Alas Landscape Modeling by Remote Sensing Image Analysis and

Geographic Ontology: Study Case of Central Yakutia (Russia)

Sébastien Gadal

1,2 a

, Moisei Zakharov

1,2 b

, Jūratė Kamičaitytė

and Yuri Danilov

Aix-Marseille Univ., CNRS, ESPACE UMR 7300, Univ., Nice Sophia Antipolis,

Avignon Univ., 13545 Aix-en-Provence, France

North-Eastern Federal University, 670000 Yakutsk, Republic of Sakha, Russian Federation

Kaunas University of Technology, Kaunas, Lithuania

Keywords: Geographic Ontology, Image Analysis, Knowledge Database, Image Processing, Alas Landscape, Remote

Sensing.

Abstract: Approaches of geographic ontologies can help to overcome the problems of ambiguity and uncertainty of

remote sensing data analysis for modeling the landscapes as a multidimensional geographic object of research.

Image analysis based on the geographic ontologies allows to recognize the elementary characteristics of the

alas landscapes and their complexity. The methodology developed includes three levels of geographic object

recognition: (1) the landscape land cover classification using Support Vector Machine (SVM) and Spectral

Angle Mapper (SAM) classifiers; (2) the object-based image analysis (OBIA) used for the identification of

alas landscape objects according to their morphologic structures using the Decision Tree Learning algorithm;

(3) alas landscape’s identification and categorization integrating vegetation objects, territorial organizations,

and human cognitive knowledge reflected on the geo-linguistic object-oriented database made in Central Ya-

kutia. The result gives an ontology-based alas landscape model as a system of geographic objects (forests,

grasslands, arable lands, termokarst lakes, rural areas, farms, repartition of built-up areas, etc.) developed

under conditions of permafrost and with a high sensitivity to the climate change and its local variabilities. The

proposed approach provides a multidimensional reliable recognition of alas landscape objects by remote sens-

ing images analysis integrating human semantic knowledge model of Central Yakutia in the subarctic Siberia.

This model requires to conduct a multitemporal dynamic analysis for the sustainability assessment and land

management.

1 INTRODUCTION

The problem of knowledge integration in remote

sensing for the studies of landscape characteristics is

one of the key gaps; especially for the modeling of the

territorial sustainability assessment and ongoing fu-

ture landscape changes (Konys, 2018). The expert ge-

ographic knowledge building based on geographic

ontologies is the most suitable approach. It allows the

integration of expert-human knowledge for the re-

mote sensing image analysis and modeling of the

complexity of the alas landscape system. The ap-

proach based on geographic ontology can be consid-

ered as a specific domain of knowledge in the field of

https://orcid.org/0000-0002-6472-9955

https://orcid.org/0000-0002-8916-2166

https://orcid.org/0000-0002-8821-5764

https://orcid.org/0000-0003-4806-4938

the artificial intelligence. Geographical ontologies

used for the alas landscape modeling includes: (a) bi-

ophysical geographic components of landscape (e.g.,

vegetation, moisture, landform (relief), climate con-

ditions, soil (Pashkevich, 2017)); (b) morphologic

forms and geometric objects characteristics (e.g.,

shape, size, textures, topology, geolocation, depres-

sion, inclination, slope), and (c) anthropogenic ob-

jects related to the land field reality measured (cap-

tured) by remote sensing, the human geographic ob-

ject’s identification (semantic meaning) and percep-

tions. The human geographic object’s identification

(semantic meaning) was formalized in Russian and

112

Gadal, S., Zakharov, M., Kami

caityt

e, J. and Danilov, Y.

Alas Landscape Modeling by Remote Sensing Image Analysis and Geographic Ontology: Study Case of Central Yakutia (Russia).

DOI: 10.5220/0009569101120118

In Proceedings of the 6th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2020), pages 112-118

ISBN: 978-989-758-425-1

Yakutsk languages in a geolinguistic database inte-

grating toponyms, territorial knowledge of local pop-

ulation, census, land uses, maps, etc. (Zamorshikova

et al, 2018). This geo-linguistic database including

the bunch of heterogeneous knowledge poses the

problem of standardization (harmonization). The ap-

plication in the alas landscape combining the

knowledge model, remote sensing and geographic on-

tologies is based on the theoretical model of ontology

(Sinha and Mark, 2010) and on the methodology de-

veloped in remote sensing applied to the subarctic ur-

ban environment in Yakutsk (Gadal and

Ouerghemmi, 2019).

