PREVENT Remote Sensing Theory (UOWM)

Remote Sensors Theory

PREVENT Project

Start

Introduction

"Remote sensing is the science and art of obtaining information about an object, area, or phenomenon from the analysis of data acquired by a device that is not in contact with the object, area, or phenomenon being examined (Lillesand et al. 2003)."There are many ways to define remote sensing in the literature. Two common elements characterize this scientific field: a) the concept of "information collection" and b) "observation from a distance (Campbell, 1987)

Index

Activities

Objectives

Evaluation

Modules

Survey

Remote sensing means acquiring information about an object without direct contact with it

-Gupta, (1991)-

Objectives

Our main goal is to provide you with the tools and knowledge you need to understand your Remote Sensing knowledge, whether you are starting from scratch or want to improve your existing skills. Throughout the course, you will be immersed in interactive lessons, stimulating case studies and practical exercises designed to reinforce your understanding and application of key concepts. Upon completion of the course, you will be equipped not only with a solid theoretical understanding, but also with the confidence and ability to tackle real-world challenges in Remote Sensing. Get ready to unleash your potential and reach new levels of success in your professional or academic career!

'Remote sensing differs from direct observation or measurement in that, in the latter, the specific observation instrument is within or in contact with the object being measured or investigated, such as a thermometer. '

+ info

Modules

Module 2

Module 1

Module 4

Module 3

Advanced Research and National ContributionsThis module focuses on the research in deep learning and remote sensing.

Remote Sensing Systems and TechniqueThis module focuses on the various satellite systems, their capabilities in remote sensing.

Foundations of Remote SensingThis section introduces the basic principles, historical development, benefits and limitations of remote sensing

Applications and Analysis in Remote SensingThis section emphasises the applications and analysis of their role in monitoring earth changes.

01

History of Remote Sensing

Satellite shots of floods in Deggendorf, Germany (before and after) 2024 https://www.euspaceimaging.com/blog/2024/07/01/satellite-imagery-for-emergency-management/

01

Summary

Although it is quite difficult to precisely define the starting point of remote sensing and its evolutionary phases, five main stages are distinguished during its development (Barrett & Curtis, 1992; Swain & Davis,1978). While Remote Sensing (RS) is its own field, it often acts as a complement to GIS analyses, adding unique information and analysis techniques. There are two types of RS, active and passive and they are generally used for different applications. Active RS involves sending out a signal and waiting for its return to the sensor. RADAR and LIDAR are examples of active RS, as they send out energy, microwave and laser pulses respectively and record the signals as they bounce back (Gennarelli G. & Catapano, 2022) Despite its significant advances, remote sensing faces some challenges that limit its effectiveness in specific applications. Errors in the recorded data can arise from various causes leading to discrepancies between actual and recorded values. In addition, understanding the complex interactions between the recorded radiation and the target objects proves difficult, as these relationships are affected by highly variable environmental factors. The huge spatial and temporal variations in the atmosphere, lithosphere and hydrosphere, combined with the complex energy-matter interaction mechanisms, further complicate the accurate analysis and interpretation of remote sensing data.

Historical Overview

1925-1945

pre 1925

1945-1955

1955-1960

1960-present

Advances in photo interpretation techniques. Focused on analysis over applications

Advanced sensors revolutionize remote sensing. Environmental monitoring expands rapidly

Development of stereoscopic aerial photography. Widely used in WWII for topographic mapping

Early aerial photography experiments. Used for strategic mapping

Applications in geology and agriculture expanded. Aerial photographs gained widespread use

Types of Remote Sensing

Passive

Active

Active RS involves sending out a signal and waiting for its return to the sensor. Examples: RADAR, LIDAR

Examples Landsat Series (NASA/USGS) Sentinel-2 (ESA).

Remote Sensing Advantages

Remote sensing is considered a modern, specialized tool that finds applications in many scientific subjects, including environmental sciences, forestry, geology, archaeology, oceanography, etc.

Summary Coverage

Accessibility

Repeated Coverage.

Data Homogeneity

Multispectral Data Characteristics

Recording Time Duration

Digital Data Form

Data Cost

Remote Sensing Disadvantages

Recorded Data

Nature and Mechanism

One of the most serious problems concerns the recorded data, which contain errors due to atmospheric conditions, the topography of the area and the functioning of the satellite system. These errors are related to differences between the actual and the satelliterecorded reflected radiation. It is very important for the further successful pplication of these data to adapt the data as closely as possible to the actual values, especially in cases where the analysis concerns temporal studies (Barrett and Curtis, 1992; Richards, 1993).

Another problem that scientists face is the difficulty in understanding the nature and mechanism of even very simple relationships and interactions between the radiation recorded by the satellite and the target objects. The changes that occur in the conditions of the atmosphere, lithosphere and hydrosphere are so large in spatial and temporal dimensions and the mechanisms of interaction between energy and matter are so complex, that simple object-detector relationships are difficult to determine (Barrett and Curtis, 1992; Sabins, 1997)

02

Remote Sensing Systems and Techniques

02

Summary

This section refers to the different types and missions of satellites used to monitor and study the Earth. First, optical satellite systems such as LANDSAT, IKONOS, World View, QuickBird, Pleiades and SPOT are described, providing detailed data for applications such as mapping, agriculture and natural disaster management. We then look at active satellite systems using radar, such as the ERS and Envisat satellites. These satellites provide data for environmental monitoring and climate change, while TerraSAR-x and Cosmo-Skymed are used for applications such as natural disaster management and environmental resources. In addition, hyperspectral satellites, exemplified by EO-1 and the Hyperion sensor, are being considered, which collect data in multiple spectral bands for detailed analysis of materials on the Earth's surface. Finally, we focus on the Copernicus programme, a European Union initiative to monitor the planet and the environment.

Optical Satellite Systems

MODIS instrument that flies aboard NASA’s Terra satellite revealed the most powerful thunderstorms, Source: https://blogs.nasa.gov/hurricanes

Remote Sensing

LANDSAT

Terra Satellite

Optical

IKONOS

Satellite Systems

ERS

Active

TerraSAR-x

Hyperspectra

EO-1 and Hyperion

Landsat

Technical Details

Landsat 3

Landsat 2

Landsat 1

Landsat 6

Landsat 5

Landsat 4

Landsat 9

Landsat 7

Landsat 8

ERS

Objectives

MISSION

ERS Instruments

ERS TOOLS

Copernicus Program

Copernicus is the Earth observation segment of the European Union’s Space program, dedicated to monitoring our planet and its environment for the benefit of all European citizens. It provides information services derived from satellite Earth Observation and insitu (non-space) data.

Next

Copernicus Program

Large volumes of global data from satellites and ground-based, airborne and seaborne measurement systems supply information that assists service providers, public authorities and other international organizations in enhancing the quality of life for European citizens and beyond.

Sentinel 1

Sentinel 2

Sentinel 3

Sentinel 4-5

Sentinel 6

Sentinel-1A and 1B, covers the entire Earth every six days.

Sentinel-2 has a high-resolution multispectral imager with 13 spectral bands.

Sentinel-3 complements the mission of Sentinel 2

Sentinel 4 and 5 focus on monitoring atmospheric composition.

