current methods of tracking methane perform well but have a large computation cost. Scientists are using Radiative Transfer Model (RTM) simulations, which allow faster and less computationally demanding retrievals.

      In this video, Jennifer Adams explains how AI was used to learn the physics of RTMs for greenhouse gases, in particular for methane retrievals from S5p.

      Jupyter Notebook Task: Methane Retrievals Using Sentinel-5P data

      In this Jupyter Notebook task, you will use physics-based machine learning for Copernicus Sentinel-5P methane retrievals. A neural network will be trained to learn the physics of Radiative Transfer Models (RTMs) for Sentinel-5p methane retrievals, allowing for faster retrievals of methane.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below.

      
      

  • Topic 5d - Estimating Precipitation
    • The prevailing element in the Earth’s water cycle is precipitation – both liquid and solid. This plays a crucial role in the hydrological budget. If we have a better understanding of the water cycle – it will have a positive impact in many different fields, such as: climatology, weather forecasting, agriculture, fishing and flooding early warning systems.

      Despite its importance, precipitation is one of the most difficult geophysical parameters to be quantified, particularly in regions where ground measurements are sparse, or do not exist (such as in the ocean, remote locations or developing countries)

      AI applied to data collected by meteorological satellites can provide a significant improvement to precipitation monitoring in these regions. These are characterized by frequent temporal sampling capabilities over large areas. Areas covered by networks of rain gauges or weather radars can also benefit from satellite precipitation estimation to compensate for the usual irregular spatial distribution of rain gauges or the orography problems that could affect weather radars.

      
      
      

  • Topic 5e - Estimating Emissions During the COVID-19 Pandemic
    • The area of estimating emissions has developed quickly in the context of the COVID-19 Pandemic. AI can be used to infer emissions, or even ‘potential’ emissions – for example what would have been observed, what energy use would have been in the absence of lockdown measures.

      To do this, different types of activity statistics are available, such as: mobility or traffic counts, flight counts and energy network fluxes.

      There has been a recent CAMS study using machine learning, (specifically gradient boosting) to study the impact of Covid on emission reductions in relation to electricity demand.

      
      

  • Topic 5f - ML for More Accurate Weather Forecasts: Part 1 - WeatherBench
    • Accurate weather forecasting is becoming increasingly more crucial to local and national crisis management planning and climate adaptation decision-making, as the frequency of extreme weather events increases due to climate change.

      Currently, weather forecasts are generated using physics-based computer simulations, which are generally effective, but perform poorly when it comes to the prediction of extreme events, and they are also computationally expensive. Data driven approaches, such as deep learning, directly learn from the best available observations and could potentially produce better forecasts.

      To look into this, WeatherBench was set up in order to evaluate deep learning models for global medium-range (i.e. several days to two weeks) weather forecasting.

      In this video, Peter Dueben explains how WeatherBench works and what data is behind it.

      
      

  • Topic 5f - ML For More Accurate Weather Forecasts: Part 2 - MET Norway
    • The Norwegian Meteorological Institute (MET Norway) delivers reliable and accurate weather predictions, from hours to several days into the future, and focuses on Norway, the European Arctic and adjacent areas.

      Their weather models accurately cover a resolution of 20 kilometres and above, so, in order to fill in gaps in the data, they use crowd-sourced in situ weather observations. This is used to improve the accuracy of their weather prediction model and therefore allows them to provide more accurate forecasts.

      In Norway, these in-situ observations are taken from Netatmo smart home weather sensors that are widely owned by the general population. Netatmo’s station density is roughly 50 times greater than MET Norway’s, resulting in temperature forecasts being able to update every hour, compared to the current six-hour frequency.

      
      

  • End of Week 5

• Week 6
  • Topic 6a – Introduction to Climate Monitoring: Part 1
    • Welcome to the last week of the course. For this week, we will be looking at how AI can help scientists to monitor our climate.

      It is very important to study the Earth’s climate so that we can understand climate change and its impacts, so that we can inform governments and decision makers, enabling them to define and implement appropriate mitigation and adaptation policies.

      Satellites, as well as in situ instruments, are an increasingly important tool in monitoring and making decisions about climate change and the environment, and they encompass a wide range of techniques used to map, measure and monitor an enormous variety of environmental parameters and processes on the Earth. With the introduction of AI, we can now optimise the way we monitor the climate.

      
      

  • Topic 6a - Introduction to Climate Monitoring: Part 2
    • The Copernicus Climate Change Service (C3S) aims to support adaptation and mitigation policies of the European Union by providing consistent and authoritative information about climate change. They offer free and open access to climate data and tools based on the best available science and endeavour to help deal with the impacts of climate change. C3S is implemented by ECMWF on behalf of the European Commission.

      The Copernicus Climate Change Service combines observations of the climate system with the latest science to develop authoritative, quality-assured information about the past, current and future states of the climate in Europe and worldwide.

