Most of the processes studied by geoscientists are characterized by variations in both space and time. These spatio-temporal phenomena have been traditionally investigated using linear statistical approaches, as in the case of physically-based models and geostatistical models. Additionally, the rising attention toward machine learning, as well as the rapid growth of computational resources, opens new horizons in understanding, modelling and forecasting complex spatio-temporal systems through the use of stochastics non-linear models.
This session aims at exploring the new challenges and opportunities opened by the spread of data-driven statistical learning approaches in Earth and Soil Sciences. We invite cutting-edge contributions related to methods of spatio-temporal geostatistics or data mining on topics that include, but are not limited to:
- advances in spatio-temporal modeling using geostatistics and machine learning;
- uncertainty quantification and representation;
- innovative techniques of knowledge extraction based on clustering, pattern recognition and, more generally, data mining.
The main applications will be closely related to the research in environmental sciences and quantitative geography. A non-complete list of possible applications includes:
- natural and anthropogenic hazards (e.g. floods; landslides; earthquakes; wildfires; soil, water, and air pollution);
- interaction between geosphere and anthroposphere (e.g. land degradation; urban sprawl);
- socio-economic sciences, characterized by the spatial and temporal dimension of the data (e.g. census data; transport; commuter traffic).

Co-organized by GM2/HS12/NH8/NP4/SSS12
Convener: Federico AmatoECSECS | Co-conveners: Fabian GuignardECSECS, Luigi LombardoECSECS, Marj Tonini
| Fri, 08 May, 16:15–18:00 (CEST)

Files for download

Session materials Download all presentations (118MB)

Chat time: Friday, 8 May 2020, 16:15–18:00

Chairperson: Federico Amato
D2167 |
| Highlight
Aoibheann Brady, Jonathan Rougier, Bramha Dutt Vishwakarma, Yann Ziegler, Richard Westaway, and Jonathan Bamber

Sea level rise is one of the most significant consequences of projected future changes in climate. One factor which influences sea level rise is vertical land motion (VLM) due to glacial isostatic adjustment (GIA), which changes the elevation of the ocean floor. Typically, GIA forward models are used for this purpose, but these are known to vary with the assumptions made about ice loading history and Earth structure. In this study, we implement a Bayesian hierarchical modelling framework to explore a data-driven VLM solution for North America, with the aim of separating out the overall signal into its GIA and hydrology (mass change) components. A Bayesian spatio-temporal model is implemented in INLA using satellite (GRACE) and in-situ (GPS) data as observations. Under the assumption that GIA varies in space but is constant in time, and that hydrology is both spatially- and temporally-variable, it is possible to separate the contributions of each component with an associated uncertainty level. Early results will be presented. Extensions to the BHM framework to investigate sea level rise at the global scale, such as the inclusion of additional processes and incorporation of increased volumes of data, will be discussed.

How to cite: Brady, A., Rougier, J., Vishwakarma, B. D., Ziegler, Y., Westaway, R., and Bamber, J.: Spatio-temporal decomposition of geophysical signals in North America, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-16232, https://doi.org/10.5194/egusphere-egu2020-16232, 2020

D2168 |
Sina Nakhostin, Jeanphilippe Malet, Mathilde Desrues, and David Michea

With the progress of Machine vision and Image processing applications and the access to massive time series of images (terrestrial imagery, satellite imagery) allowing the computation of displacement fields, challenging tasks such as detection of surface motion became achievable. This calls for fast, flexible and automated procedures for modeling and information retrieval.

While supervised learning paradigms have been finding extended application within the field of remote sensing, the scarcity of reliable labeled data to be used within the training phase, sets a noticeable limitation for the generalization of these procedures in the shadow of huge spatial, spectral or temporal redundancy. Although this downside can to some extent be ameliorated by enriching training samples through active learning techniques, relying merely on supervised approaches, is a hindrance while analyzing large stacks of remote-sensing data. In addition, the process of information retrieval becomes more challenging when the data is not the direct acquisition of the scene but other derivatives (after applying different image processing steps) of it. Modeling of the motion maps and extracting high-level information from them and/or fusion of these maps with other available features of the domain (with the aim of increasing the accuracy of the underlying physical patterns) are good examples of such situation, calling to break-free from the supervised learning paradigm.
Dimensionality Reduction (DR) techniques are a family of mathematical models which work based on matrix factorization. The unsupervised DR techniques seek to provide a new representation of the data within a lower (thus more interpretable) sub-space. After finding this new representative space the original data is being projected onto this new-found subspace in order to 1) reduce the redundancy within data and 2) emphasize the most important factors within it. This will indirectly help clarifying the best (observation) sampling strategies for characterization and emphasis on the most significant detectable pattern within data.  

Spatio-temporal clustering aims at improving the result of clustering by bringing in the spatial information within an image, including coherent regions or textures and fuse them to the information provided across the temporal(or spectral in case of hyperspectral imagery) dimension. One way to reach this goal is to come up with image pyramid of the scene using methods including Gaussian Pyramids and/or Discrete Wavelet Transform and then iteratively clustering the scene beginning from the coarsest to the finest resolution of the pyramid, with the membership probabilities passed on to the next level in each iteration.
Applying the combination of the two mentioned techniques on the stacks of consecutive motion maps (produced by multi-temporal optical/SAR offset tracking) representing the surface behavior of different landslides, a more accurate classification of regions based on their landslide characteristics is expected to be achieved in a complete unsupervised manner. Extensive comparisons can then be made to evaluate the several existing clustering solutions in separation of specific known surface movements. Examples of application of these techniques to SAR derived offset-tracking glacier and landslide displacement fields, and optical terrestrial landslide displacement fields will be presented and discussed.

How to cite: Nakhostin, S., Malet, J., Desrues, M., and Michea, D.: Surface motion information retrieval from dense time series of spaceborne and terrestrial co-registered images, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-17518, https://doi.org/10.5194/egusphere-egu2020-17518, 2020

D2169 |
Viktoria Wichert and Holger Brix

A sensor network surrounds the island of Helgoland, supplying marine data centers with autonomous measurements of variables such as temperature, salinity, chlorophyll and oxygen saturation. The output is a data collection containing information about the complicated conditions around Helgoland, lying at the edge between coastal area and open sea. Spatio-temporal phenomena, such as passing river plumes and pollutant influx through flood events can be found in this data set. Through the data provided by the existing measurement network, these events can be detected and investigated.

 Because of its important role in understanding the transition between coastal and sea conditions, plans are made to augment the sensor network around Helgoland with another underwater sensor station, an Underwater Node (UWN). The new node is supposed to optimally complement the existing sensor network. Therefore, it makes sense to place it in an area that is not yet represented well by other sensors. The exact spatial and temporal extent of the area of representativity around a sensor is hard to determine, but is assumed to have similar statistical conditions as the sensor measures. This is difficult to specify in the complex system around Helgoland and might change with both, space and time.

Using an unsupervised machine learning approach, I determine areas of representativity around Helgoland with the goal of finding an ideal placement for a new sensor node. The areas of representativity are identified by clustering a dataset containing time series of the existing sensor network and complementary model data for a period of several years. The computed areas of representativity are compared to the existing sensor placements to decide where to deploy the additional UWN to achieve a good coverage for further investigations on spatio-temporal phenomena.