Sentinel-6 involves two satellites: Sentinel-6A and Sentinel-6B

Sentinel High Level Operations Plan (HLOP)

link

What is the Copernicus Programme?

Climate change refers to long-term shifts in temperatures and weather patterns. Human activities have been the main driver of climate change, primarily due to the burning of fossil fuels like coal, oil and gas.

03

This image originally appeared in the NASA Earth Observatory Story: Running Through Paris Heat, Data acquired June 7, 2024. https://visibleearth.nasa.gov

Applications and Analysis in Remote Sensing

03

Summary

Remote sensing plays a vital role in monitoring and mitigating natural disasters. Soil moisture estimation using active and passive sensors aids in predicting floods, landslides, and water resource management. Techniques like thermal infrared and microwave sensing provide accurate soil moisture data, critical for environmental studies. Mapping land cover and use, supported by satellite imagery, offers fast and comprehensive spatial data for resource management and urban planning. Techniques such as DInSAR and Persistent Scatterers Interferometry enable precise measurement of surface deformations caused by earthquakes and volcanoes. Thermal sensors on satellites like MODIS and Landsat enhance active volcano monitoring, contributing to disaster preparedness and response. Satellite imagery is increasingly integrated into post-earthquake damage assessments, enabling faster rescue operations and improving disaster response strategies.

Recording and Earth Observation Equipment

Recording equipment is characterized by their spatial, radiometric, spectral and temporal resolution. These characteristics determine the quality and accuracy of the collected data.

Temporal Resolution:

Spectral Resolution:

Radiometric Resolution:

Spatial Resolution:

Refers to the frequency with which a satellite system captures images of the same area. This is important for monitoring changes on the Earth's surface over time.

Related to the number of spectral channels used by the satellite. More spectral channels allow the distinction of more types of surface features.

Describes the system's ability to record differences in the intensity of reflected or emitted energy. This resolution is often expressed in bits.

Refers to the equipment's ability to detect small features on the Earth's surface. The smaller the pixel, the greater the spatial resolution.

Digital Image Analysis

Digital image processing is a process that involves converting analog signals into digital values. These values are used to store and analyze data from satellites or other recording systems. This data is recorded in binary numbers (bits), representing the intensity of the brightness of the scanned surface. Images, also known as raster data, is used for easy management of pixel values by processing software. Although raster data is convenient for digital analysis, it presents difficulties in depicting discrete areas or points compared to vector data.

'Digital images are two-dimensional arrays of small areas called pixels.'

The arrangement of pixels in rows and columns allows processing software to analyze the data, providing useful information on specific topics. Satellite image processing software manages remote sensing data and offers analysis capabilities for scientific and commercial applications. Some of the most common software includes ERDAS IMAGINE, ArcGIS, ENVI, IDRISI and Geomatica.

Analysis of Optical Satellite Images

The process of analyzing satellite images includes several stages for ensuring accurate and useful results. These stages include image restoration and preprocessing, image enhancement, image classification and the interpretation of satellite images and digital aerial photographs.

Stage 3

Stage 2

Stage1

Image Classification

Image Enhancement

Image Restoration or Preprocessing

Remote Sensing for Natural Disaster Monitoring and Mitigation

Soil Sensing

Mapping Land Cover

Soil deformation

Recording and assessment of post-earthquake damage

Soil Sensing

Determination of soil moisture content using active and passive sensors from Space

Soil moisture is a critical variable for understanding water resources and natural hazards such as floods and landslides. Accurate estimation of spatial and temporal variations in moisture is important for environmental studies and improving flood forecasts, especially in medium and large drainage basins where floods often cause disasters. The condition of surface soil moisture is critical for assessing rainwater infiltration or runoff, so this information is vital for flood prediction models. Moreover, soil moisture in mountainous areas is a key factor for landslides. Since the 1970s, research in this field has used techniques that cover the entire electromagnetic spectrum, from optical to microwave.

'Digital images are two-dimensional arrays of small areas called pixels.'

Remote sensing techniques use different wavelengths, energy sources, and sensor responses to estimate soil moisture. Solar radiation measures reflected sunlight, while microwaves and thermal infrared are more commonly used. Thermal methods, like thermal inertia and temperature/vegetation index, are particularly effective in areas with little or no vegetation. These techniques enhance natural hazard prediction and water resource management.

Mapping Land Cover

Mapping land cover and land use is vital for scientific research, spatial planning and natural resource management. "Land cover" refers to the physical characteristics of the surface, while "land use" describes how the land is utilized. A land use plan captures the interaction between humans and the environment, impacting significant economic activities. Advances in satellite monitoring systems have made remote sensing techniques realistic and attractive tools for natural resource research and management. Land use maps are particularly useful in agriculture and natural resource management and their updating is necessary due to the continuous changes in natural resources and human activities.

'Digital images are two-dimensional arrays of small areas called pixels.'

Remote sensing offers fast and accurate landscape representation, providing data in digital form and across a wide range of spectral channels. Although it does not completely replace field observations, it significantly reduces the time and cost of data collection, offering a comprehensive view of the study area. It also facilitates access to remote areas and monitoring changes in land use/cover through temporal data

Monitoring and assessment of soil deformation

Active sensors use phase difference to measure Earth's surface deformation through the technique of differential interferometry (DInSAR). This technique uses data from Synthetic Aperture Radar (SAR) instruments. Currently, there are several satellites suitable for obtaining InSAR data, such as the European Space Agency's Envisat satellites, the Canadian Radarsat-1 and 2, the German TerraSAR-X and the Italian Cosmo-Skymed satellites, along with many planned future SAR missions.

'Digital images are two-dimensional arrays of small areas called pixels.'

Remote sensing offers fast and accurate landscape representation, providing data in digital form and across a wide range of spectral channels. Although it does not completely replace field observations, it significantly reduces the time and cost of data collection, offering a comprehensive view of the study area. It also facilitates access to remote areas and monitoring changes in land use/cover through temporal data

Recording and assessment of post-earthquake damage

Recording and assessing damage after an earthquake is critical, especially when the destruction is geographically extensive or in remote areas. This process is important for rescue teams and civil protection services and must be done quickly and accurately. The "change detection" technique using pre- and post-earthquake images is a fast method for assessing damage. As space technologies improve, they are increasingly integrated into disaster management actions. However, support with space techniques remains largely misunderstood by most emergency and civil protection services.

'Digital images are two-dimensional arrays of small areas called pixels.'

Remote sensing offers fast and accurate landscape representation, providing data in digital form and across a wide range of spectral channels. Although it does not completely replace field observations, it significantly reduces the time and cost of data collection, offering a comprehensive view of the study area. It also facilitates access to remote areas and monitoring changes in land use/cover through temporal data

Observation of active volcanoes using thermal radiation

Although there is no satellite designed exclusively for volcanic applications, many thermal sensors for military, urban and industrial applications can be adapted for volcano monitoring. For example, weather satellites provide data for examining volcanic hot spots, although meteorological sensors typically measure lower temperatures. The turn of the new millennium brought new NASA satellites, such as Terra, Landsat-7, Aqua and EO-1, which enable thermal recording of active volcanoes.