      The portfolio of service products includes:

      • Consistent estimates of multiple Essential Climate Variables
      • Global and regional reanalyses (covering a comprehensive Earth system domain: atmosphere, ocean, land, carbon)
      • Products based on observations alone (gridded; homogenised station series; reprocessed Climate Data Records)
      • A near-real-time climate monitoring facility
      • Multi-model seasonal forecasts
      • Climate projections at global and regional scales

      This wealth of climate information is the basis for generating a wide variety of climate indicators aimed at supporting adaptation and mitigation policies in Europe in a number of sectors such as:

      • Water management
      • Energy
      • Agriculture and forestry
      • Health
      • Tourism
      • Infrastructure
      • Insurance
      • Disaster risk reduction
      • Transport
      • Coastal areas

  • Topic 6b – Mapping Local Climate Zones in Urban Areas
    • By 2050, Berlin summers could be as hot as in Canberra, Australia. Pankow, a district in the city’s north, has already declared a climate emergency in 2019 and is planning ahead. It is planting trees from the Mediterranean that can withstand the heat, and has calculated computer simulations for sunshine and cold air corridors for the construction of 1200 new apartments. A few changes, like swapping asphalt and concrete that store heat against greenery that soaks up water and provides shade, can make a difference on the local scale. Many of these changes on a local scale then make a difference on the bigger scale.

      To understand local climate in cities, scientists have developed the Local Climate Zone classification scheme, as part of the So2Sat project. The aim is to create a 4D urban map of the world.

      It differentiates between 17 zones based mainly on surface structures (such as building and tree density) as well as surface cover (green, pervious soils versus impervious grey surfaces) There are algorithms that calculate these maps from freely available satellite imagery, but there’s still room for improvement by adapting or developing suitable and advanced Convolutional Neural Network (CNN) architectures that generalize well.

      
      

  • Topic 6c – Cloud Classification
    • Clouds play a crucial role in the climate system. They are the source of all the Earth’s precipitation and have a significant impact on the Earth’s radiative budget. Changes in clouds impact the environment and these changes feedback on cloud formation and behaviour.

      FDL’s Cumulo project can tell the difference between different types of clouds – for example, cirrus and cumulus clouds. The project has massive compute costs - and it will be used operationally by institutions such as NASA or the MetOffice.

      Jupyter Notebook Task - Cumulo

      For this Jupyter Notebook task, you will use data from the MODIS instrument aboard Aqua and IRES net model to classify cloud types.

      Clouds are important to classify as a source of uncertainty in future climate projections comes from limitations in modelling clouds and in understanding how different cloud types interact with the climate system.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below!

      
      

  • Topic 6d – ML To Help Navigate in Changing Polar Regions
    • Over the past few decades, temperatures have increased about twice as fast in the Arctic as in the mid-latitudes, a phenomenon known as ‘Arctic amplification’, and floating sea ice cover has been shrinking at an alarming rate, especially during summer. This loss of sea ice is the reason for Arctic amplification, as it is lowering the albedo, a measure of how much a surface reflects solar energy, of the Arctic. When bright and reflective ice melts, it gives way to a darker ocean, which amplifies the warming trend because the ocean surface absorbs more heat from the Sun than the surface of snow and ice.

      FDL’s Cumulo project can tell the difference between different types of clouds – for example, cirrus and cumulus clouds. The project has massive compute costs - and it will be used operationally by institutions such as NASA or the MetOffice.

  • Topic 6e - Mapping Climate Indicators: Part 1 - Mangrove forests and coral reefs
    • 
      

      Coastal mangrove forests sequester a lot of carbon - taking carbon dioxide (CO2) out of the atmosphere and holding onto it. They can sequester more carbon per hectare than any other forests on Earth. They also help protect communities and habitats from storm surges, but sea-level rise could wipe them out.

      Coral reefs are one of the most diverse ecosystems on Earth and some estimates say that over 1 billion people depend on food from coral reefs. They also can create and protect land by dissipating wave energy from storms and tsunamis. However, they are particularly sensitive to increases in ocean temperature - they can be severely affected by even small changes in seawater temperature, which can kill them.

      The World Wide Fund for Nature (WWF) has used drones to train an ML model to be able to detect the extent of corals and mangroves on Sentinel-2 data. Doing this allows them to create change maps to see how the benthic habitats are changing over time - are the protective actions we’re taking actually working? Or is climate change ruining it all?

      
      

  • Topic 6e - Mapping Climate Indicators: Part 2 - Seagrass
    • Seagrass beds are highly-valued components of coastal ecosystems due to their role in nutrient cycling, carbon sequestration, food-web structure and ecosystem services.

      Alterations to seagrass can be an indication of climate change, as increased global temperature causes a change in growth rates of seagrass as well as other physiological functions of the plants themselves.