A challenge that occurs during the clustering analysis is to determine whether the spatial areas of representativity remain stable enough over time to base the decision of long-term sensor placement on its results. I compare results across different periods of time and investigate how fast areas of representativity change spatially with time and if there are areas that remain stable over the course of several years. This also allows insights on the occurrence and behavior of spatio-temporal events around Helgoland in the long-term.    

Whether spatial areas of representativity remain stable enough temporally to be taken into account for augmenting sensor networks, influences future network design decisions. This way, the extended sensor network can capture a greater variety of the spatio-temporal phenomena around Helgoland, as well as allow an overview on the long-term behavior of the marine system.

How to cite: Wichert, V. and Brix, H.: Augmenting the sensor network around Helgoland using unsupervised machine learning methods, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-15212, https://doi.org/10.5194/egusphere-egu2020-15212, 2020

D2170 |
Arnaud Adam and Isabelle Thomas

Transport geography has always been characterized by a lack of accurate data, leading to surveys often based on samples that are spatially not representative. However, the current deluge of data collected through sensors promises to overpass this scarcity of data. We here consider one example: since April 1st 2016, a GPS tracker is mandatory within each truck circulating in Belgium for kilometre taxes. Every 30 seconds, this tracker collects the position of the truck (as well as some other information such as speed or direction), leading to an individual taxation of trucks. This contribution uses a one-week exhaustive database containing the totality of trucks circulating in Belgium, in order to understand transport fluxes within the country, as well as the spatial effects of the taxation on the circulation of trucks.

Machine learning techniques are applied on over 270 million of GPS points to detect stops of trucks, leading to transform GPS sequences into a complete Origin-Destination matrix. Using machine learning allows to accurately classify stops that are different in nature (leisure stop, (un-)loading areas, or congested roads). Based on this matrix, we firstly propose an overview of the daily traffic, as well as an evaluation of the number of stops made in every Belgian place. Secondly, GPS sequences and stops are combined, leading to characterise sub-trajectories of each truck (first/last miles and transit) by their fiscal debit. This individual characterisation, as well as its variation in space and time, are here discussed: is the individual taxation system always efficient in space and time?

This contribution helps to better understand the circulation of trucks in Belgium, the places where they stopped, as well as the importance of their locations in a fiscal point of view. What are the potential modifications of the trucks routes that would lead to a more sustainable kilometre taxation? This contribution illustrates that combining big-data and machine learning open new roads for accurately measuring and modelling transportation.

How to cite: Adam, A. and Thomas, I.: From machine learning to sustainable taxation: GPS traces of trucks circulating in Belgium , EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-9456, https://doi.org/10.5194/egusphere-egu2020-9456, 2020

D2171 |
Brianna Pagán, Nele Desmet, Piet Seuntjens, Erik Bollen, and Bart Kuijpers

The Internet of Water (IoW) is a large-scale permanent sensor network with 2500 small, energy-efficient wireless water quality sensors spread across Flanders, Belgium. This intelligent water management system will permanently monitor water quality and quantity in real time. Such a dense network of sensors with high temporal resolution (sub-hourly) will provide unprecedented volumes of data for drought, flood and pollution management, prediction and decisions. While traditional physical hydrological models are obvious choices for utilizing such a dataset, computational costs or limitations must be considered when working in real time decision making.

In collaboration with the Flemish Institute for Technological Research (VITO) and the University of Hasselt, we present several data mining and machine learning initiatives which support the IoW. Examples include interpolating grab sample measurements to river stretches to monitor salinity intrusion. A shallow feed forward neural network is trained on historical grab samples using physical characteristics of the river stretches (i.e. soil properties, ocean connectivity). Such a system allows for salinity monitoring without complex convection-diffusion modeling, and for estimating salinity in areas with less monitoring stations. Another highlighted project is the coupling of neural network and data assimilation schemes for water quality forecasting. A long short-term memory recurrent neural network is trained on historical water quality parameters and remotely sensed spatially distributed weather data. Using forecasted weather data, a model estimate of water quality parameters are obtained from the neural network. A Newtonian nudging data assimilation scheme further corrects the forecast leveraging previous day observations, which can aid in the correction for non-point or non-weather driven pollution influences. Calculations are supported by an optimized database system developed by the University of Hasselt which further exploits data mining techniques to estimate water movement and timing through the Flanders river network system. As geospatial data increases exponentially in both temporal and spatial resolutions, scientists and water managers must consider the tradeoff between computational resources and physical model accuracy. These type of hybrid approaches allows for near real-time analysis without computational limitations and will further support research to make communities more climate resilient.

How to cite: Pagán, B., Desmet, N., Seuntjens, P., Bollen, E., and Kuijpers, B.: Data driven methods for real time flood, drought and water quality monitoring: applications for Internet of Water, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-9291, https://doi.org/10.5194/egusphere-egu2020-9291, 2020

D2172 |
Kwo-Sen Kuo and Michael Rilee

The only effective strategy to address the volume challenge of Big Data is “parallel processing”, e.g. employing a cluster of computers (nodes), in which a large volume of data is partitioned and distributed to the cluster nodes. Each of the cluster nodes processes a small portion of the whole volume. The nodes, working in tandem, can therefore collectively process the entire volume within a much-reduced period of time. In the presence of data variety, however, it is no longer as straightforward, because naïve partition and distribution of diverse geo-datasets (packaged with existing practice) inevitably results in misalignment of data for the analysis. Expensive cross-node communication, which is also a form of data movement, thus becomes necessary to bring the data in alignment first before analysis may commence.


Geoscience analysis predominantly requires spatiotemporal alignment of diverse data. For example, we often need to compare observations acquired by different means & platforms and compare model output with observations. Such comparisons are meaningful only if data values for the same space and time are compared. With the existing practice of packaging data using the conventional array data structure, it is nearly impossible to spatiotemporally align diverse data. Because, while array indices are generally used for partition and distribution, for different datasets (even data granules) the same indices most-often-than-not refer to different spatiotemporal neighborhoods. Partition and distribution using conventional array indices thus often results in data of the same spatiotemporal neighborhoods (from different datasets) reside on different nodes. Comparison thus cannot be performed until they are brought together to the same node.


Therefore, we need indices that tie directly and consistently to spatiotemporal neighborhoods to be used for partition and distribution. SpatioTemporal Adaptive-Resolution Encoding (STARE) provides exactly such indices, which can replace floating-point encoding of longitude-latitude and time as a more analytics-optimized alternative. Moreover, data packaging can base on STARE indices. Due to its hierarchical nature, geo-spatiotemporal data packaged based on STARE hierarchy offers essentially a reusable partition for distribution adaptable to various computing-and-storage architectures, through which spatiotemporal alignment of geo-data from diverse sources can be readily and scalably achieved to optimize parallel analytic operations.

How to cite: Kuo, K.-S. and Rilee, M.: Analytics Optimized Geoscience Data Store with STARE-based Packaging, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-20339, https://doi.org/10.5194/egusphere-egu2020-20339, 2020

D2173 |
| Highlight
Giovanni Marchisio and Rasmus Houborg

Planet operates the largest constellation of Earth-observing satellites in human history collecting 1.3 million 29 MP multispectral images over 250 million km2 daily at a resolution of 3-5 meters. This amounts to more than twice the Earth’s total landmass every day and to more than 10 times the area covered by all other commercial and public sources combined, including Sentinel and Landsat, and at a higher resolution. To date we have collected an average of 1,200 images for every point on the surface of the planet. This provides an unparalleled amount of data from which to establish historical baselines and train and refine machine learning algorithms. Intersecting dense time series of global observations with modern deep learning solutions allows us to take a daily pulse of the planet like it has never been done before.