'Digital images are two-dimensional arrays of small areas called pixels.'

This established the first global satellite monitoring system for volcanoes, monitoring all active and potentially active volcanoes daily. Thermal data from satellite sensors are increasingly accessible for example, MODIS and Landsat data are available at no cost. Thermal remote sensing has provided valuable insights into volcanic behavior, despite the absence of specialized sensors.

04

Advanced Research and National Contributions

04

Summary

Remote sensing plays a vital role in monitoring and mitigating natural disasters. Soil moisture estimation using active and passive sensors aids in predicting floods, landslides, and water resource management. Techniques like thermal infrared and microwave sensing provide accurate soil moisture data, critical for environmental studies. Mapping land cover and use, supported by satellite imagery, offers fast and comprehensive spatial data for resource management and urban planning. Techniques such as DInSAR and Persistent Scatterers Interferometry enable precise measurement of surface deformations caused by earthquakes and volcanoes. Thermal sensors on satellites like MODIS and Landsat enhance active volcano monitoring, contributing to disaster preparedness and response. Satellite imagery is increasingly integrated into post-earthquake damage assessments, enabling faster rescue operations and improving disaster response strategies.

Advanced Research and National Contributions

Greek Initiatives

Research

Methodologies

Greek Initiatives in Natural Disaster Management

Research and Advances in Deep Learning and Remote Sensing for Natural Disaster Management

Technologies and Advanced Techniques in Deep Learning and Remote Sensing

The process of analyzing satellite images includes several stages for ensuring accurate and useful results. These stages include image restoration and preprocessing, image enhancement, image classification and the interpretation of satellite images and digital aerial photographs.

Methodologies in Deep Learning and Remote Sensing

CNNs

GANs

RNNs

CD

SSMs

10

DMCNN

SAR

Geo-Computational Techniques

UAVs

EO

Here you can put an important title

Your content is good, but it‘ll engage much more if it’s interactive. Capture your audience's attention with an interactive photo or illustration.

Research and Advances in Deep Learning and Remote Sensing for Natural Disaster Management (1/3)

Deep Learning and Remote Sensing technologies have been widely integrated within disaster risk management systems in recent years, thanks to the growing availability of high-quality data/products and advanced systems for their analysis. Remote Sensing is progressively supporting disaster mapping and monitoring because it can allow quick and accurate physical observation of the earth’s surface before, during and after disasters.

Park et al. (2022)

Long et al. (2021)

Psomiadis et al. (2019)

Wang et al. (2021)

Goldberg et al. (2020)

Stephenson et al. (2022)

Elmahdy et al. (2020)

Dinh et al. (2022)

Barmpoutis et al. (2020)

Peng (2022)

Taskin et al. (2022)

Hakim et al. (2022)

Here you can put an important title

Your content is good, but it‘ll engage much more if it’s interactive. Capture your audience's attention with an interactive photo or illustration.

Research and Advances in Deep Learning and Remote Sensing for Natural Disaster Management (2/3)

Deep Learning and Remote Sensing technologies have been widely integrated within disaster risk management systems in recent years, thanks to the growing availability of high-quality data/products and advanced systems for their analysis. Remote Sensing is progressively supporting disaster mapping and monitoring because it can allow quick and accurate physical observation of the earth’s surface before, during and after disasters.

Dinh et al. (2022)

Zhang et al., (2022)

Asaly et al. (2022)

Akhyar et al. (2023)

Feng et al. (2022)

Chen et al. (2023)

Ananias et al. (2022)

Garcia et al. (2023)

Sun et al. (2022)

Kim & Muminov (2023)

Lee (2022)

Jia & Ye (2023)

Here you can put an important title

Your content is good, but it‘ll engage much more if it’s interactive. Capture your audience's attention with an interactive photo or illustration.

Research and Advances in Deep Learning and Remote Sensing for Natural Disaster Management (3/3)

Deep Learning and Remote Sensing technologies have been widely integrated within disaster risk management systems in recent years, thanks to the growing availability of high-quality data/products and advanced systems for their analysis. Remote Sensing is progressively supporting disaster mapping and monitoring because it can allow quick and accurate physical observation of the earth’s surface before, during and after disasters.

Ntinopoulos et al. (2023)

Shastry et al. (2023)

Li et al. (2023)

Yang et al. (2024)

Wu et al. (2024)

Sundriyal et al. (2024)

Greek Initiatives in Natural Disaster Management

In Greece, various initiatives and projects have been implemented to enhance natural disaster management using advanced technologies.

Maestro

SEASFire

Ofire+

GET

DISARM

Activities Show what you know!

Activity 2

Activity 1

Click on the correct consepts

Matching Questions

Activity 1

1. Copernicus Program
2. Sentinel-1
3. Sentinel-2
4. Sentinel-3
5. Sentinel-4 and 5
6. Sentinel-6

Activity 2 (1/2)

Solution

Click on the correct concepts

Choose 4 terms

CONCEPT 1

After an earthquake, remote sensing techniques are used to assess the extent and impact of the damage, including destroyed infrastructure, surface ruptures, and changes in terrain. Choose 4 terms that relate to post-earthquake damage assessment:

Seismic Wave Detection

Change Detection

Damage Classification

Displacement Mapping

Synthetic Aperture Radar

Optical Imagery

Persistent Scatterers

Unmanned Aerial Vehicles

Activity 2 (2/2)

Solution

Click on the correct concepts

Select 4 correct answers

CONCEPT 2

Which combination of technologies would you choose to effectively assess the earthquake damage in a densely populated urban area? Select the 4 correct answers from the given options.

Recurrent Neural Networks

Convolutional Neural Networks

Unmanned Aerial Vehicles

Synthetic Aperture Radar

Semantic Segmentation Net

Generative Adversarial Net

Maestro Telemetry System

Persistent Scatterers

Evaluation

In this section, you will have the opportunity to test your acquired knowledge throughout the course. Our interactive quiz will provide a detailed assessment of your understanding of key topics. Get ready to challenge your skills and reinforce your learning as you move towards mastering the fundamental concepts. Don't miss the chance to demonstrate everything you've learned so far!

1/10

2/10

3/10

4/10

5/10

6/10

7/10

8/10

01:00

9/10

10/10

Course completed!

Hakim et al. (2022)

Investigate the volcanic landforms in the Hantangang River Volcanic Field (HRVF), highlighting their geoheritage value. The study employs 3D printing of a terrain model and Q-LavHA simulation to estimate a prehistorical eruption from two source vents in North Korea. Artificial neural network (ANN) and support vector machine (SVM) algorithms are used to classify the lava area, with SVM demonstrating higher accuracy and efficiency. The single eruptive vent scenario showed better accuracy than the Q-LavHA simulation, but multiple vent scenarios improved the overall accuracy.

Technical Details

Launch Date: March 1, 1984 Status: Decommissioned January 2013 Sensors: TM, MSS Altitude: 705 km Inclination: 98.2° Orbit: polar, sun-synchronous Equatorial Crossing Time: nominally 9:45 AM (± 15 min.) local time (descending node) Period of Revolution : 99 minutes; ~14.5 orbits/day Repeat Coverage : 16 days

1945-55

This period is characterized by the development of photo-interpretation methods and techniques. Emphasis was placed on analysis and photo interpretation methods rather than their applications.