      In this video, Georgios Sylaios explains how machine learning was used to understand the distribution of seagrass species over the Mediterranean Sea.

      
      

  • Topic 6f – How ML Can Speed Up Our Response to Natural Disasters
    • Extreme weather, sea level rise and other climate change impacts are causing an increase in flooding worldwide.

      Satellite data can be used for flood monitoring, however, it is difficult to interpret optical imagery as floodwaters do not look blue, they are usually brown due to sediment loads. This can hinder the fast response needed for dangerous flood events.

      To combat this, we can use machine learning to process flood images much quicker in order to generate vitally needed flood maps. This was completed by Oxford researchers and ESA’s Phi-lab, who developed the WorldFloods AI model.

      WorldFloods is a training data set that combines, in ‘machine-learning ready form,’ several existing databases of satellite imagery of historical flood events. The dataset contains pairs of Sentinel-2 images and flood extent maps covering 119 global flood events. This is an expert dataset that was labelled by flood experts – which means it has a higher confidence level and a higher accuracy. Typically, this kind of expert labelling takes about six months of work.

      You need experts to label the dataset because it can be hard to identify what is a flood and what isn’t a flood. For example, a flood can look different depending on what colour composite you’re looking at.

      Worldfloods is specifically designed for deployment in specialized hardware in space on low-cost satellites in low Earth orbit. So, in the future, flood segmentation mapping will be able to be run on-board satellites. The new CubeSat satellites such as PhiSat include chips that can run the algorithm – which would significantly speed up response times. This is ultimately what the WorldFloods dataset will be used for.

      Jupyter Notebook Task: ML4Floods: Training

      In this last Jupyter Notebook you will have a go at flood mapping with machine learning using WorldFloods data which combines, in “machine-learning ready form”, several existing databases of satellite imagery of historical flood events. The dataset contains pairs of Sentinel-2 images and flood extent maps covering 119 global flood events.

      ML algorithms have the potential to offer significantly faster and more accurate flood mapping than traditional methods.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      If you have any questions, comments or would just like to discuss this Notebook with your fellow learners, please write in the discussion below!

      Jupyter Notebook Task - ML4Floods: Inference

      This notebook guides you step-by-step through the inference process of a flood extent segmentation model using Sentinel-2 data.

      You can view the video for this tutorial both here on FutureLearn as well as at the top of the Jupyter Notebook.

      For this Jupyter Notebook workflow tutorial please follow this link to go to the WEkEO Jupyter Hub or click on the link in the ‘See Also’ section below.

      
      

  • Topic 6g – Neural Networks and In Situ Observations to Improve Weather Forecasting
    • Weather buoys are instruments which collect weather and ocean data within the world’s oceans.

      In this video, José María García-Valdecasas talks about a project which uses a neural network linked to a network of buoys around the coast of Spain, which is helping to fill in gaps missing from the in situ data that the buoys collect and to improve model forecasts in the area of the in situ instruments. It is currently used for making more accurate predictions of wave patterns for port operations, but it is expected that the same methodology & technology can be used in climate projections or long term hindcasts.

      
      

  • Topic 6h – ML and Ocean Data For Human Health
    • Aquatic ecosystems have undergone complex environmental changes in recent years due to climate change. But climate change is not just an environmental issue, it is also a fundamental threat to human health and well-being.

      Increases in the number of warm days and nights, and the frequency and intensity of both droughts and heavy rainfall events, can have an impact on waterborne diseases. As high temperatures can alter pathogen survival, replication and virulence, heavy rainfall events can mobilize pathogens and compromise water and sanitation infrastructure, and drought can concentrate pathogens in limited water supplies.

      Cholera is a waterborne epidemic disease in humans caused by the bacterial pathogen Vibrio Cholerae. It affects 1.3 to 4 million people globally each year, with 21,000 to 143,000 reported fatalities. The survival of Vibrio pathogens outside a human host in the environment is an important factor of cholera disease transmission and this is being exacerbated by climate change.

      This topic looks at a novel study where machine learning is used to link environmental state to the risk of cholera occurrence. It is a rare example of Copernicus marine data being used to link directly to a possible human health outcome. The question addressed is highly relevant to large numbers of people.

      This data approach can be used by organisations such as the WHO, MSF, NGOs, national health authorities, disaster planning authorities, medical supply chain managers, water management authorities, epidemiologists and the general public.

• Final Course Round-Up
  • Round-Up
    • We have now come to the end of this course on AI for Earth Monitoring. Congratulations on completing the course and thank you for your participation over the last six weeks.

      We hope you have enjoyed this course and that you have gained a useful insight into how AI is advancing the fields of Earth observation and monitoring. Furthermore, we hope that this will help you in your further work, interests and decisions, and we wish you the very best for the future.

      If you have any final remarks or thoughts, we would be delighted to hear them in the comments below!