The daily temporal cadence and higher resolution at global scale is unlocking new challenges and opportunities. These range from tracking and discovering previously unknown natural phenomena to improving existing approaches for modeling vegetation phenology and monitoring human impact on the environment. We will provide a brief overview of recent success stories from our university partner ecosystem. For instance, spatio-temporal analytics based on millions of observations has enabled researchers to show that sub-seasonal fluctuations in surface water of Arctic-Boreal can increase carbon emissions and affect global climate to an extent that has eluded traditional satellite remote sensing. The new data source has also enabled intraday measurements of river flows, the first ever measurements of crop water usage and evapotranspiration from space, field level sowing date prediction on a nearly daily basis and improved detection of early-season corn nitrogen stress.

The second part of our presentation covers Planet’s own internal development of spatio-temporal deep learning solutions which target the interaction between geosphere and anthroposphere. Man-made structures such as roads and buildings are among the information layers that we are beginning to extract from our imagery reliably and at a global scale. Our deep learning models, with about seven million parameters, are trained on several billion labeled pixels representative of a wide variety of terrains, densities, land cover types and seasons worldwide. The outcome is a pipeline that has produced the most complete and current map of all the roads and buildings worldwide. It reveals details not available in popular mapping tools, in both industrialized cities and rural settlements. The high temporal cadence of these spatial information feeds increases our confidence in tracking permanent change associated with urbanization and improves our knowledge of how human settlements grow. Applications include tracking urban sprawl at a country level in China, deriving land consumption rates for countries in Sub-Saharan Africa, identifying construction in flood risk zones worldwide, and timely augmentation of OpenStreetMap in disaster management situations that affect developing countries. With continually refreshed imagery from space, such maps can be updated to highlight new changes around the world, opening up new possibilities to improve transparency and help life on Earth.

How to cite: Marchisio, G. and Houborg, R.: Modeling and Capturing New Phenomena from Very High Cadence Earth Observations, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-12492, https://doi.org/10.5194/egusphere-egu2020-12492, 2020

D2174 |
Min Zhang

With the rapid development of urbanization, many problems become more serious in big cities, such as traffic congestion. Different urban land use type can have different influence on traffic, therefore, the analysis of relationship between urban traffic and urban land use is important for better understanding of urban traffic status. This study firstly utilizes spatial data analysis method and time series analysis method to obtain urban traffic pattern from the spatial and temporal perspective, using one-week traffic sensor data, we measure the urban commuting patterns, which include weekday mode and weekend mode. Secondly, this study analyzes the relationship between traffic status and land use type in traffic analysis zone (TAZ) level, which indicates traffic status has spatial autocorrelation, besides, commercial land use and mixed land use type may result in more serious traffic congestion. The research can be of value for urban understanding and decision making in areas of urban management, urban plan and traffic control.

How to cite: Zhang, M.: Understanding traffic distribution pattern from the perspective of urban land use, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-4099, https://doi.org/10.5194/egusphere-egu2020-4099, 2020

D2175 |
Guolei Zhou

Abstract: Urban shrinkage has become a global phenomenon, occurring all over the world. Faced with the reduction of population, shrinking cities will continue to lose the vitality of development, and it is difficult to achieve the past glory. How to realize the sustainable development of shrinking cities will become an important issue, which deserves our in-depth study. We will apply big data to analyze the spatiotemporal changes of the population of shrinking cities. Urban shrinkage will lead to a series of chain reactions, reflected in all aspects of social and economic development. Shops are closing down constantly, and there are very few visitors to the commercial streets at night. Industrial enterprises have moved to other places, and urban employment and financial income have declined sharply. The city's economy will face the risk of collapse. Consequently, the lack of maintenance of infrastructure and public service facilities, the decrease of residents' income and the decrease of residents' happiness lead to the lack of cohesion in urban society. The sustainable development of shrinking cities will face great difficulties. Therefore, we must be aware of the seriousness of this problem and take necessary actions to reduce the negative effects of urban shrinkage.

Key words: Shrinking cities, sustainable development, society, economy

How to cite: Zhou, G.: Negative Effects of Shrinking Cities and the Dilemma of Their Sustainable Development, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-13040, https://doi.org/10.5194/egusphere-egu2020-13040, 2020

D2176 |
| Highlight
Laura Bindereif, Tobias Rentschler, Martin Bartelheim, Marta Díaz-Zorita Bonilla, Philipp Gries, Thomas Scholten, and Karsten Schmidt

Land cover information plays an essential role for resource development, environmental monitoring and protection. Amongst other natural resources, soils and soil properties are strongly affected by land cover and land cover change, which can lead to soil degradation. Remote sensing techniques are very suitable for spatio-temporal mapping of land cover mapping and change detection. With remote sensing programs vast data archives were established. Machine learning applications provide appropriate algorithms to analyse such amounts of data efficiently and with accurate results. However, machine learning methods require specific sampling techniques and are usually made for balanced datasets with an even training sample frequency. Though, most real-world datasets are imbalanced and methods to reduce the imbalance of datasets with synthetic sampling are required. Synthetic sampling methods increase the number of samples in the minority class and/or decrease the number in the majority class to achieve higher model accuracy. The Synthetic Minority Over-Sampling Technique (SMOTE) is a method to generate synthetic samples and balance the dataset used in many machine learning applications. In the middle Guadalquivir basin, Andalusia, Spain, we used random forests with Landsat images from 1984 to 2018 as covariates to map the land cover change with the Google Earth Engine. The sampling design was based on stratified random sampling according to the CORINE land cover classification of 2012. The land cover classes in our study were arable land, permanent crops (plantations), pastures/grassland, forest and shrub. Artificial surfaces and water bodies were excluded from modelling. However, the number of the 130 training samples was imbalanced. The classes pasture (7 samples) and shrub (13 samples) show a lower number than the other classes (48, 47 and 16 samples). This led to misclassifications and negatively affected the classification accuracy. Therefore, we applied SMOTE to increase the number of samples and the classification accuracy of the model. Preliminary results are promising and show an increase of the classification accuracy, especially the accuracy of the previously underrepresented classes pasture and shrub. This corresponds to the results of studies with other objectives which also see the use of synthetic sampling methods as an improvement for the performance of classification frameworks.

How to cite: Bindereif, L., Rentschler, T., Bartelheim, M., Díaz-Zorita Bonilla, M., Gries, P., Scholten, T., and Schmidt, K.: Synthetic sampling for spatio-temporal land cover mapping with machine learning and the Google Earth Engine in Andalusia, Spain, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-1153, https://doi.org/10.5194/egusphere-egu2020-1153, 2019

D2177 |
Lucia Saganeiti, Ahmed Mustafà, Jacques Teller, and Beniamino Murgante

This paper presents a spatiotemporal analysis to simulate and predict urban growth. It is important to study the growth of cities to understand their implications for environmental sustainability and infrastructure needs. The aim of this work is to predict future scenarios in low-density settlements in order to control and regulate the processes of urban transformation.