Close

Convolutional Neural Networks

Unmanned Aerial Vehicles

Synthetic Aperture Radar

Semantic Segmentation Net

Objectives

ERS was the first ESA programme in Earth observation to provide microwave spectrum-based environmental monitoring. The missions’ range of instruments were capable of monitoring the land, oceans, and atmosphere, and more specifically sea ice, geology, forestry, wave phenomena bathymetry, meteorological events and many more scientific fields.

Earth Observation (EO) Technologies

Involves using satellite and airborne data to monitor and assess natural and human-made disasters. EO technologies are used for post-disaster response, damage assessment, recovery and mitigation, roviding detailed, real-time data for better decisionmaking. roviding detailed, real-time data for better decision making.

Description

Methods: Divided into unsupervised and supervised methods. Unsupervised classification: Identifies natural spectral groups automatically without external information. Supervised classification: Uses samples of known identity to guide the classification. Algorithms: Include minimum distance and maximum likelihood. Accuracy evaluation: Classification accuracy is evaluated by comparing results with reference data. Results: Produces a thematic map that can be integrated into GIS

Unmanned Aerial Vehicles (UAVs)

Drones equipped with high-resolution cameras and sensors provide real-time data from disaster sites. UAVs are used for search and rescue operations and post-disaster damage assessment, quickly surveying large areas and locating survivors.

Drone video of the Palisades devastation

Wu et al. (2024)

Present a study on post-flood disaster house damage classification using a Dual-View Convolutional Neural Network (DV-CNN) model, published in Sustainable Cities and Society. This model, incorporating ResNet-50, transfer learning and the Concentration-Based Attention Module (CBAM), enhances the efficiency and generalization of damage assessment. Validated with data from the "July 20 Heavy rainstorm in Zhengzhou," the DV-CNN achieved a 92.5% accuracy in classifying damage levels, outperforming other models. The study underscores the model's reliability and generalizability, offering a valuable reference for post-flood damage assessment in rural houses.

Zhang et al., (2022)

Explore the use of high-spatial-resolution thermal infrared data to predict earthquakes using a "heating core" filter. Their study addresses two primary gaps: the variation of thermal anomalies with different earthquake magnitudes and the challenge of spotty thermal anomaly distributions in high-resolution data. The research involves resampling data, applying a "heating core" filter to isolate seismic-related thermal anomalies and using time–distance–magnitude windows for correlation. Results show no significant difference in thermal anomalies across earthquake magnitudes and the model can predict earthquakes within 200 km and 20 days of anomaly appearance. This binary prediction model offers a valuable reference for earthquake prediction.

Sundriyal et al. (2024)

Employ an integrated machine learning and remote sensing approach to evaluate landslide hazards and risk hotspots in the NW Himalaya. Utilizing a Multilayer Perceptron (MLP) to generate a landslide susceptibility map, combined with rainfall intensity data, the study produces a comprehensive hazard map. Incorporating land use and land cover data, the resulting risk map indicates that approximately 5% of the area falls in high-risk zones. The research highlights that around 53% of the population resides in high to very high landslide risk areas, emphasizing the need for sustainable development and urban planning in this vulnerable region.

Garcia et al. (2023)

Propose a new CNN framework for the semi-automatic detection of relict landslides in rainforest areas, using a dataset generated by a k-means clustering algorithm with a pre-training step to fine-tune the CNN training process. The study compares the proposed framework with standard methods using CBERS-04A WPM images and tests 42 combinations of three CNNs (Unet, FPN and Linknet) and two augmented datasets. The framework shows higher recall rates, but precision remains low due to false positives. Despite limitations in detecting relict landslides due to spectral similarities with non-landslide areas, the framework demonstrates improved accuracy in landslide detection.

Shastry et al. (2023)

Develop a deep learning model for mapping floods using Maxar WorldView imagery, addressing the challenge of surface obstruction by clouds and vegetation. Their approach involves creating a semantically labeled dataset representative of North American surface water variability, which is used to train a convolutional neural network (CNN) within the Deep Earth Learning, Tools and Analysis (DELTA) framework. The model achieves high precision (98%) and recall (94%) during validation. However, a comparison with hydraulic models reveals a 62% underprediction of flood extents, largely due to obstructions, with 74% attributable to vegetation and 9% to clouds. The study underscores the need to combine flood models with remote sensing data for accurate inundation mapping.

Psomiadis et al. (2019)

Describe a synergistic approach using remote sensing and GIS techniques for flash-flood monitoring and damage assessment in the Thessaly Plain area, Greece. The study focuses on a flash flood event in May 2016, employing Landsat-7 ETM+ and Sentinel-1 SAR images to detect flooded areas. Various water indices and a threshold method were applied to Landsat-7 and Sentinel-1 data, respectively. Additionally, high-resolution DEM and Sentinel-2 images were used to refine inundation delineation, estimate flood water depth and assess land use/cover of the flooded regions. This integrated approach successfully delineated flooded areas and evaluated the financial impact on affected cultivations, demonstrating the effectiveness of combining optical and radar data with GIS modeling for accurate flood mapping and damage evaluation.

Sentinel 2

Its mission includes monitoring vegetation, soil conditions and water resource management. The Sentinel-2 mission objectives include:

1. Observing land, vegetation, soil and water cover, inland waterways, coastal areas
2. Mapping changes in land cover
3. Managing food security
4. Monitoring forests
5. Determining various plant indexes (leaf area chlorophyll, water content

6. Providing information on water pollution 7. Assisting in disaster relief (flood, volcanic eruption and landslide imaging) 8. Monitoring climate change

Description

Techniques: Include contrast enhancement, filtering, principal component analysis and spectral band combination. Visual improvement: Enhances visual interpretation of digital images. Contrast enhancement: Increases grey level gradients. Filters: Low and high-frequency pass filters enhance or eliminate image details.

Sun et al. (2022)

Developed a forest fire susceptibility model using the Light Gradient Boosting Machine (LightGBM) algorithm to produce an accurate fire susceptibility map. Focusing on a subtropical national forest park in Jiangsu, China, the study used eight variables derived from topographic, climatic, human activity and vegetation factors. The LightGBM model was compared with logistic regression (LR) and random forest (RF) models. Results showed that temperature was the primary fire driving factor, with LightGBM outperforming LR and RF in F1-score, accuracy (88.8%) and AUC (0.935), demonstrating superior predictive performance.

Goldberg et al. (2020)

Utilize operational satellite observations to map and monitor floods caused by ice jams and snowmelt, particularly in high-latitude regions. The study employs the Visible Infrared Imaging Radiometer Suite (VIIRS) and the Advanced Baseline Imager (ABI) to estimate water fractions through mixed-pixel decomposition, deriving flood maps from water fraction differences pre- and post-flooding. The high temporal resolution of ABI and the detailed spatial resolution of VIIRS enable effective observation of ice movement, snowmelt status and flood water evolution, aiding in flood prediction and dynamic monitoring. The JPSS and GOES-R flood products uniquely include suprasnow/ice flood types and snow/ice masks, enhancing their utility for river forecasters and wide-end users.