2 ISSUES OF CLIMATE

CHANGES OF THE ALAS

LANDSCAPE

The interest to study the alas landscape is related to

the impacts of climate change and modeling the land-

scape processes. The alas landscapes are localized in

Taiga boreal forest and permafrost territories of the

subarctic region of Siberia, Alaska and Canada. They

are formed as a result of thawing of the permafrost

layer in the exposed parts of the taiga. It makes the

alas one of the key indicators of the permafrost deg-

radation and vulnerability of landscapes in the sub-

arctic region (Fedorov, 2019) and has a major contri-

bution on carbon (CH

and CO₂) emission. They have

impact on the forms of steppe azonal soils which have

a high bio-productivity for the agriculture. Agricul-

ture in the subarctic regions is the main framework of

the traditional local lifestyles of Yakutsk and Ekenkii.

According to it, the cultural, social and territorial sig-

nificance of the alas is defined (Desyatkin et al,

2018).

Therefore, the modeling of the alas landscape

needs to integrate both the human semantic factors

and the geo-physical reality, associating quantitative

and semantic parameters defining ontologies of the

alas geographic space. The approach for recognizing

the geographic ontologies merges manually collected

data from the geographic landscape (field observa-

tions, population knowledge formalized in geo-lin-

guistic database), maps and remote sensing data:

Landsat series, 5 TM, 7 ETM+, 8 OLI and Sentinel 2.

Geographic object ontologies are extracted from the

satellite images by semi-automatic or automatic ap-

proach of image analysis (Clouard et al, 2013) using

machine learning algorithms Spectral Angle Mapper

(SAM) on the Landsat series images according to the

better performance (overall accuracy). Two ap-

proaches were applied for the Sentinel 2 images: (a)

the Support Vector Machine (SVM) for the recogni-

tion of the geographic objects by pixel-based classifi-

cation (Li et al 2009, Ouerghemmi et al, 2017) show-

ing the land cover landscape of the alas and the envi-

ronment around, and (b) the OBIA based on the De-

cision Tree algorithm (DT) (Vopham et al, 2018) and

the morphological image filtering based on the con-

trast and texture information through large-scale

smoothing for image segmentation (Maragos, 1989,

Sofou et al, 2005) that helped to identify automati-

cally the current different geographic objects defining

the actual alas. Results are integrated and merged in

GIS. It must be noted that the application of the DT

algorithm on the Landsat TM, ETM+, OLI is not rel-

evant because of the too low GSD of 30m.

2.1 Study Area

The Central Yakutia lowland is characterized by the

widespread distribution of Alas. Throughout the ter-

ritory there are about 16000 alas between Vilyuy,

Lena and Aldan Rivers. The study area is located at

40 km distance to the North from the city of Yakutsk

covering the surface of 30000 hectares (Fig. 1). The

study area is a boreal forest represented by pine and

larch. The soil basis is permafrost with a layered and

lenticular texture, where thermokarst cryogenic pro-

cesses are developed, which cause the formation of

numerous alas.

Figure 1: The study area [© Open Street Map]; the extent

of the Sentinel 2A RGB scene at 26th of July 2019.

Alas Landscape Modeling by Remote Sensing Image Analysis and Geographic Ontology: Study Case of Central Yakutia (Russia)

113

3 IMAGE ANALYSIS OF ALAS

LANDSCAPE STRUCTURE

3.1 Methodology

The geographic objects as ontological elements are

the basis of the structure and organization of a land-

scape (Gadal, 2012). In remote sensing they are iden-

tified by image processing and image analysis (here

by machine learning, landscape object classifications,

and OBIA image segmentation) with the knowledge

GIS database and the implementation of ontological

rules (Fig. 2).

Figure 2: Implementation of image analysis with geo-

graphic ontologies for alas landscape recognition.

The typical alas landscapes consist of the follow-

ing elementary geographic objects: lake, grassland,

woodland, built-up and arable lands for agriculture.

First, they are identified and analyzed according to

the land cover classes produced from the biophysical

parameters measured spectrally by Landsat series and

Sentinel 2 data.

The second level of the geographical objects

recognition by object-oriented classification algo-

rithms of DT is resulted from knowledge acquisition

(Clouard et al, 2010) structured in the geo-linguistic

database and integrated in GIS, and the OBIA to ex-

tract the morphology structures of the alas landscape.

The categorization of the alas landscapes objects gen-

erated is followed-on by the fusion of land cover

maps and alas landscape object classification.