We applied an integrated approach based on the multinomial logistic regression (MLR) and the cellular automata (CA) for urban sprinkling modelling. Our case study is the Basilicata region, in southern Italy, which is affected by urban sprinkling literally defined as: “a small amount of urban territory distributed in scattered particles".

Built-up density maps were created on the basis of three regional building datasets (1989, 1998 and 2013) with four density classes: no built up, low density, medium density and high density. These sources were used for calibrating and validating the model as well as for future simulation of urban sprinkling. Two components were considered for the calculation of the transition potential from one density class to another. For the first built up development causative factors were calibrated using the MLR for the expansion and densification processes. The causative factors consider elevation and slope as physical factors, Euclidian distance to railway station, different type of street, large size city and medium size city as proximity factors, population density and employment rate as socio economic factors and zoning for land use policies in the study area.

The second causative factor is the CA neighbourhood effects that have been calibrated using a multi objective genetic algorithm (MOGA) as in (Mustafa et al., 2018). The transition potential was calibrated for the 1989-1998 time period and the calibration outcomes were used to simulate the 2013 map. The calibrated map was then used for the simulation of the 2013 map which was compared with the actual map of 2013 (validation). We then used our calibrated model to simulate urban growth in the year 2030.

The robustness of MLR has been evaluated with the Receiver Operating Characteristic (ROC) index. The Fuzzines index has been used for evaluating the validation process accuracy.

The results of the 2030 prediction show the greatest variations in class 1 (low density) that represent the sprinkling of urban cells in the territory.

Keywords: Low-density, Cellular Automata, Multinomial Logistic Regression, Urban Sprinkling, Basilicata Region.

How to cite: Saganeiti, L., Mustafà, A., Teller, J., and Murgante, B.: Predict urban growth in a low-density context: Basilicata region study case, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-16187, https://doi.org/10.5194/egusphere-egu2020-16187, 2020

D2178 |
Dan Li, Jigang Qiao, and Yihan Zhang

Territory spatial planning is a guide and blueprint for future territorial development in China. It means to form a scientific, rational, intensive, and efficient spatial protection and development pattern in territory space. The first task according to the government is to delimitate the functional zones of ecology, agriculture, urban zones, and delineation of ecological protection red lines, permanent basic farmland boundaries, and urban development boundaries ("three zones and three lines"). Currently China used a resource and environment carrying capacity and land space development suitability evaluation ("double evaluations") to complete the delimitation task. However, the process of these evaluations and demarcation is relatively complicated, high-level human intervention, and the operability is not strong, therefore it is not practically at municipal or county level. We proposed a new delineation framework, methods, and software tools for the delimitation work, based on a spatial optimization and simulation coupling approach, and is verified by an example in Guangzhou, a super metropolis city in China. It shows that this method can rapidly and efficiently delimit urban ecological and agricultural zones based on regional geographic background conditions, by using an ant colony intelligent optimization algorithm, and using a cellular automata model to delineate compact urban zones. Compared with the "three zones" division plan in the "Guangzhou Land and Space Master Plan (2018-2035) Draft" which is published by local government, the delimitated functional zones by proposed method can meet the quantitative requirements of the draft, while providing more detailed and realistic spatial pattern of the three functional zones, which can be very useful for municipal and county level territory spatial planning work.

How to cite: Li, D., Qiao, J., and Zhang, Y.: Delimitating functional zones at municipal or county level in China based on a spatial optimization and simulation coupling approach, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7181, https://doi.org/10.5194/egusphere-egu2020-7181, 2020

D2179 |
Angela Pilogallo, Lucia Saganeiti, Francesco Scorza, and Beniamino Murgante

By the end of this century effects of land-use change on ecosystem services are expected to be more significant than other world-wide transformation processes such as climate change, altering atmospheric concentrations of greenhouse gases or distribution of invasive alien species.

In recent years, scientific literature has been embellished with numerous land-use models that aim to explore the behaviour of land use systems under changing environmental conditions and different territorial transformations explaining the different dynamics that contribute to it, and to formulate scenario analyses to be followed up by development strategies.

In addition, it should be noted that a dimension of the nexus between planning and sustainability that is important but still too little explored, is the assessment of territorial changes and development dynamics through the alterations analysis induced on processes, functions and complex systems.

While land-use models can help investigate the effects of a combination of drivers at different scales, ecosystem services approach can help in better understand the trade-offs between different development scenarios making explicit the relations that every variation induces within the relationship between man and territory  and among different environmental components.

In this framework is set the present work that aims to integrate scenario analysis of the Basilicata region (Italy) development with assessments of alterations induced on the capacity to deliver ecosystem services. Although this region is very poorly populated and characterised by low settlement density, it is not immune to the global phenomenon of land take associated with high territorial fragmentation.

The building stock increase due to real development dynamics and relative demographic increase typical of the post-war period, was followed by a further built up environment growth - in contrast with the demographic trend - and a significant land take due to massive construction of renewable energy production plants.

Changing model have been applied to identify and classify the driving forces for land use changes and predict future development scenarios.

In order to contribute to the development of decision support systems, scenarios resulting from the implementation of different policies are analyzed with the ecosystem services approach.

How to cite: Pilogallo, A., Saganeiti, L., Scorza, F., and Murgante, B.: Investigating the effects of land use change on ecosystem services: the Basilicata region (Italy) case study, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-18512, https://doi.org/10.5194/egusphere-egu2020-18512, 2020

D2180 |
Cunjin Xue and Changfeng Jing

A marine heatwave (MHW) is defined as a coherent area of extreme warm sea surface temperature that persists for days to months, which has a property of evolution from production through development to death in space and time. MHWs usually relates to climatic extremes that can have devastating and long-term impacts on ecosystems, with subsequent socioeconomic consequences. Long term remote sensing products make it possible for mining successive MHWs at global scale. However, more literatures focus on a spatial distribution at a fixed time snapshot or a temporal statistic at a fixed grid cell of MWHs. As few considering the temporal evolution of MWHs, it is greater challenge to mining their dynamic changes of spatial structure. Thus, this manuscript proposes a process-oriented approach for identifying and tracking MWHs, named as PoAITM. The PoAITM considers a dynamic evolution of a MWH, which consists of three steps. The first step uses the threshold-based algorithm to identifying the time series of grid pixels which meets the MWH definition, called as MWH pixels; the second adopts the spatial proximities to connect the MWH pixels at the snapshots, and transforms them spatial objects, called as MWH objects; the third combines the dynamic characteristics and spatiotemporal topologies of MWH objects between the previous and next snapshots to identify and track them belonging to the same ones. The final extract MWH with a property from production through development to death is defined as a MWH process. Comparison with the prevail methods of tracking MHWs, The PoAITM has three advantages. Firstly, PoAITM combines the spatial distribution and temporal evolution of MWH to identify and track the MWH objects. The second considers not only the spatial structure of MWH at current snapshot, also the previous and next ones, to track the MWH process, which ensures the MWH completeness in a temporal domain. The third is the dynamic behaviors of MWH, e.g. developing, merging, splitting, are also found between the successive MWH objects. Finally, we address the global MWHs exploring from the sea surface temperature products during the period of January 1982 to December 2018. The results not only show well-known knowledge, but also some new findings about evolution characteristics of MWHs, which may provide new references for further study on global climate change.