Ananias et. al (2022)

Introduce the Algal Bloom Forecast (ABF) framework for predicting algal blooms in inland water bodies using machine learning and remotely sensed data. The fully automated ABF framework leverages MODIS images, environmental data and spectral indices to build anomaly detection models with SVM, RF and LSTM methods. Case studies in Erie (USA), Chilika (India) and Taihu (China) lakes demonstrate the framework's effectiveness. The RF model within the ABF framework achieved the best predictions, evaluated through metrics such as global accuracy, kappa coefficient, F1- Score and R2-Score.

Copernicus Program

The European Commission oversees the Programme, which is executed in collaboration with 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), various EU Agencies, Mercator Ocean, the European Environment Agency (EEA) and the Joint Research Center (JRC).

Technical Details

Launch Date: April 15, 1999 Status: operational despite Scan Line Corrector (SLC) failure May 31, 2003 Sensors: ETM+ Altitude: 705 km Inclination: 98.2° Orbit: polar, sun-synchronous Equatorial Crossing Time: nominally 10 AM (± 15 min.) local time (descending node) Period of Revolution : 99 minutes; ~14.5 orbits/day Repeat Coverage : 16 days

Sentinel-6

Sentinel-6 involves two satellites: Sentinel-6A was launched in November 2020, while Sentinel-6B is set to launch in 2025. Its primary mission is to provide high-precision information on the topography of the ocean worldwide, while its secondary mission is radio occultation for climate change monitoring and weather forecasting. For these purposes, the Sentinel-6 is equipped with a Synthetic Aperture Radar Altimeter (POSEIDON-4) and a GNSS-RO respectively. The Sentinel-6 has the ability to map up to 95% of the Earth’s ocean every 10 days, with the gathered information complementing the oceanic data from Sentinel-3. Based on its missions, Sentinel-6’s objectives include:

1. Monitoring changes in sea level
2. Forecasting for operational oceanography
3. Providing information on ocean currents, wind speed and wave height for maritime safety
4. Protecting and managing coastal zones

link

What is Remote Sensing?

• Understanding of Basic Remote Sensing Principles

Analysis of the fundamental concepts of remote sensing, along with its advantages and disadvantages.

• Historical development of remote sensing

Review of the evolution of the technology, from the early use of photography to the modern application of satellite systems.

• Use of active and passive systems

Presentation of active and passive remote sensing techniques, with emphasis on RADAR and LIDAR systems for remote sensing.

• Natural Disaster Management through Remote Sensing

Applications of remote sensing in monitoring and response to natural disasters such as floods, earthquakes and fires.

• Satellite Image Analysis

Synthesis and analysis of satellite images through digital processing and categorization techniques for scientific use.

• Integration with Geographic Information Systems (GIS)

Use of GIS for the analysis and presentation of remote sensing data, focusing on their application to natural disaster management.

• Greek Initiatives in Disaster Management

Presentation of Greek projects that use remote sensing for monitoring and management of natural disasters, such as SEASFire, Maestro telemetry system and DISARM.

A Landsat Timeline | Landsat Science

Landsat 8 successfully launched on Feb. 11, 2013 and the Landsat data archive continues to expand. Landsat 5 delivered high-quality, global data of Earth’s land surface for 28 years and 10 months, officially setting a new Guinness World Record title for “Longest-operating Earth observation satellite.”

Technical Details

The DISARM project,

is a part of the Interreg Balkan-Mediterranean 2014-2020 program, focuses on creating an early warning system for fire risk in Greece, Cyprus and Bulgaria. Developed by Geospatial Enabling Technologies, the system uses data from meteorological stations and satellite information to predict fire risks. The platform includes temperature, humidity, wind speed and specialized fire risk indices, along with satellite data on active fires and weather forecasts. The website and tools were built using opensource software and international standards .

Park et al. (2022)

Address the increasing need for rapid detection and monitoring of natural disasters due to climate change. They emphasize the role of remote sensing techniques in managing disasters across multiple spatial and temporal domains. Despite the challenges in developing robust monitoring and assessment methods for complex disaster mechanisms, recent advances in satellite, airborne and ground remote sensing, along with novel image analysis techniques, offer promising solutions. Their study highlights the integration of various remote sensing data for comprehensive disaster monitoring and assessment, aiming to mitigate disaster risks

Deep Learning-Based Change Detection (CD)

Utilizes multi-temporal remote sensing imagery and deep learning to detect changes on the Earth's surface. This approach helps in updating land use, assessing natural hazards and analyzing urban sprawl by automatically learning and adapting to high-level feature representations.

Geo-Computational Techniques

Includes the integration of GIS, LIDAR, UAVs and advanced computational techniques like machine learning and deep learning. Used for modeling, visualization and prediction of natural hazards at local to global scales, enhancing disaster prediction and management.

Convolutional Neural Networks (CNNs)

CNNs are the cornerstone of image-based deep learning models. They have been extensively applied in the analysis of satellite and aerial imagery to detect and classify disaster-affected areas. For instance, post-disaster damage assessment using CNNs has proven effective in identifying destroyed buildings and infrastructure with high accuracy.

Dinh et al. (2022)

Evaluate the performance of various optimizers for Deformable-DETR in assessing natural disaster damage. Utilizing deep learning techniques and UAV remote sensing, the study enhances disaster response by improving the efficiency and convergence time of Deformable DETR, a Transformer-based object detection method. The researchers analyze several optimizers to enhance Deformable DETR's performance, demonstrating its adaptability and effectiveness for rapid building damage assessment in disaster scenarios.

TerraSAR-X and TanDEM-X Objectives

Flying in close formation, the objective of the TerraSAR-X and TanDEM-X satellites is to simultaneously image Earth's terrain from different angles with unprecedented accuracy for research and development purposes as well as scientific and commercial applications

Technical Details

Launch Date: March 5, 1978 Status: put into standby mode: March 31, 1983; decommissioned: Sept. 7, 1983 Sensors: RBV, MSS Altitude: nominally 900 km Inclination: 99.2° Orbit: polar, sun-synchronous Equatorial Crossing Time: nominally 9:42 AM mean local time (descending node) Period of Revolution : 103 minutes; ~14 orbits/day Repeat Coverage : 18 days

Earthquakes release energy from the Earth and transfer this energy to the surface, causing significant effects in the affected areas. Satellite images provide a unique tool for capturing these areas after seismic events, offering a rapid damage assessment. Damage assessment after an earthquake must consider the following parameters: 1. Spatial resolution: The analysis of space data in relation to the characteristics of built-up areas (dense or sparse construction). 2. System repeatability: The frequency with which the satellite images the same area. Although a system may have high repeatability, acquiring data in real or near real-time is critical. 3. Detection and recognition capability: This capability is particularly important in sparsely built areas.