The data analyzed include the expert and the do-

main knowledge (landscape theories, concepts, indic-

ative and cognitive analysis products) (Gadal and

Ouerghemmi, 2019). The ontological approach ena-

bles us to obtain numerous representations of all as-

pects of the alas landscape spatial organization

through the accurate acquired knowledge structured

in the GIS database. It is used both for training, learn-

ing and the validation of results, the spatial analysis

and modeling in GIS.

The integrity and the logic coherence of all pro-

cesses taking place in geographic space is ensured by

the integration of semantic geographic object ontolo-

gies modelled by the geostatistical analysis of geo-

linguistics database (toponyms, territorial knowledge

of local population, census, land uses, maps, etc.)

with the categorization of alas landscape based on

GIS knowledge database.

The obtained alas landscape model (Fig. 3) is a

multidimensional model including knowledge and

geo-linguistic GIS database adapted for machine

learning training and simulation of geospatial pro-

cesses: rural exodus, permafrost degradation, etc.

having impact on the landscape’s changes and their

consequences (sustainability, re-greening).

Figure 3: Alas landscape model map with knowledge and

geo-linguistic GIS database. Terra-Metrics, DigitalGlobe

2012.

3.2 Image Processing

3.2.1 Recognition of Objects of Landscape

Land Cover Change

The recognition of the landscape objects and the

change dynamics modeling between 2007 and 2019

is based on the use of the Landsat 5 TM (date:

15.07.2007), Landsat 8 OLI (bands 2 to 7: VIR-NIR-

SWIR) pansharpened at 15m (date: 26.07.2013) and

Sentinel 2A (bands 2 to 5: VIR- NIR at 10m) (date:

24.07.2019) acquired in July.

Landscape land covers produced and mapped by

supervised classification are used as training data of

the GIS database. Landscape objects are recognized

by the spectral signatures related to the geographic

objects of the GIS database. Several classifier algo-

rithms were tested: Minimum Distance (MD), Maxi-

mum Likehood (ML), the K-Nearest Neighbors (K-

NN), Spectral Angle Mapper (SAM) and Support

Vector Machine (SVM).

The results show that the optimal machine learn-

ing classifiers are obtained with SAM for the Landsat

5 TM and the Landsat 8 OLI data confirming the re-

sults presented by Pertopoulos et al. (2010) (Tab. 1).

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

114

The most reliable results for the landscape land cover

were proceeded with the SVM based on Kernel

method for the Sentinel 2A images (Pal and Mather,

2005) (Fig. 4, 5).

Table 1: Results obtained from applied classifiers.

Classifi-

cation

algorithm

O.A. / Kappa

2007 2013 2019

MD 72%/0,45 67%/0,44 45%/0.37

ML 43%/0,31 56%/0.39 54/0.36

K-NN non non 34%/0.21

SAM 77%/ 0.57 81%/ 0.64 67%/0,59

SVM non non 83%/0,65

The accuracy of landscape land cover classifica-

tions is provided with 50 randomly generated poly-

gons (3x3 pixels size). The random set of vector data

(by 3x3 pixels) constitutes the validation data set.

They are used for the accuracy assessment of the clas-

sification by confusion matrix and Kappa coefficient.

Figure 4: Land cover classification by SAM for Landsat 5

TM in 2007 (a) and Landsat 8 OLI in 2013 (b).

Figure 5: Land cover classification result for Sentinel 2A

with SVM classifier.

3.2.2 Alas Recognition by Object-based

Image Analysis

The recognition of alas is based on two criteria that

ontologically characterize geographical objects: mor-

phology and their own geometries. They lend them-

selves to OBIA procedures. The satellite images used

are from the VIR and NIR bands of Sentinel 2A, in

particular because of their spatial resolution of 10m

that enables to recognize the geometric shapes and the

structures of the alas. The alas as a unique geographic

object is characterized ontologically by its geometry

and by the geographic objects that compose it such as

lakes, meadows, swamps, the presence of farms and

forests surrounding it. These geographic objects are

identified by classification (part 3.2.1.) with the land-

scape cover maps (fig. 4, fig. 5).