How to cite: Xue, C. and Jing, C.: A process-oriented approach for mining marine heatwaves with a time series of raster formatted products, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-2251, https://doi.org/10.5194/egusphere-egu2020-2251, 2020

D2181 |
Li-hong Shi

Abstract: Agglomeration of the manufacturing industries is not only a fundamental driving force for urban development However, a large number of manufacturing industries produce sewage and thus have negative effects on regional environment. This study first estimates the degree of clustering of pollution intensive manufacturing industries in the developed region of China at city level by introducing the Kernel density distribution function, and then evaluates the pollution distribution pattern by dividing the study area into several environmental units according to the naturally integrated characteristics of the primary streams. Furthermore, we quantitatively analyze the mechanism of the response of water environment quality to industrial distribution by utilizing the bi-variate spatial autocorrelation model. Results show that pollution-intensive manufacturing industries form clusters in suburban and non-sensitive areas. Besides, the density of pollution sources gradually decreases from the chief canals to the peripheral areas. Spatial autocorrelation analysis shows that spatial-relationships show differences according to industry categories: the agglomeration of textile, petrochemical and metallurgical industries prominently affects the spatial heterogeneity of water pollution distribution while the effects of the agglomeration of food manufacturing and paper-making industry on water pollution are not significant. Based on the spatial autocorrelation between industrial agglomeration and pollution distribution, we divide the study area into four types: high-agglomeration and high-pollution area, low-agglomeration and low-pollution area, low-agglomeration and high-pollution area, high-agglomeration and low-pollution area. Based on the that, we analyze the formation scheme and provide policy suggestions regarding industrial development. This paper provides a new perspective for the study of the interaction between industrial agglomeration and environment effects, and will be benefit the sustainable development of cities.

Key words: industrial agglomeration; Kernel Density Distribution function; water pollution; manufacturing industry; spatial autocorrelation

How to cite: Shi, L.: Study on spacial-correlation between water pollution and industrial agglomeration in the developed regions of China, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6382, https://doi.org/10.5194/egusphere-egu2020-6382, 2020

D2182 |
Artur Kohler

Groundwater contamination resulted from chemical releases related to anthropogenic activity often proves to be a persistent feature of the affected groundwater regime.  The affected volume (i.e. where the concentration of hazardous substances exceeds a certain threshold) is a complex and dynamic entity commonly called “contaminant plume”.  The plume can be described as a spatially dependent concentration pattern with temporal behavior.  Persistent plumes are regularly monitored, concentration data gained by repeated sampling of monitoring points and laboratory analyses of the samples are used to assess the actual state of the plume.  The change of the concentrations at certain points of the plume facilitates the assessment of the temporal behavior of the plume.  Repeated sampling of the monitoring points provides concentration time series.

Concentration time series are evaluated for trends.  Methods include parametric (regression using least squares) and non-parametric methods.  Mann-Kendall statistic is a commonly used, well known non parametric method.

When using Mann-Kendall statistics consecutive concentration data are compared to each other, their cumulative relation defines Mann-Kendall statistic ‘S’.  However, when comparing concentration data laboratory uncertainties are usually neglected.  Allowing for laboratory uncertainties, rises the question of what concentrations are considered equal, less or more than other concentrations.  In addition aggravating concentration data will change the previous equal – more - less status of two concentrations, thus changing the Mann-Kendall statistics value, which sometimes results in differences in trend significance.

How to cite: Kohler, A.: Application of nonparametric trend analysis to concentration time series, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-13476, https://doi.org/10.5194/egusphere-egu2020-13476, 2020

D2183 |
Amirhossein Hassani, Adisa Azapagic, and Nima Shokri

Soil salinity is among the major threats affecting the soil fertility, stability and vegetation. It can also accelerate the desertification processes, especially in arid and semi-arid regions. An accurate estimation of the global extent and distribution of the salt-affected soils and their temporal variations is pivotal to our understanding of the salinity-induced land degradation processes and to design effective remediation strategies. In this study, using legacy soil profiles data and a broad set of climatic, topographic, and remotely sensed soil surface and vegetative data, we trained ensembles of classification and regression trees to map the spatio-temporal variation of the soil salinity and sodicity (exchangeable sodium percentage) at the global scale from 1980 to 2018 at a 1 km resolution. The User’s Accuracies for soil salinity and sodicity classification were 88.05% and 84.65%, respectively. The 2018 map shows that globally  ̴ 944 Mha of the lands are saline (with saturated paste electrical conductivity > 4 ds m-1), while  ̴ 155 Mha can be classified as sodic soils (with sodium exchange percentage > 15%). Our findings and provided dataset show quantitatively how soil salinization is influenced by a broad array of climatic, anthropogenic and hydrologic parameters. Such information is crucial for effective water and land-use management, which is important for maintaining food security in face of future climatic uncertainties. Moreover, our results combined with the quantitative methodology developed in this study will provide us with an opportunity to delineate the role of anthropogenic activities on soil salinization. This information is useful not only for developing predictive models of primary and secondary soil salinization but also for natural resources management and policy makers.

How to cite: Hassani, A., Azapagic, A., and Shokri, N.: Spatio-temporal variability of global soil salinization delineated by advanced machine learning algorithms, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-15891, https://doi.org/10.5194/egusphere-egu2020-15891, 2020

D2184 |
Sibo Zhang and Wei Yao

In the past, soil moisture can be retrieved from microwave imager over most of land conditions. However, the algorithm performances over Tibetan Plateau and the Northwest China vary greatly from one to another due to frozen soils and surface volumetric scattering. The majority of western Chinese region is often filled with invalid retrievals. In this study, Soil Moisture Operational Products System (SMOPS) products from NOAA are used as the learning objectives to train a  machine learning (random forest) model for FY-3C microwave radiation imager (MWRI) data with multivariable inputs: brightness temperatures from all 10 MWRI channels from 10 to 89 GHz, brightness temperature polarization ratios at 10.65, 18.7 and 23.8 GHz, height in DEM (digital elevation model) and statistical soil porosity map data. Since the vegetation penetration of MWRI observations is limited, we exclude forest, urban and snow/ice surfaces in this work. It is shown that our new method performs very well and derives the surface soil moisture over Tibetan Plateau without major missing values. Comparing to other soil moisture data, the volumetric soil moisture (VSM) from this study correlates with SMOPS products much better than the MWRI operational L2 VSM products. R2 score increases from 0.3 to 0.6 and ubRMSE score improves significantly from 0.11 m3 m-3 to 0.04 m3 m-3 during the time period from 1 August 2017 to 31 May 2019. The spatial distribution of our MWRI VSM estimates is also much improved in western China. Moreover, our MWRI VSM estimates are in good agreement with the top 7 cm soil moisture of ECMWF ERA5 reanalysis data: R2 = 0.62, ubRMSD = 0.114 m3 m-3 and mean bias = -0.014 m3 m-3 for a global scale. We note that there is a risk of data gap of AMSR2 from the present to 2025. Obviously, for satellite low frequency microwave observations, MWRI observations from FY-3 series satellites can be a benefit supplement to keep the data integrity and increase the data density, since FY-3B\-3C\-3D satellites launched in November 2010\September 2013\November 2017 are still working today, and FY-3D is designed to work until November 2022.

