Use Cases

Permafrost thawing is one of the major concerns on a global scale with a direct contribution to land degradation. It covers approximately 24% of the land from the northern hemisphere and stores about half of the organic carbon on a global scale. Because it is a product of a cold climate, permafrost thawing has the potential to change the chemical and mechanical properties of the soil, transform terrestrial ecosystems into aquatic ecosystems and lead to massive loss of vegetation cover trigger landslides and increases soil erosion by increasing the water flow over the terrain and to generate thermokarst. In this perspective, a regular monitoring system is deemed necessary and extremely useful for the decision-makers in charge of regulating this unavoidable process.

Field sampling for lands degradation monitoring is difficult to achieve in the harsh climate from the Arctic region, especially when the land degradation is assessed from different processes perspectives, like landslide occurrences, thermokarst of terrestrial ecosystems transformations. The seasonal freeze-thaw process can impact the accuracy assessment of the impact of permafrost thawing on land degradation by providing false positives, while detecting from satellite imagery and field studies. Physical modelling in combination with satellite imagery assimilation has proved to help, but still, it remains an open discussion regarding its applications.
EO-PERSIST will use super resolution models applied on satellite imagery to improve the landslide occurrence and thermokarst detection. On the complex data cube created from various datasets available for the Arctic region, long-short term memory artificial intelligence algorithm is going to be used for detecting the general trends in land degradation in the Arctic region.

Three specific applications are going to be explored:

1. Assessing the relationship between the permafrost thawing and landslide occurrences using InSAR analysis and Sentinel-1 imagery along with Long Short-Term Memory artificial intelligence techniques.
2. Mapping areas with possible thermokarst induced by permafrost thawing.
3. Mapping the transformation of terrestrial ecosystems into aquatic ones using use Sentinel-1 and Sentinel-2 imagery.

Climate change poses a serious threat to the operations of industrial sites located in the Arctic region. Permafrost variations resulting in temperature fluctuations and anomalies have a huge impact on the standardized operations of the industrial sites, but also on their surrounding environments which in return can cause security, health and economic issues.

Many applications of thermal remote sensing for industrial monitoring and risk assessment have been done in the past years, using historical EO data. These kind of applications are very important to the industry, since analysis of EO data can provide evaluation of hazard levels or prevent extensive damages before the accident happens.

In this use case, EO-PERSIST will develop a downscaling temperature time series and anomalies methodology fusing time series of thermal (Landsat series, MODIS, Sentinel-3), optical (Sentinel-2) and meteorological models (ERA-5) with state of art deep learning architectures. approaches can be categorized in terms of the type of data needed to train the model. This use case will be focused on two basic categories, recognized also in the framework of Industrial Risk Applications:
1. Site monitoring/risk assessment for chemical gases and plume detection and/or identification.
2. Areas monitoring/risk assessment for gas, hot water and high voltage transportation infrastructure
leak detection.

Permafrost coastlines are highly vulnerable to climate change ecosystems covering 34% of the total global coastlines ‘extent. Monitoring permafrost coastlines spatiotemporal changes are mainly due to erosion caused by internal and external factors linked to climate change such as the global warming, sea level rise, air temperature, reduced sea ice cover and increased storm intensity. The monitoring and understanding of the observed changes of permafrost coastlines in Artic have been a topic of rising concern and of key research priority to be addressed by the global community, particularly so in the framework of future IPCC climate change scenario projections. In this context, it has become more relevant than ever the examination of permafrost coastal sites socio-economic impacts using climatic data projections from IPCC.

Exposure assessment implications related to changes in structural population size, migration, land use and changes in population distribution. EO has gained high importance in mapping and monitoring permafrost coastline dynamics, particularly so since the launch of sophisticated imaging sensors by various Space Agencies such as ESA’s Sentinel missions. The latter, combined with the developments in hardware technology (e.g. HPC, cloud-computing) and GIS-based systems and geospatial analysis methods has clearly opened up new opportunities towards the monitoring of permafrost shores.