The detection, identification and extraction of alas

from their morphological characteristics lends itself

well to the OBIA object-oriented machine learning

classification algorithms. The OBIA alas recognition

consists of four phases: (a) The dimensionality reduc-

tion by the Principal Component Analysis (PCA) al-

gorithm, and linear combination of original bands that

contains 91-95% of spectral information in first three

components (Fig 6a). (b) Smoothing filtering that is

applied for the suppression of noise and homogenize

statistically the alas regions. This filtering method as-

sociates with each pixel of the image the closest local

mode in the density distribution of the combined re-

gion (Comaniciu and Meer, 1999, 2002). (c) The im-

age segmentation by Large Scale MeanShift non-par-

ametric and iterative clustering method (Michel et al,

2015). The segmentation produces a labeled image

with tile-wise processing where neighbor pixels

whose range distance is below range radius and op-

tionally spatial distance below spatial radius are

merged into the same raster value (Fig. 6c). The aver-

age spatial neighborhood radius is 25 meters.

Figure 6: Object recognition from Sentinel 2A (a) RGB of

PCA components; (b) MeanShiftSmoophing filtering; (3)

Image segmentation of alas.

The most complete extraction of alas is obtained

with the following parameters: the spatial radius – 50;

the range radius – 25; the Minimum segment size –

10. The main geometrical parameters distinguishing

alas and thermokarstic basins concern their shapes:

Alas Landscape Modeling by Remote Sensing Image Analysis and Geographic Ontology: Study Case of Central Yakutia (Russia)

115

alas has a shape close to the circle, does not have dis-

tinct angles, and is not characterized by elongation

contrary to thermokarstic shapes. (d) The fourth

phase consists in separating the alas from another an-

thropogenic objects: agricultural parcels. The identi-

fication methodology uses a decision tree algorithm.

This machine learning algorithm required a dataset

that selected 30 polygons generated using the GIS da-

tabase, more specifically the alas toponyms, topo-

graphic maps and Open Street Map data. The classi-

fication structure defined by the decision tree is esti-

mated from the training data using a statistical proce-

dure of overall mean and standard deviation for 3

PCA components. The decision rules developed are

used to associate each segment of the image with one

of the geographic object classes. The O.A. with Sen-

tinel 2A data is =88.4%; a Kappa coefficient is 0,83

for the Decision Tree classifier.

The result is two broad types of geographic ob-

jects in which elementary ontological and geographic

structures are reflected: alas and arable lands of agri-

culture (Fig. 7). The information was obtained

through the interpretation of remote sensing data by

artificial intelligence with the integration of expert

knowledge for learning and validation of the obtained

results. This GIS-structured information set provides

the basis for a comprehensive spatial analysis using

semantic ontology to model the landscape dynamics

of the boreal forests of alas and permafrost.

Figure 7: Alas identification by OBIA with Sentinel 2A im-

age.

4 ALAS LANDSCAPE

MODELING

In this part, some applications of the image analysis

data obtained in the modeling of the alas landscapes

are presented. The ecological status of the alas land-

scapes plays an important socio-economic role, as

well as in the dynamic processes of the permafrost.

The results of spatial remote sensing image analysis

using GIS-integrated geographic ontologies are de-

signed to serve as a basis and environment for this

research. Thus, the modeling of geo-linguistic land-

scape features has been developed in GIS with the

creation of an extensive geographically referenced

database. It aims to understand better the land use and

to measure the complexity of landscape structure.

The first method consists of performing modeling

analysis land covering changes within alas. The four

categories were identified according to the structure

of alas vegetation: (a) Termokarst lake, when the alas

basin is completely filled with water; (b) Complex

alas, when vegetation changes from steppe to coastal-

aquatic; (c) herbaceous alas, when vegetation is rela-

tively uniform throughout all alas territories; (d) pas-

ture alas, with active arable or hay areas. The appli-

cation of knowledge about the vegetation species of

alas allows interpreting land cover classes and to as-

sess the stability and transformation of the alas land-

scape. In total, the alas include 5 typical land cover

classes. A highly moistened alas section belongs to

the grass, less – often to the forest due to close vege-

tative indicators (chlorophyll intensity). In some

cases, the presence of forests inside the alas indicates

the presence of a hill, that are indicators of the pres-

ence of permafrost heaving tubercles – pingo. The ar-

able land is the middle alas section, the most bio-pro-

ductive section often used as a forage base for cattle-

breeding and horse-breeding. The peripheral portion

of alas with a lack of moisture forms the azonal steppe

vegetation and stands out as a class of bare soils. Ta-

ble 2 shows examples of alas with the percentage

structure of the land cover and the corresponding cat-

egory.

This categorization allows us to consider the stage

of spatio-temporal development of alas, water con-

tent, and to obtain data on anthropization and eco-

nomic activity. This model also has a perspective for

the development of simulation strategies for the trans-

formation of plains landscapes of permafrost.