How to cite: Zhang, S. and Yao, W.: Remote Sensing of Surface Soil Moisture from FengYun MicroWave Radiation Imager (MWRI) Data Using a Machine Learning Technique , EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-311, https://doi.org/10.5194/egusphere-egu2020-311, 2019

D2185 |
Lei Wang, Haoran Sun, Wenjun Li, and Liang Zhou

Crop planting structure is of great significance to the quantitative management of agricultural water and the accurate estimation of crop yield. With the increasing spatial and temporal resolution of remote sensing optical and SAR(Synthetic Aperture Radar) images,  efficient crop mapping in large area becomes possible and the accuracy is improved. In this study, Qingyijiang Irrigation District in southwest of China is selected for crop identification methods comparison, which has heterogeneous terrain and complex crop structure . Multi-temporal optical (Sentinel-2) and SAR (Sentinel-1) data were used to calculate NDVI and backscattering coefficient as the main classification indexes. The multi-spectral and SAR data showed significant change in different stages of the whole crop growth period and varied with different crop types. Spatial distribution and texture analysis was also made. Classification using different combinations of indexes were performed using neural network, support vector machine and random forest method. The results showed that, the use of multi-temporal optical data and SAR data in the key growing periods of main crops can both provide satisfactory classification accuracy. The overall classification accuracy was greater than 82% and Kappa coefficient was greater than 0.8. SAR data has high accuracy and much potential in rice identification. However optical data had more accuracy in upland crops classification. In addition, the classification accuracy can be effectively improved by combination of classification indexes from optical and SAR data, the overall accuracy was up to 91.47%. The random forest method was superior to the other two methods in terms of the overall accuracy and the kappa coefficient.

How to cite: Wang, L., Sun, H., Li, W., and Zhou, L.: crops planting area identification and analysis based on multi-source high resolution remote sensing data, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-13227, https://doi.org/10.5194/egusphere-egu2020-13227, 2020

D2186 |
Yannik Roell, Amélie Beucher, Per Møller, Mette Greve, and Mogens Greve

Predicting wheat yield is crucial due to the importance of wheat across the world. When modeling yield, the difference between potential and actual yield consistently changes because of technology. Considering historical yield potential would help determine spatiotemporal trends in agricultural development. Comparing current and historical production in Denmark is possible because production has been documented throughout history. However, the current winter wheat yield model is solely based on soil. The aim of this study was to generate a new Danish winter wheat yield map and compare the results to historical production potential. Utilizing random forest with soil, climate, and topography variables, a winter wheat yield map was generated from 876 field trials carried out from 1992 to 2018. The random forest model performed better than the model based only on soil. The updated national yield map was then compared to production potential maps from 1688 and 1844. While historical time periods are characterized by numerous low production potential areas and few highly productive areas, present-day production is evenly distributed between low and high production. Advances in technology and farm practices have exceeded historical yield predictions. Thus, modeling current yield could be unreliable in future years as technology progresses.

How to cite: Roell, Y., Beucher, A., Møller, P., Greve, M., and Greve, M.: Comparing a random forest based prediction of winter wheat yield to historical production potential, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-138, https://doi.org/10.5194/egusphere-egu2020-138, 2019

D2187 |
Sheng Sheng, Hua Chen, Chong-Yu Xu, Wen Zhang, Zhishuai Li, and Shenglian Guo

Reliable estimation of grid precipitation dataset from gauge observations is crucial for hydrological modelling and water balance studies. Datasets developed by common precipitation interpolation methods are mainly derived from the spatial relationship of gauges while neglecting the trend contained in the antecedent precipitation. Precipitation data can be viewed as an intrinsically related matrix, with columns representing temporal relationships and rows representing spatial relationships. A method, called combinatorial point spatiotemporal interpolation based on singular value decomposition (CPST-SVD) that combines traditional interpolators and matrix factorization and considers the spatiotemporal correlation of precipitation, is proposed to improve estimation. Two widely used approaches including the inverse distance weighting (IDW) and universal kriging (UK) were combined to the new method respectively to interpolate precipitation data. Hourly precipitation data from several flood events were selected to verify the performance of the new method in the time period between 2012 and 2018 under different meteorological conditions in Hanjiang Basin, China. The Funk SVD algorithm and the stochastic gradient descent (SGD) algorithm were introduced for matrix factorization and optimization. The performances of all methods in the leave-one-out cross-validation were assessed and compared by five statistical indicators. The results show that CPST-SVD combined with IDW has the highest accuracy, followed by CPST-SVD combined with UK, IDW and UK in descending order. Through combination, estimation errors in precipitation interpolation can be greatly reduced, especially for the situation that the distribution of surrounding gauges is not so uniform or the precipitation in the target gauge is non-zero. In addition, the larger the amount of precipitation event, the greater the improvement of error. Therefore, this study provides a more accurate method for interpolating precipitation based on the assessment of both temporal and spatial information.

How to cite: Sheng, S., Chen, H., Xu, C.-Y., Zhang, W., Li, Z., and Guo, S.: A combinatorial method for improving hourly precipitation interpolation based on singular value decomposition, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6276, https://doi.org/10.5194/egusphere-egu2020-6276, 2020

D2188 |
İsmail Sezen, Alper Unal, and Ali Deniz

Atmospheric pollution is one of the primary problems and high concentration levels are critical for human health and environment. This requires to study causes of unusual high concentration levels which do not conform to the expected behavior of the pollutant but it is not always easy to decide which levels are unusual, especially, when data is big and has complex structure. A visual inspection is subjective in most cases and a proper anomaly detection method should be used. Anomaly detection has been widely used in diverse research areas, but most of them have been developed for certain application domains. It also might not be always a good idea to identify anomalies by using data from near measurement sites because of spatio-temporal complexity of the pollutant. That’s why, it’s required to use a method which estimates anomalies from univariate time series data.

This work suggests a framework based on STL decomposition and extended isolation forest (EIF), which is a machine learning algorithm, to identify anomalies for univariate time series which has trend, multi-seasonality and seasonal variation. Main advantage of EIF method is that it defines anomalies by a score value.

In this study, a multi-seasonal STL decomposition has been applied on a univariate PM10 time series to remove trend and seasonal parts but STL is not resourceful to remove seasonal variation from the data. The remainder part still has 24 hours and yearly variation. To remove the variation, hourly and annual inter-quartile ranges (IQR) are calculated and data is standardized by dividing each value to corresponding IQR value. This process ensures removing seasonality in variation and the resulting data is processed by EIF to decide which values are anomaly by an objective criterion.

How to cite: Sezen, İ., Unal, A., and Deniz, A.: Anomaly Detection by STL Decomposition and Extended Isolation Forest on Environmental Univariate Time Series, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-18471, https://doi.org/10.5194/egusphere-egu2020-18471, 2020

D2189 |
Xiao Feng

Air pollution poses a serious threat to human health. A large number of studies have shown that certain diseases are closely related to air pollution. Understanding the spatiotemporal distribution of air pollutants and their health effects are of great significance for pollution prevention. This study takes Hubei Province, China as an example. It integrates measured ground air quality data, natural environment data, and socioeconomic data, and uses machine learning to improve the land use regression model to simulate the spatial distribution of concentration PM2.5 / O3 from 2015 to 2018 in the study area. The combined pollutant concentration data and population raster data were used to assess the deaths from specific diseases (stroke, ischemic heart disease, lung cancer) caused by air pollutants. The result shows that high concentrations of pollutants are concentrated in the more economically developed eastern regions of Hubei Province, and the economically backward western regions have good air quality. In addition, the distribution of deaths caused by exposure to air pollution is similar to that of pollutants, which is higher in eastern part of Hubei province. However, the total number of deaths in the province is decreasing year by year. This result shows that environmental governance policies have alleviated the threat of air pollution to human health to some extent. It shows that Hubei Province should combine actual conditions and spatial-temporal distribution characteristics of pollutants to make appropriate environmental protection measures.