In this use case, EO-PERSIST will use advanced geospatial modeling approaches (including multivariate models, raster gap-filling approaches) to map coastline changes of Arctic permafrost areas. Shorelines will be extracted in a GIS environment using the latest algorithms. Furthermore, long-term analysis of coastline erosion or accretion rates will be performed also through GIS techniques (e.g. CHET tool). Accuracy assessment of the permafrost coastlines change monitoring will be performed using standardized approaches considering variables that affect permafrost coastal erosion31. These variables will be defined by LU. A time-series analysis of the EO data will allow extracting information on the coastal dynamics over a longer period of study. Advanced 3-way coupled modeling system CHAOS (Chemical Hydrological Atmospheric Ocean wave System) will be used to assess the impact of the Arctic permafrost change to the prevailed atmospheric conditions.

Moreover, in this use case EO-PERSIST will assess the economic impact of permafrost changes allowing the estimation of changes in infrastructures, healthcare employment, income, and levels of business activity that may result from the implementation of the different use cases. Spatial information, regional and microeconomic data will be used to predict the economic impact assessment. Subject to available secondary data, integrated assessment econometric models will allow EO-PERSIST to evaluate economic losses of permafrost changes under combined climate and socio-economic scenarios. Econometric models will be developed to predict the socioeconomic assessment of permafrost changes considering variables such as population density, per capita income, land use, farming production etc. These methods will allow projecting the costs and benefits of governmental agencies that are likely to occur as a result of the climate change scenarios. The qualitative assessment of the social impact will be based on discourse analysis in media and relevant social media, qualitative online synchronous interviews, focus groups and Transgressive Action Research (TAR). TAR addresses the gap for more inclusive and participatory forms of knowledge creation and engagement, where research participants are co-researchers gathering visual and audio data such as videos and photos that capture their perspectives and visions of resilience, attractiveness and well-being of their coastal area of residence.

Monitoring of the active layer freeze-thaw cycle is considered an essential indicator for the state of the permafrost environments. Due to the harsh environmental conditions, many permafrost areas are not accessible and most of the observation sites are located at easy-to-access locations. According to ESA`s Permafrost project and several recent studies21, TSInSAR (Time series Interferometric Synthetic Aperture Radar) is a cost-effective and powerful technique able to reveal ground deformation patterns that are mainly related to seasonal active layer freeze-thaw cycle. A

To our knowledge, the full potetial of the TSInSAR methodologies over permafrost regions has not been fully exploited yet. Specifically, the unwrapping of the interferometric phase was performed only in spatial dimension without exploiting the temporal behavior of the interferometric phase. Also, the interferometric phase information of the distributed scatterers is not fully exploited. Spatiotemporal unwrapping which can vastly improve the quality of the interferometric observations and exploitation of the phase information of distributed scatterers can significantly improve the spatial detail and accuracy of ground deformation patterns23. In this UC, we will develop an innovative methodological pipeline that will include the missing components from the state-of-the-art approaches through the expertise and cooperation of our multidisciplinary team by exploiting multi-source datasets

The main objective of this UC is to develop an innovative TSInSAR algorithm for permafrost regions, able to a) exploit interferometric measurements of distributed scatterers and b) exploit the spatiotemporal information in the unwrapping procedure. In order to achieve this, we propose the following innovative methodological steps:

1. Identify pilot areas in permafrost regions that our methodological pipeline will be developed and validated (NTUA, FMI, CloudFerro). This includes the collection of available SAR and other datasets required for the development and the validation of the algorithm (CloudFerro). Based on the availability
of EO datasets (NTUA) and in-situ observations (FMI), some pilot areas will be selected. We point out that we will consider pilot areas with diverse geophysical characteristics in order to ensure the transferability of the developed algorithm (FMI).

2. Next we will apply state-of-the-art approaches based on current studies and projects (see section 1.1.2) (NTUA). This step output will be used as the baseline for the comparison with our developed method.

3. In order to ensure the innovative aspect of the developed algorithm, information from GNSS (YETITMOVES) and atmospheric models (LU) will be utilized for the development and the validation phase of the algorithm. The output of this step is going to be an TSInSAR approach which will exploit information from distributed scatterers and include spatiotemporal unwrapping functionalities (NTUA).

4. In this step we will retrieve information for the active layer freeze-thaw cycle inferred from TSInSAR results. FMI will be responsible for inferring variables related to active layer freeze-thaw cycle from InSAR results. Error assessment with available in-situ measurements will be performed (NTUA, FMI).