Sentinel 3

Sentinel-3's primary objective is to measure sea surface topography, land and sea surface temperature and ocean color. It provides high-accuracy data to support environmental and climate forecasting and monitoring. The Sentinel-3 complements the mission of Sentinel 2 with objectives such as:

1. Determining sea surface topography and height, as well as significant wave height
2. Measuring ocean and land surface temperature
3. Determining ocean and land surface colour
4. Mapping sea and land ice topography
5. Monitoring sea and inland water quality, pollution and biological productivity
6. Modelling climate change
7. Identifying changes in land use
8. Mapping forest cover
9. Detecting wildfires
10. Providing indices of vegetation state
11. Forecasting weather and climate
12. Measuring Earth’s thermal radiation for atmospheric applications

Semantic Segmentation Networks

Uses CNNs to identify and locate areas of interest within satellite imagery. Essential for analyzing natural disasters such as wildfires, floods and hurricanes by providing precise and accurate damage assessments, enhancing response efforts and resource allocation.

EarthExplorer

The EarthExplorer (EE) user interface is an online tool developed by the United States Geological Survey (USGS) for search, discovery, and ordering. It enables users to search satellite, aircraft, and other remote sensing inventories using interactive and text-based query capabilities.

Recurrent Neural Networks (RNNs)

RNNs, particularly Long Short-Term Memory (LSTM) networks, are used for time-series prediction, making them suitable for forecasting natural disasters. Studies have demonstrated their capability in predicting flood events by analyzing temporal sequences of rainfall data and river discharge rates.

Peng, (2022)

Reviews recent achievements and challenges in remote sensing approaches for meteorological disaster monitoring. The key issues identified include task arrangement, information extraction and multi-temporal change detection. Accurate monitoring requires determining timescales, sensor planning and constructing representation models. Extracted information is then processed and compared over time to detect disaster evolution. While successful applications exist, gaps remain in process monitoring. Future research on sensor planning, information representation and multisource data fusion is needed to enhance the monitoring and understanding of meteorological disasters.

Close

Change Detection

Damage Classification

Synthetic Aperture Radar

Unmanned Aerial Vehicles

IKONOS, World View, QuickBird, Pleiades and SPOT Satellites

Its capabilities include capturing a 3.2m multispectral, Near-Infrared (NIR) 0.80-meter panchromatic resolution at nadir. Its applications include both urban and rural mapping of natural resources and natural disasters, tax mapping, agriculture and forestry analysis, mining, engineering, construction, and change detection. It can yield relevant data for nearly all aspects of environmental study.

Feng et al. (2022)

Investigate the uncertainties in machine learning models for assessing earthquake-induced landslide susceptibility. Using Artificial Neural Network (ANN), Random Forest (RF), Support Vector Machine (SVM) and Logistic Regression (LR), the study evaluates model uncertainty through susceptibility zoning, risk area statistics and the area under the ROC curve. The results show that landslides tend to cluster spatially, with real landslide occurrences in high-risk areas at 86% for SVM, 87% for RF, 82% for LR and 61% for ANN. The ROC area for RF, SVM, LR and ANN is 90.92%, 80.45%, 73.75% and 71.95%, respectively. Accuracy declines when predicting landslides from different earthquakes.

Yang et al. (2024)

Present a study on a refined fire detection and band selection method in hyperspectral remote sensing imagery using sparse-VIT. This research, published in Infrared Physics & Technology, aims to enhance fire detection accuracy by leveraging advanced hyperspectral imaging techniques. Conducted by researchers from the Shanghai Institute of Technical Physics, the University of Chinese Academy of Sciences and the Beijing Institute of Remote Sensing Information, the study demonstrates significant improvements in remote sensing applications for fire monitoring and management, providing a robust framework for future developments in the field.

Elmahdy et al. (2020)

focus on flash flood (FF) susceptibility mapping and prediction in the northern United Arab Emirates (NUAE), employing a hybrid approach integrating machine learning and geohydrological models. The study tests three machine learning models: boosted regression tree (BRT), classification and regression trees (CART) and naive Bayes tree (NBT). The BRT model demonstrated the highest performance, evaluated using precision, recall, F1 score and ROC curve. The FF susceptibility map was further refined by dividing it into seven basins and calculating new FF conditioning parameters. Results indicated that mountainous and narrow basins like RAK, Masafi, Fujairah and Rol Dadnah have the highest FF occurrence and magnitude, while wider alluvial plains like Al Dhaid have the lowest. This approach enhances the accuracy of FF susceptibility mapping, providing a valuable tool for disaster management in arid regions.

Li et al. (2023)

Integrate the SLIDE model within CAESAR-Lisflood to quantitatively model the ‘rainfall-landslide-flash flood’ disaster chain mechanism under landscape evolution in mountainous areas. Applied to the Wenchuan earthquake-stricken area, this integrated model predicts landslide susceptibility under extreme rainfall and identifies high-risk areas such as mountain valleys and steep gullies. The study highlights the significant influence of landslide legacy effects on erosion and deposition processes, enhancing the model’s applicability for disaster management and reduction in earthquake-affected regions.

Lee, (2022)

Investigates forest fire trends and characteristics in North Korea using remote sensing techniques and digital topographic data. The study analyzes MODIS data from 2004 to 2015 and Landsat data to estimate burned areas in South Hamgyong Province (SHP) and Gangwon Province (GWP). Findings indicate that fires in SHP are more frequent and severe, particularly in coniferous areas, which are more susceptible to fire due to their combustible resin. Large fires predominantly occur on windward open slopes, with fewer fires in shallow valleys and high ridges. The study highlights the need for North Korea to develop measures against large fire damage.

Barmpoutis et al. (2020)

Provide a comprehensive review of early forest fire detection systems using optical remote sensing technologies. The study addresses the escalating threat of large-scale forest fires, emphasizing the need for effective prevention, early warning and response strategies. It categorizes fire detection systems into three types: terrestrial, airborne and spaceborne and evaluates various flame and smoke detection algorithms used by these technologies. The review highlights the strengths and weaknesses of each system, aiming to guide future research in developing more accurate and reliable early warning fire systems to mitigate the impacts of forest fires on the environment and human life.

The analysis of a time series of SAR images extends the possible applications of interferometry, allowing the detection of small displacements on the order of a few millimeters and reducing error sources. There are two main techniques:

a. The SBAS technique (Berardino P. and Sansosti, 2002), which requires many images to create multiple simple interferograms. Through a processing procedure, these interferograms allow precise recording of deformation. b. The Persistent Scatterers (PS) technique, which also requires a large number of SAR images and focuses on ground features that remain stable over time (Ferretti A. & Rocca, 2001). This technique provides point information on deformation, mainly from human constructions and bare rocks.

Land cover changes are detected via remote sensing by analyzing variations in radiation values. Satellite images provide spatial insights, large-area coverage, and temporal data for monitoring dynamic phenomena. Creating thematic land cover/use maps involves three stages: pre-processing, enhancement, and classification. Classification assigns pixel properties to land categories, and its accuracy is critical. Satellite imagery over time enables change monitoring. Supervised classification relies on statistical data and precise class definitions, requiring expertise for effective mapping.

Asaly et al. (2022)

Explore the detection of earthquake precursors using remote sensing technologies and machine learning methods. They apply a support vector machine (SVM) technique to GPS ionospheric total electron content (TEC) time series data to identify potential earthquake precursors. After filtering out solar and geomagnetic influences, their method achieved 85.7% accuracy for true negative predictions and 80% for true positive predictions for large earthquakes (Mw > 6). The model's performance is validated with various skill scores, including an accuracy of 0.83, precision of 0.85, recall of 0.8, Heidke skill score of 0.66 and true skill statistics of 0.66.