How to cite: Feng, X.: Spatiotemporal distribution of major pollutants and their health impacts in Hubei Province from 2015 to 2018 based on machine learning to improve LUR, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-3822, https://doi.org/10.5194/egusphere-egu2020-3822, 2020

D2190 |
Wei Yao, Octavian Dumitru, Jose Lorenzo, and Mihai Datcu

This abstract describes the Data Fusion tool of the Horizon 2020 CANDELA project. Here, Sentinel-1 (synthetic aperture radar) and Sentinel-2 (multispectral) satellite images are fused at feature level. This fusion is made by extracting the features from each type of image; then these features are combined in a new block within the Data Model Generation sub-module of the Data Fusion system.

The corresponding tool has already been integrated with the CANDELA cloud platform: its Data Model component on the platform is acting as backend, and the user interaction component on the local user machine as frontend. There are four main sub-modules: Data Model Generation for Data Fusion (DMG-DF), DataBase Management System (DBMS), Image Search and Semantic Annotation (ISSA), and multi-knowledge and Query (QE). The DMG-DF and DBMS sub-modules have been dockerized and deployed on the CANDELA platform. The ISSA and QE sub-modules require user inputs for their interactive interfaces. They can be started as a standard Graphical User Interface (GUI) tool which is linked directly to the database on the platform.

Before using the Data Fusion tool, users have to prepare the already co-registered Sentinel-1 and Sentinel-2 products as inputs. The S1tiling service provided on the platform is able to cut out the overlapping Sentinel-1 area based on Sentinel-2 tile IDs.

The pipeline of the Data Fusion tool starts from the DMG-DF process on the platform, and the data will be transferred via Internet; then local end users can perform semantic annotations. The annotations will be ingested into the database on the platform via Internet.

The Data Fusion process consists of three steps:

  • On the platform, launch a Jupyter notebook for Python, and start the Data Model Generation for Data Fusion to process the prepared Sentinel-1 and Sentinel-2 products which cover the same area;
  • On the local user machine, by clicking the Query button of the GUI, users can get access to the remote database, make image search and queries, and perform semantic annotations by loading quick-look images of processed Sentinel-1 and Sentinel-2 products via Internet. Feature fusion and image quick-look pairing are performed at runtime. The fused features and paired quick-looks help obtain better semantic annotations. When clicking on another ingestion button, the annotations are ingested into the database on the platform;
  • On the platform, launch a Jupyter notebook for Python, and the annotations and the processed product metadata can be searched and queried.

Our preliminary validation results are made based on visual analysis, by comparing the obtained classification maps with already available CORINE land cover maps. In general, our fused results generate more complete classification maps which contain more classes.

How to cite: Yao, W., Dumitru, O., Lorenzo, J., and Datcu, M.: Data Fusion on the CANDELA Cloud Platform, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19933, https://doi.org/10.5194/egusphere-egu2020-19933, 2020

D2191 |
Alexandra Moshou, Antonios Konstantaras, Emmanouil Markoulakis, Panagiotis Argyrakis, and Emmanouil Maravelakis

The identification of distinct seismic regions and the extraction of features of theirs in relation to known underground fault mappings could provide most valuable information towards understanding the seismic clustering phenomenon, i.e. whether an earthquake occurring in a particular area can trigger another earthquake in the vicinity. This research paper works towards that direction and unveils the potential presence and extent of distinct seismic regions in the area of the Southern Hellenic Seismic Arc. To achieve that, a spatio-temporal clustering algorithm has been developed based on expert knowledge regarding the spatial and timely influence of an earthquake  in its nearby vicinity using seismic data provided by the Geodynamics Institute of Athens, and is further supported by geological observations of underground faults’ mappings beneath the addressed potentially distinct seismic regions. This is made possible thanks to advances in deep learning and graphics processing units’ 3D technology that encompass parallel processing architectures, which comprise of blocks of multiple cores with parallel threads providing the necessary foundation in terms of hardware for accelerated processing for parallel seismic big data analysis. Seismic data are normally stored in massive continuously expanding matrices, as wide areas seismic covering is thickening, due to the establishment of denser recording networks, and decades of data are being stacked together. This research work embodies that technology for the development and implementation of a Cuda parallel processing agglomerative spatio-temporal clustering algorithm that enables the import of expert knowledge for the investigation of the potential presence of distinct seismic regions in the vicinity under investigation. The overall spatio temporal clustering results are also in accordance with empirical observations reported in the literature throughout the vicinity of the Hellenic Seismic Arc.

Indexing terms: parallel processing, heterogeneous parallel programming, Cuda, distinct seismic regions, seismic clustering, spatio-temporal clustering


Axaridou A., I. Chrysakis, C. Georgis, M. Theodoridou, M. Doerr, A. Konstantaras, and E. Maravelakis. 3D-SYSTEK: Recording and exploiting the production workflow of 3D-models in cultural heritage. IISA 2014 - 5th International Conference on Information, Intelligence, Systems and Applications, 51-56, 2014.

Drakatos G. and J. Latoussakis. A catalog of aftershock sequences in Greece (1971–1997): Their spatial and temporal characteristics. Journal of Seismology. 5, 137–145, 2001.

Konstantaras A.J. Classification of distinct seismic regions and regional temporal modelling of seismicity in the vicinity of the Hellenic seismic arc. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 6 (4), 1857-1863, 2012.

Konstantaras A.J., E. Katsifarakis, E. Maravelakis, E. Skounakis, E. Kokkinos and E. Karapidakis. Intelligent spatial-clustering of seismicity in the vicinity of the Hellenic Seismic Arc. Earth Science Research 1 (2), 1-10, 2012.

Maravelakis E., A. Konstantaras, K. Kabassi, I. Chrysakis, C. Georgis and A. Axaridou. 3DSYSTEK web-based point cloud viewer. IISA 2014 - 5th International Conference on Information, Intelligence, Systems and Applications, 262-266, 2014.

Moshou Alexandra, Eleftheria Papadimitriou, George Drakatos, Christos Evangelidis Vasilios Karakostas, Filippos Vallianatos, and Konstantinos Makropoulos Focal Mechanisms at the convergent plate boundary in Southern Aegean, Greece, Geophysical Research Abstracts, Vol. 16, EGU2014-12185, 2014, EGU General Assembly 2014

How to cite: Moshou, A., Konstantaras, A., Markoulakis, E., Argyrakis, P., and Maravelakis, E.: A Deep-Learning Parallel Processing Agglomerative Algorithm for the Identification of Distinct Seismic Regions in the Southern Hellenic Seismic Arc, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-10768, https://doi.org/10.5194/egusphere-egu2020-10768, 2020

D2192 |
Yongshang Wang

The new geodetic technology system characterized by high precision, real-time and popularization has gradually formed in the wave of new technological revolution. Taking advantage of modern information technology,the new geodetic technology system has new features for technology sharing and data sharing, in which geodetic data and positioning applications and geodetic applications can meet the requirements of real-time, public-sharedand interactivity. Geodetic Data Standards are the cornerstone in process of geodetic informationization, which describes scientifically the geodetic data processing, management and service, as well as the theory, methods and procedures used for implementation. Along with the development of geodetic informationization, the width and depth of data standardization application will far exceed the level of traditional standardization. With the advancement of measurement technology, there are less technical constraints in the relatively simple instruments operation process while more complicated data structure to be analyzed, and the more urgent socialization services requirements to be met. The focus of geodetic standardization will shift from operational standards to data standards. In this paper, the content, characteristics, classification principles and methods of geodetic data are studied for the new characteristics of modern geodetic informationization.

