Quantifying Landscape Changes through Land Cover Transition Potential Analysis and Modeling (on the example of the Black Tisza river basin)

Quantifying Landscape Changes through Land Cover Transition Potential Analysis and Modeling (on the

example of the Black Tisza river basin)

Alexander MKRTCHIAN1, Daria SVIDZINSKA2

1 Ivan Franko National University of Lviv, Lviv, 79000, Ukraine; [email protected];

2 Taras Shevchenko National University of Kyiv, Kyiv, 01601, Ukraine; [email protected]

The study area

Ukraine, Transcarpathian region

24.14° – 24.55° E, 48.07° – 48.40° N

Total area: 567.72 km2

Elevation range: 447 – 2 004 m a.s.l.

Black Tisza watershed

Black Tisza

Data sets used:Dataset PurposeLandsat 5 ETM+ Surface Reflectance data,bands 1 – 4scenes for 08 Jul 1989, 02 Aug 1998, 15 Jul 2009

Distinguishing land cover classes

SRTM 1 Arc-Second (30 meters)Global Digital Elevation Dataset

Morphometric factors distribution

OpenStreetMap vector data Road network location

WorldClim dataset Heat conditions modeling

Software used:Software PurposeQGIS core functionality Data visualization, map

creation

QGIS Semi-Automatic Classification plugin Land cover classification

QGIS MOLUSCE plugin Land cover changes analysis and simulation

R Logistic regression modeling

R hier.part package Hierarchical partitioning of variables

FragStats Landscape metrics calculation

Land cover classes, 1989 Land cover classes, 1998

Image classification results

Class ID Class area, ha Changes in class area

(1989/98), %

Changes in class area

(1998/09), %1989 1998 2009

1 371 285 318 -23.18 11.58

2 336 252 125 -25.00 -50.40

3 31434 29357 29423 -6.61 0.22

4 2080 2174 877 4.52 -59.66

5 11691 12396 13306 6.03 7.34

6 10468 11916 12331 13.83 3.48

The Land cover class area distributions and changes

Class 1 represents artificial surfaces like buildings, paved roads, etc.Class 2 represents fresh clearances where forest has been recently removed.Class 3 relates to forests in general: while there are different types of forests in the study area, distinguishing between them is complicated because of the prevalence of forests where different deciduous and coniferous tree species are mixed with varied proportions. Class 4 is formed by the merging of three spectrally different primary classes each of which occurs almost exclusively on highland locations above 1600 m a.s.l.Class 5 corresponds to shrubs and tenuous and low tree stands, mostly – secondary communities formed on places of former clearances and abandoned agricultural lands.Class 6 represents the grassland vegetation (except of grasslands located above treeline) – natural meadows, pastures, and scattered plots of different annual crops.

Deforestation extent

1989-98 1998-09

Deforestation defined as the transition of forest class 3 to any of the other Land cover classes

Explanatory variables:

To analyze the factors and to reveal the driving forces of LC changes,the set of spatially distributed variables were suggested to serve as explanatory variables

Heat resourcesTerrain ruggedness

Explanatory variables:

Cost distance from the major roads(terrain ruggedness as cost surface)

Cost distance from grasslands class 3(terrain ruggedness as cost surface)

Explanatory variables:

Cost distance from artificial surfaces class 1(terrain ruggedness as cost surface)

Cost distance from forest class 4(terrain ruggedness as cost surface)

Explanatory variable Deforestation, 1989-98

Deforestation, 1989-98

z I z IElevation -100.4 1.7 -50.1 0.45Terrain ruggedness 57.7 1.58 -48.1 0.5Cost distance from major roads -69.5 0.54 -69.8 0.85Cost distance fromartificial surfaces class

-22.9 0.12 -87.7 2.12

Cost distance fromgrasslands class

-75.1 2 -56.7 0.59

Heat resources 99.8 1.53 46.1 0.35Overall AIC 294667 298221

Multiple regression andhierarchical partitioning results:

The relative influence of the main factors of deforestation

1989-98 1998-09

1998-091989-98

Multiple regression andhierarchical partitioning results:

The relative influence of themain factors of reforestation

Explanatory variable Reforestation, 1989-98

Reforestation, 1998-09

z I z IElevation 33.1 0.77 6.8 0.47Terrain ruggedness 28.7 0.23 89.0 3.42Cost distance from major roads 72.0 2.19 46.7 1.15Cost distance fromartificial surfaces class

30.5 0.1 26.5 0.13

Cost distance from forest class -116.7 10 -117.6 8.73Heat resources -40.8 0.66 -10.7 0.33Overall AIC 198257 287568

Boxplots of some major factors relationships with land cover transition types

Actual land cover classes, 2009 Simulated land cover classes, 2020

The land cover projections have been produced by MOLUSCE plug-in for QGIS,based on the land cover transition potential modeling

Land cover classes simulation certainty, 2020Land cover classes simulation, 2020

Simulation method % of correct predictions kappa

LRM 62.4 0.42ANN 71.1 0.54

Validation results of simulations for 2009

The MOLUSCE plug-in also produces the spatially distributed measures of simulation certainty.The simulation can also be validated by comparison with actual data, when these are available

The temporal dynamics of forest class landscape metrics

Year Total Area, ha

% of Landscape

Perimeter-area fractal dimension

Total Core Area, ha

% of Core Area

Total edge,km

Mean patch area, ha

1989 31434 55.8 1.274 23844 75.85 1762 77.9

1998 29357 52.07 1.264 22469 76.54 1663 65.4

2009 29423 52.19 1.272 21694 73.73 1763 49.4

Forest ecosystem fragmentation has a number of ecological effects on forest species and communities,introducing edge effects leading to habitat and forest biodiversity loss.

Several landscape metrics characterizing forest habitat fragmentation have been calculated using FragStats 4.2,and their changes for the two decades have been assessed.

While the total forested area has increased a little from 1998 to 2009, the total core area in fact continued to decrease due to intensification of the fragmentation of forest patches, as is evident from the changes in class landscape metrics.

Thank you for attention Alexander [email protected]

Daria [email protected]