EO-PERSIST will develop innovative TSInSAR algorithm for permafrost regions, able to a) exploit interferometric measurements of distributed scatterers and b) exploit the spatiotemporal information in the unwrapping procedure. In order to achieve this, the following innovative methodological steps will be implemented:
1. Identify pilot areas in permafrost regions for the development and validation of this Use case.
2. Next we will apply state-of-the-art approaches based on current studies and projects utilizing, information from GNSS and atmospheric models (LU).
4. Retrieval of information for the active layer freeze-thaw cycle inferred from TSInSAR results

Snow cover has a notable impact on soil freezing and thawing (F/T) in northern latitudes due its insulating properties; consequently, the presence of snow cover, as well as the timing and duration of soil freezing conditions can be linked to permafrost active layer dynamics. The insulation properties of snow are largely determined by the thickness and density of the snowpack (i.e. snow mass), which can be expressed as the Snow Water Equivalent (SWE). Passive microwave (MW) sensors provide currently the only global remote sensing source of information for SWE. Spanning several decades, time series of passive MW observations, assimilated with in situ observations, have been used to derive the magnitude and trend of SWE over the Northern Hemisphere (Figure 7). However, the spatial resolution as well as the thematic accuracy of stand-alone retrievals from passive MW observations still fail to meet the requirements of many operational applications in e.g. hydrology.

Physical and hemispheric scale snow models can provide SWE estimates which are on par in accuracy with remote sensing retrievals. It has also been suggested that physical models be used in data assimilation schemes with EO in order to achieve improved precision, but there remain significant questions regarding the wider applicability of these schemes. On the other hand, the surface F/T state of soils can be detected directly by currently operational active and passive MW sensors. Compared to passive MW sensors, SAR, provide spatial resolutions of up to less than 1 meter, enabling to derive small scale spatial variations in e.g. freezing properties. However, e.g. the state-of-the art Copernicus Sentinel-1 platform provides observations only every 6 days in Central Europe (now 12 days with the failure of Sentinel-1B in December 2021). Furthermore, Sentinel-1, as well as other similar sensors (e.g. the RadarSat Constellation Mission, RCM) operate at C-band which is less well suited for soil freeze detection, being susceptible to changes in the overlying vegetation.

Together these features may result in missed detection of freezing in areas experiencing e.g. crop damages from ephemeral (short-lived) freezing events. Several future low frequency SAR missions promise to increase the capability for soil F/T and SWE detection from active MW observations. The NASA-ISRO SAR mission (NISAR) and Copernicus Radar Observation System for Europe in L-band (ROSE-L) missions, in particular, have the objective of providing L-band observations over land surfaces suitable for the retrieval of both parameters. While the soil F/T state can be derived from the backscatter coeffiient, SWE retrievals will exploit interferometric pairs of images separated by a temporal baseline to estimate SWE accumulation30. The sensors promise unprecedented accuracy and spatial resolution for both SWE and soil F/T retrievals. These may also be exploited in downscaling schemes with existing passive MW sensors as well as physical snow distribution models.

The objective in this UC is to develop and test capabilities of L-band SAR imagery for retrieving both soil F/T state and SWE, with the eventual aim of exploiting these to investigate their impact on permafrost active layer dynamics.
The primary source of remote sensing information will be from ALOS2 and Sentinel-1 observations. Time series of ALOS2 imagery are available for several test sites via a ESA-JAXA collaboration where FMI is participating, while all Sentinel-1 data are distributed publicly. In the absence of L-band observations for some sites, we will apply the C-band observations from Sentinel-1. To this end we propose the following methodological steps:

1. Down select representative test sites where data on permafrost active layer dynamics are available. A reference site experiencing seasonal soil freezing and thawing but no permafrost will be included.
2. Consolidate algorithms for retrieving soil F/T state and SWE from SAR data (applicable to L- and C-bands)
3. Generate time series of SWE and soil F/T state from ALOS2 and Sentinel-1 for the selected test sites
4. Generate physical snow distribution model estimates for same sites for comparison and evaluation
5. Collect and consolidate in situ observations for evaluating EO and snow distribution model estimates
6. Use the EO-PERSIST platform for uploading SWE products