How to cite: Wang, Y.: Analysis and Construction of Geodetic Data Classification Standard , EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-6771, https://doi.org/10.5194/egusphere-egu2020-6771, 2020

D2193 |
Claudie Beaulieu, Matthew Hammond, Stephanie Henson, and Sujit Sahu

Assessing ongoing changes in marine primary productivity is essential to determine the impacts of climate change on marine ecosystems and fisheries. Satellite ocean color sensors provide detailed coverage of ocean chlorophyll in space and time, now with a combined record length of just over 20 years. Detecting climate change impacts is hindered by the shortness of the record and the long timescale of memory within the ocean such that even the sign of change in ocean chlorophyll is still inconclusive from time-series analysis of satellite data. Here we use a Bayesian hierarchical space-time model to estimate long-term trends in ocean chlorophyll. The main advantage of this approach comes from the principle of ”borrowing strength” from neighboring grid cells in a given region to improve overall detection. We use coupled model simulations from the CMIP5 experiment to form priors to provide a “first guess” on observational trend estimates and their uncertainty that we then update using satellite observations. We compare the results with estimates obtained with the commonly used vague prior, reflecting the case where no independent knowledge is available.  A global average net positive chlorophyll trend is found, with stronger regional trends that are typically positive in high and mid latitudes, and negative at low latitudes outside the Atlantic. The Bayesian hierarchical model used here provides a framework for integrating different sources of data for detecting trends and estimating their uncertainty in studies of global change.

How to cite: Beaulieu, C., Hammond, M., Henson, S., and Sahu, S.: Long-term trends in ocean chlorophyll: update from a Bayesian hierarchical space-time model, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-21208, https://doi.org/10.5194/egusphere-egu2020-21208, 2020

D2194 |
| Highlight
Fabian Guignard, Federico Amato, Sylvain Robert, and Mikhail Kanevski

Spatio-temporal modelling of wind speed is an important issue in applied research, such as renewable energy and risk assessment. Due to its turbulent nature and its very high variability, wind speed interpolation is a challenging task. Being universal modeling tools, Machine Learning (ML) algorithms are well suited to detect and model non-linear environmental phenomena such as wind.

The present research proposes a novel and general methodology for spatio-temporal interpolation with an application to hourly wind speed in Switzerland. The methodology is organized as follows. First, the dataset is decomposed through Empirical Orthogonal Functions (EOFs) in temporal basis and spatially dependent coefficients. EOFs constitute an orthogonal basis of the spatio-temporal signal from which the original wind field can be reconstructed. Subsequently, in order to be able to reconstruct the signal at spatial locations where measurements are unknown, the spatial coefficients resulted from the decomposition are interpolated. To this aim, several ML algorithms were used and compared, including k-Nearest Neighbors, Random Forest, Support Vector Machine, General Regression Neural Networks and Extreme Learning Machine. Finally, wind field is reconstructed with the help of the interpolated coefficients.

A case study on real data is presented. Data consists of two years of wind speed measurements at hourly frequency collected by Meteoswiss at several hundreds of stations in Switzerland, which has a complex orography. After cleaning and handling of missing values, a careful exploratory data analysis was carried out, followed by the application of the proposed novel methodology. The model is validated on an independent test set of stations. The outcome of the case study is a time series of hourly maps of wind field at 250 meters spatial resolution, which is highly relevant for renewable energy potential assessment.

In conclusion, the study introduced a new way to interpolate irregular spatio-temporal datasets. Further developments of the methodology could deal with the investigation of alternative basis such as Fourier and wavelets.



N. Cressie, C. K. Wikle, Statistics for Spatio-Temporal Data, Wiley, 2011.

M. Kanevski, A. Pozdnoukhov, V. Timonin, Machine Learning for Spatial Environmental Data, CRC Press, 2009.

How to cite: Guignard, F., Amato, F., Robert, S., and Kanevski, M.: Spatio-Temporal Modeling of Wind Speed Using EOF and Machine Learning , EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-9186, https://doi.org/10.5194/egusphere-egu2020-9186, 2020

D2195 |
| Highlight
Federico Amato, Fabian Guignard, and Mikhail Kanevski

Global climate has been the focus of an increasing number of researches over the last decades. The ratification of the Paris Agreement imposes to undertake the necessary actions to limit the increase in global average temperature below 1.5oC to ensure a reduction of the risks and impacts of climate change.

Despite the importance of its spatial and temporal distribution, warming has often been investigated only in terms of global and hemispheric means. Moreover, although it is known that climate is characterised by strong nonlinearity and chaotic behaviour, most of the studies in climate science adopt statistical methods valid only for stationary or linear systems. Nevertheless, it has already been shown that warming trends are characterised by strong nonlinearities, with an acceleration in the increase of temperatures since 1980.

In this work, we investigate the complex nature of global temperature trends. We study the maximum temperature at two meters above ground using the NCEP CDAS1 daily reanalysis data, with a spatial resolution of 2.5o by 2.5o and covering the time period from 1 of January 1948 to 30 of November 2018. For each spatial location, we characterize the corresponding temperature time series using methods from Information Theory. Specifically, we analysed the temperature by computing the Fisher Information Measure [1] (FIM) and the Shannon Entropy Power [2] (SEP) in a temporal sliding window, which allows to follow the temporal evolution of the two parameters. We find a significant change in the spatial patterns of the dynamic behaviour of temperatures starting from the early eighties. Specifically, two different patterns are recognizable. In the period from 1948 to the early eighties the latitudes higher than 60oN and lower than 60oS show high levels of SEP and low levels of FIM. The situation completely revers starting from 1980s, and in a faster way for the latitudes higher than 60oN, so that tropical and temperate zones are now characterized by high levels of entropy. The stronger growth of SEP is measured in the northern mid-latitudes. These regions are also known to have been characterized by higher warming trends. Finally, a drastic difference between oceans and land surfaces is detectable, with the former generally interested by significant increases of SEP since the eighties.

[1] Fisher, R.  A Theory of statistical estimation. Math. Proc. Camb. Philos. Soc.22, 700–725, DOI:  10.1017/S0305004100009580 (1925).

[2] Shannon, C. E.  A mathematical theory of communication. Bell Syst. Tech. J.27, 379–423, DOI: 10.1002/j.1538-7305.1948.tb01338.x (1948).

How to cite: Amato, F., Guignard, F., and Kanevski, M.: Spatio-temporal global patterns of 70 years of daily temperature using Fisher-Shannon complexity measures, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-9206, https://doi.org/10.5194/egusphere-egu2020-9206, 2020