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

116

Table 2: Examples of alas landscape categorization by land

cover structure.

Name Year

Land cover classes (%)

Alas

category

forest

Ara

ble

land

grass

bare

soil

water

Ton-Bas

(Тон-

Бас)

2007 17 2 34 0 48

termokarst

lake

2013 18 6 23 17 35

complex

alas

2019 5 10 63 20 2

complex

alas

Mokh-

sogo-

lokh

(Мохсо

голлох)

2007 53 0 9 0 37

termokarst

lake

2013 1 0 99 0 0

herba-

ceous alas

2019 5 38 57 0 0

pasture

alas

Sier-Bie

(Сиэр-

Бие)

2007 1 0 81 0 18

herba-

ceous alas

2013 1 45 53 0 1

pasture

alas

2019 1 47 42 0 11

pasture

alas

The second method is based on landscape seman-

tic analysis using the geo-linguistic database. The alas

landscapes are the main source of livelihood for the

local population, which implies the presence of a cul-

tural layer of information of the geographical ontol-

ogy, within the framework of local knowledge about

the geographical space. Semantics provides a transi-

tion from the object level of the geographical ontol-

ogy, where the main concept is geographical objects

represented in our case as land-use classes - the alas -

and the results of their modeling according to the ex-

pert knowledge (alas categories, vegetation species),

to the spatial level of the geographical ontology. This

territorial knowledge generates geographic models

recognition and identification of which constitute a

new challenge for artificial intelligence in space and

airborne remote sensing.

The spatial organization of the local economy in-

volves the location of villages and farms in the alas.

Due to their rarity, they cannot be recognized by im-

age processing from medium spatial resolution im-

ages such as Sentinel 2A/B, Landsat 5/6/7/8, and are

included in the class of arable (agricultural) land.

When their presence was known, we reclassified the

pixel sets of the arable (agricultural) land value into

the buildings class by comparing them with the se-

mantic layer of settlements and sayylyks (summer

houses).

An initial geo-linguistic analysis also allows us to

identify the alas used in agriculture, which may have

been abandoned recently. Compared to satellite im-

ages analysis, the absence of a semantic layer can in-

dicate the age and time of alas formation. The GIS

semantic database also allows the dissemination of re-

mote sensing knowledge and facilitates the interpre-

tation of landscape ontology. It provides a semantic

description of the results of analysis and processing

of remote sensing images.

5 CONCLUSIONS

Geographic ontologies image analysis in artificial in-

telligence is a necessary part of knowledge that com-

bines multi-level and heterogeneous concepts of an

object and a geographic space, and as well geo-lin-

guistics in order to implement a systematic approach

to the modeling of alas landscapes. Although, the de-

veloped model confronts the technical issues, such as

the standardization and harmonization of heterogene-

ous information flows, it also establishes the rules for

solving decision-making problems, for environmen-

tal and agricultural monitoring under the conditions

of continuously expanding permafrost in subarctic

climate.

This research significantly contributes to the de-

velopment of application of knowledge model to im-

prove remote sensing image processing analysis and

landscape modeling for the territorial and geographic

studies. The novelty of the research comprises the in-

tegration of the geo-linguistic database reflecting the

semantic meaning of people's knowledge as part of

the geographic ontology of alas landscapes. It allows

to analyze the value, the age and the evolution of alas

increasing the accuracy of recognition of alas objects.

One of the main challenges is to hybridize geo-

graphical knowledge with artificial intelligence used

to analyze the landscapes and the cross processes of

anthropization and geophysical evolution of perma-

frost by spatial and airborne remote sensing.

ACKNOWLEDGEMENTS

This research was supported by the French National

Research Agency (ANR) through the PUR project

(ANR-14-CE22- 0015), the Russian Science Founda-

tion № 15-18-20043 “Landscape Ontology: Seman-

tics, Semiotics and Geographical Modeling”, and the

Vernadski Grant of the French Embassy in the Rus-

sian Federation.

Alas Landscape Modeling by Remote Sensing Image Analysis and Geographic Ontology: Study Case of Central Yakutia (Russia)

117

REFERENCES

Clouard, R., Renouf, A., Revenu, M., 2010. An Ontology-

Based Model for Representing Image Processing Ob-

jectives. International Journal of Pattern Recognition

and Artificial Intelligence, World Scientific Publishing,

pp.1181-1208.

Comaniciu, D. and Meer, P. 1999. Mean shift analysis and

applications. International Conference on Computer

Vision, 2, pp. 1197-1203.