Akhyar et al. (2023)

Provide a comprehensive review of deep learning methodologies, particularly convolutional neural networks (CNNs), used for natural disaster management systems. The study highlights the use of semantic segmentation networks to analyze satellite imagery and remote sensing data for disaster evaluation and response. Despite the effectiveness of models like SegNet, U-Net, FCNs, FCDenseNet, PSPNet, HRNet and DeepLab in tasks such as forest fire delineation, flood mapping and earthquake damage assessment, challenges remain in retaining spatial information and optimal feature representation. This review underscores the importance of extracting features from multiple levels of semantic representation to enhance disaster management efforts.

Kim & Muminov, (2023)

Propose an advanced YOLOv7 model for detecting forest fire smoke using UAV images. The model enhancements include incorporating the CBAM attention mechanism, adding an SPPF+ layer for better focus on smaller smoke regions and introducing decoupled heads for effective data extraction. A BiFPN is used for multiscale feature fusion, with learning weights to prioritize critical feature mappings. Tested on a dataset of 6500 UAV images, the proposed approach achieved an AP50 of 86.4%, outperforming previous detectors by 3.9%, demonstrating its efficacy in early wildfire smoke detection.

Technical Details

Launch Date: January 22, 1975 Status: removed from operational status: February 5, 1982; decommissioned: July 27, 1983 Sensors: RBV, MSS Altitude: nominally 900 km Inclination: 99.2° Orbit: polar, sun-synchronous Equatorial Crossing Time: nominally 9:42 AM mean local time (descending node) Period of Revolution : 103 minutes; ~14 orbits/day Repeat Coverage : 18 days

Greek GIS company GET

The final datasets and detected flood polygons are displayed dynamically via the GET SDI Portal, showcasing the potential of Sentinel-1 data in crisis management. The RGB image created from Sentinel-1 data on February 11, 2018 (pre-crisis) and February 23, 2018 (crisis) highlights flooded areas in red due to the contrast in backscatter values. The flood polygons were derived by applying a global threshold to the difference between the pre-crisis and crisis images

https://www.getmap.eu/company/?lang=en

The final datasets and detected flood polygons are displayed dynamically via the GET SDI Portal, showcasing the potential of Sentinel-1 data in crisis management. The RGB image created from Sentinel-1 data on February 11, 2018 (pre-crisis) and February 23, 2018 (crisis) highlights flooded areas in red due to the contrast in backscatter values. The flood polygons were derived by applying a global threshold to the difference between the pre-crisis and crisis images

Sentinel 1

The launch

Date: Sentinel-1A - 03 April 2014 Sentinel-1B - 25 April 2016 Sentinel-1C - 5 December 2024 Site: Kourou, French Guiana Rocket: Sentinel-1A and -B on Soyuz Sentinel-1C on Vega-C

Technical Details

Launch Date: October 5, 1993 Status: lost at launch Sensor: ETM

Stephenson et al. (2022)

Introduce a novel deep learning-based approach for damage mapping using InSAR coherence time series to improve the separation of disasterinduced damage from other surface changes. This method leverages recurrent neural networks (RNNs) to analyze the full time history of synthetic aperture radar (SAR) observations, detecting anomalous variations in surface properties.

Technical Details

Launch Date: July 16 , 1982 Status: decommissioned, June 15, 2001 Sensors: TM, MSS Altitude: 705 km Inclination: 98.2° Orbit: polar, sun-synchronous Equatorial Crossing Time: nominally 9:45 AM (± 15 min.) local time (descending node) Period of Revolution : 99 minutes; ~14.5 orbits/day Repeat Coverage : 16 days

About Terra

Approximately the size of a small school bus, the Terra satellite carries five instruments that take coincident measurements of the Earth system:

1. Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)
2. Clouds and Earth’s Radiant Energy System (CERES)
3. Multi-angle Imaging Spectroradiometer (MISR)
4. Measurements of Pollution in the Troposphere (MOPITT)
5. Moderate Resolution Imaging Spectroradiometer (MODIS)

Jia & Ye, (2023)

Conduct a comprehensive review of deep learning (DL) applications in Earthquake Disaster Assessment (EDA), analyzing 204 articles to explore the current state, development and challenges. They categorize EDA objects into disaster objects (earthquakes and secondary disasters) and physical objects (buildings, infrastructure and areas). The study examines the use of remote sensing, seismic and social media data in EDA, highlighting their advantages and limitations. It also evaluates six DL models, including CNN, MLP, RNN, GAN, TL and hybrid models, across different earthquake stages (pre-, during-, post- and multi-stage). CNNs are notably prominent for image classification in assessing building damage. The review identifies challenges in training data and DL models, suggesting opportunities in new data sources, multimodal DL and emerging concepts, providing valuable insights for researchers and practitioners.

1925-45

This period saw extensive use of aerial photography, mainly for topographic mapping using stereoscopic aerial photographs. World War II greatly boosted the improvement of aerial photography technology.

Sentinel 4 and 5

Sentinel-4 is a future geostationary mission which monitor key air quality trace gas concentrations and aerosols over Europe to support services pertaining to air quality applications and climate protocol. The Sentinel-4 and Sentinel-5 missions focus on monitoring atmospheric composition. The Sentinel-4, the Sentinel-5P and the Sentinel 5 were conceived to complement Copernicus Atmosphere Monitoring Service (CAMS), which provides aggregated information on worldwide air pollution, health, solar energy, greenhouse gases and climate forcing. The mission objectives of the Sentinel-4 include:

1. Measuring air quality
2. Monitoring stratospheric ozone
3. Measuring solar radiation
4. Monitoring climate change

link

1955-60

Aerial photographs became more popular and their applications expanded beyond topographic mapping to include geology, agriculture, environment, forestry, archaeology, etc.

Synthetic Aperture Radar (SAR)

Uses radar to create detailed images of the Earth's surface, even through cloud cover and in darkness. SAR is used for earthquake damage assessment and volcanic eruption monitoring, providing critical data for early damage assessment and predicting eruptions.

Deep Multi-Instance Convolutional Neural Networks (DMCNN)

A new deep learning model designed for disaster classification in high-resolution remote sensing images. Detects and classifies various natural disasters, offering robust disaster management capabilities by accurately identifying and classifying affected regions.

Taskin et al. (2022)

Present an ensemble deep learning architecture based on shared blocks for shallow landslide susceptibility mapping. This approach aims to address limitations in model variance and generalization. By combining Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) models, the ensemble model (CNN-RNN-LSTM) was applied to Trabzon province, Turkey. The ensemble achieved the highest modeling performance with an accuracy of 0.93, surpassing individual models. The ensemble model improved overall accuracy by up to 7% and demonstrated a significant enhancement (~4%) in susceptibility map accuracy, as confirmed by the Wilcoxon signed-rank test.