Comaniciu, D., and Meer, P. 2002. Mean shift: A robust

approach toward feature space analysis. IEEE Transac-

tions on pattern analysis and machine intelligence,

24(5), pp. 603-619.

Desyatkin, A., Takakai, F. and Nikolaeva, M. 2018. Land-

scape Microzones within Thermokarst Depressions of

Central Yakutia under Present Climatic Conditions. Ge-

osciences 2018, 8. MDPI

Fedorov, A., 2019. Permafrost Landscapes: Classification

and Mapping. Geosciences 2019, 9. MDPI.

Gadal, S., 2012. Geographic Space Ontology, Locus Ob-

ject, and Spatial Data Representation Semantic Theory.

Universal Ontology of Geographic Space: Semantic

Enrichment for Spatial Data, IGI Global, pp.28-52,

9786-1-4666-032-27-1.

Gadal, S., and Ouerghemmi, W., 2019. Knowledge Models

and Image Processing Analysis in Remote Sensing: Ex-

amples of Yakutsk (Russia) and Kaunas (Lithuania). In

Proceedings of the 5th International Conference on Ge-

ographical Information Systems Theory, Applications

and Management, Heraklion, Crete – Greece. pp. 282-

288.

Konys, A., 2018. An Ontology-Based Knowledge Model-

ling for a Sustainability Assessment Domain. In Sus-

tainability Journal 10.

Kordjamshidi, P., Moens, M-F. 2014. Global Machine

Learning for Spatial Ontology Population. Journal of

Web Semantics pp. 1-29

Li, J., Carlson, B.E., and Lacis A.A., 2009. A study on the

temporal and spatial variability of absorbing aerosols

using total ozone mapping spectrometer and ozone

monitoring instrument aerosol index data. J. Geophys.

Res., 114.

Maragos, P. 1989 A representation theory for morphologi-

cal image and signal processing. IEEE Transactions on

Pattern Analysis and Machine Intelligence 11.6 pp.

586−599.

Michel, J., Youssefi, D., & Grizonnet, M. (2015). Stable

mean-shift algorithm and its application to the segmen-

tation of arbitrarily large remote sensing images. IEEE

Transactions on Geoscience and Remote Sensing,

53(2), pp. 952-964.

Ouerghemmi, W., Gadal, S., Mozgeris, G., Jonikavičius,

D., and Weber, C., 2017. Urban objects classification

by spectral library: feasibility and applications. In 2017

Joint Urban Remote Sensing Event (JURSE), Dubai,

pp. 1- 4.

Sinha, G., Mark, D., 2010. Toward A Foundational Ontol-

ogy of the Landscape. MDPI.

Sofou, A., Evangelopoulos, G., Maragos, P. 2005. Soil Im-

age Segmentation and Texture Analysis: A Computer

Vision Approach. IEEE Geoscience and Remote Sens-

ing Letters, Vol. 2, No. 4, pp. 394-398

Pal, M. and Mather, P. 2005. Support vector machines for

classification in remote sensing. In International Jour-

nal of Remote Sensing Vol.5. pp. 1007-1011

Pashkevich, M., 2017. Classification and Environmental

Impact of Mine Dumps. In Assessment, Restoration and

Reclamation of Mining Influenced Soils. pp. 1-32

Petropoulos, G., Vadrevu, K., Xanthopoulos, G., Karantou-

nias, G., Scholze, M. 2010. A Comparison of Spectral

Angle Mapper and Artificial Neural Network Classifi-

ers Combined with Landsat TM Imagery Analysis for

Obtaining Burnt Area Mapping. Sensors. pp. 1967-

1985.

Puebla-Martınez, M., Perea-Ortega, J., Simon-Cuevas, A.,

and Romero, F. 2018. Automatic Expansion of Spatial

Ontologies for Geographic Information Retrieval.

Communications in Computer and Information Sci-

ence. pp. 659-670

VoPham, T., Hart, J., Laden, F. and Chiang, Y. 2018.

Emerging trends in geospatial artificial intelligence (ge-

oAI): potential applications for environmental epidemi-

ology. Environ Health 17, 40

Zamorshchikova, L., Khokholova, I., Ikonnikova, A., Sam-

sonova, M. and Lebedeva, V. 2018. Toponymic Land-

scape of Central Yakutia: Etymological Analysis of Ge-

ographical Names. Journal of Siberian Federal Univer-

sity. Humanities & Social Sciences.

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

118