ERS Instruments:

• The Along-Track Scanning Radiometer (ATSR) measured sea-surface temperatures and cloud-top temperatures.
• The Global Ozone Monitoring Experiment (GOME) was a nadir-scanning ultraviolet and visible spectrometer for global monitoring of atmospheric ozone.
• The Microwave Radiometer (MWR) measured the integrated atmospheric water vapour column and cloud liquid water content, as correction terms for the radar altimeter signal.
• The Radar Altimeter (RA) was a Ku-band (13.8 GHz) nadir-pointing active microwave sensor designed to measure the time return echoes from ocean and ice surfaces.
• The Synthetic Aperture Radar (SAR) wave mode provided two-dimensional spectra of ocean surface waves.
• The Wind Scatterometer (WS) obtained information on wind speed and direction at the sea surface for incorporation into models, global statistics and climatological datasets.
• The Laser Retro Reflector (LRR) was a passive device used as a reflector by ground-based SLR stations using high-power pulsed lasers.
• The Precise Range And Range-Rate Equipment (PRARE) was a compact, space-borne, two-way, two-frequency microwave satellite tracking system.

Description

Critical stage: Prevents amplification of imperfections in digital processing. Atmospheric corrections: Necessary for accurate results. Geometric corrections: Transforms images into maps for GIS use. Conversion process: Involves converting image coordinates to a cartographic projection system using polynomial algorithms and control points.

The SEASFire project

focuses on developing a real-time forest fire monitoring and management system. It integrates satellite data, UAVs and ground sensors to provide comprehensive situational awareness and support for fire suppression efforts.

The system aims to enhance decision-making capabilities for forest fire management through advanced technologies and data analytics. By leveraging these tools, SEASFire seeks to improve the efficiency and effectiveness of firefighting operations, ultimately reducing the impact of forest fires. Led by the National Observatory of Athens and funded by ESA, the project aims to predict seasonal wildfire patterns in Europe using advanced deep learning models and Earth Observation data. The project focuses on understanding the spatiotemporal connections between atmospheric conditions and fire regimes (“SeasFire –Earth System Deep Learning for Seasonal Fire Forecasting in Europe,” 2024).

1960-Present

This stage is characterized by the active development of satellites and sensors. In 1960, the first meteorological satellite was launched, marking a new era of intense activity and research in remote sensing. During this period, some satellite recording systems, initially developed exclusively for military purposes, began to be widely used, as more advanced systems were developed for military applications.

Wang et al., (2021)

Identify the Separable Channel Attention Network (SCANet) as a promising deep technology for managing natural disasters, particularly landslides. SCANet leverages a Poolformer encoder and SCA-FPN decoder to enhance the accuracy of landslide detection from remote sensing images. By improving pixel-level prediction and reducing computational complexity, SCANet significantly outperforms existing methods, aiding rapid rescue and ecological restoration efforts post-disaster.

The Maestro telemetry system

developed by Aristotle University of Thessaloniki, is designed to predict and manage forest fires. It involves placing wireless sensors in forest areas to collect data of temperature, humidity, wind speed, flame and smoke detection. This data, transmitted via the internet, is used to predict fire risks and inform civil protection plans. The system will be tested in a controlled fire at the university's farm to evaluate sensor sensitivity.

Ofire+ from OMIKRON SA

is an innovative system designed to enhance operational readiness for managing wildfire crises in municipalities and regions. It provides a combination of scientific and technological solutions, offering daily meteorological data, fire ignition points, fire behavior simulations, vegetation fire characteristics and a fire weather index. The system includes a web/cloud-based management application for municipal authorities and a mobile app for personnel, volunteers, residents and visitors, allowing autonomous use for individual properties..

Funded by the EU and national resources, it aims to improve crisis management and response (Omikron S.A., 2024)

https://ofireplus.com/

Ntinopoulos et al. (2023)

Explore the fusion of remotely-sensed fire-related indices for wildfire prediction using artificial intelligence. The study employs Artificial Neural Networks (ANN) and Radial Basis Function Networks (RBF) to predict wildfire ignitions in Greece, utilizing indices such as the Fire Weather Index (FWI), Fosberg Fire Weather Index (FFWI), Normalized Difference Vegetation Index (NDVI) and Normalized Difference Moisture Index (NDMI). Additionally, they introduce a new index, "Vegetation-Enhanced FWI" (FWIveg), combining FWI with NDVI information. Developed through the Google Earth Engine platform, this index is optimized using a genetic algorithm. The methodology's robustness was demonstrated by predicting the 2018 Mati wildfire in Attica, Greece, showing the effectiveness of integrating FWIveg with neural networks for fire prediction.

Generative Adversarial Networks (GANs)

GANs are employed to enhance the resolution of remote sensing images, thereby improving the accuracy of disaster impact assessments. They generate high-resolution images from low-resolution inputs, which are crucial for detailed analysis in disaster-stricken areas.

MISSION

The ERS-1 and ERS-2 satellites, launched in the 1990s, use synthetic aperture radar (SAR) for Earth mapping. Envisat, launched in 2002, provided data for monitoring the environment and climate changes until 2012.

Long et al. (2021)

Present a study on using deep learning for emergency monitoring of high-level landslide disasters in the Jinsha River area. Combining satellite remote sensing images with various landslide-inducing factors, the study establishes two detection models: Deep Belief Networks (DBN) and Convolutional Neural-Deep Belief Networks(CDN). The models' performance is analyzed based on parameters such as the number of neurons and learning layers, with DBN and CDN achieving detection accuracies of 97.56% and 97.63%, respectively. This research demonstrates the feasibility of these models in accurately monitoring landslides, providing valuable insights for disaster management in the region.

Dinh et al. (2022)

Evaluate the performance of various optimizers for Deformable-DETR in the context of natural disaster damage assessment. Leveraging recent advances in computer vision and UAV remote sensing, the study aims to enhance disaster response by improving damage detection methods. Deformable DETR, an enhancement of the Transformer-based DETR object detection method, is examined for its efficiency and convergence time. The researchers analyze multiple optimizers to boost the performance of Deformable DETR, demonstrating its suitability and effectiveness for rapid damage assessment in disaster scenarios.

Technical Details

Launch Date: July 23, 1972 Status: expired, January 6, 1978 Sensors: RBV, MSS Altitude: nominally 900 km Inclination: 99.2° Orbit: polar, sun-synchronous Equatorial Crossing Time: nominally 9:42 AM mean local time (descending node) Period of Revolution : 103 minutes; ~14 orbits/day Repeat Coverage : 18 days

Chen et al. (2023)

Propose a K-Net-based hybrid semantic segmentation method for extracting lake water bodies, addressing inefficiencies and dangers in traditional methods. This approach introduces dynamic semantic kernels to iteratively refine feature information, significantly enhancing extraction accuracy from remote sensing images. Validation on a Google dataset demonstrates the model's superiority, with the UperNet +K-Net model using Swin-l achieving the highest mean intersection over union (mIoU) of 97.77%. The incorporation of the K-Net module consistently improves mIoU across all tested models.

Before 1925

This period is characterized by experimentation with applications of photography from balloons and airplanes for topographic mapping. From the start, these photographs highlighted the value of aerial photography, especially during World War I when they were used to locate and map strategic positions.