2020
Wellmann, Thilo; Lausch, Angela; Scheuer, Sebastian; Haase, Dagmar
In: Ecological Indicators, vol. 111, pp. 106029, 2020.
Abstract | Links | BibTeX | Tags: Leipzig, Machine learning, Random forest, RapidEye, Remote Sensing, Species Distribution Models, Spectral trait variations, Spectral traits, Urban birds
@article{wellmann2020earth,
title = {Earth observation based indication for avian species distribution models using the spectral trait concept and machine learning in an urban setting},
author = {Thilo Wellmann and Angela Lausch and Sebastian Scheuer and Dagmar Haase},
url = {https://thilowellmann.de/wp/wp-content/uploads/2020/01/WellmannEtAl_BreedngbirdsEO_SDM_Leipzig_AcceptedManuscript.pdf},
doi = {10.1016/j.ecolind.2019.106029},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
journal = {Ecological Indicators},
volume = {111},
pages = {106029},
publisher = {Elsevier},
abstract = {Birds respond strongly to vegetation structure and composition, yet typical species distribution models (SDMs) that incorporate Earth observation (EO) data use discrete land-use/cover data to model habitat suitability. Since this neglects factors of internal spatial composition and heterogeneity of EO data, we suggest a novel scheme deriving continuous indicators of vegetation heterogeneity from high-resolution EO data.
The deployed concepts encompass vegetation fractions for determining vegetation density and spectral traits for the quantification of vegetation heterogeneity. Both indicators are derived from RapidEye data, thus featuring a continuous spatial resolution of 6.5 m. Using these indicators as predictors, we model breeding bird habitats using a random forest (RF) classifier for the city of Leipzig, Germany using a single EO image.
SDMs are trained for the breeding sites of 44 urban bird species, featuring medium to very high accuracies (59–90%). Analysing similarities between the models regarding variable importance of single predictors allows species groups to be determined based on their preferences and dependencies regarding the amount of vegetation and its spatial and structural heterogeneity. When combining the SDMs, models of urban bird species richness can be derived.
The combination of high-resolution EO data paired with the RF machine learning technique creates very detailed insights into the ecology of the urban avifauna, opening up opportunities of optimising greenspace management schemes or urban development in densifying cities concerning overall bird species richness or single species under threat of local extinction.},
keywords = {Leipzig, Machine learning, Random forest, RapidEye, Remote Sensing, Species Distribution Models, Spectral trait variations, Spectral traits, Urban birds},
pubstate = {published},
tppubtype = {article}
}
Birds respond strongly to vegetation structure and composition, yet typical species distribution models (SDMs) that incorporate Earth observation (EO) data use discrete land-use/cover data to model habitat suitability. Since this neglects factors of internal spatial composition and heterogeneity of EO data, we suggest a novel scheme deriving continuous indicators of vegetation heterogeneity from high-resolution EO data.
The deployed concepts encompass vegetation fractions for determining vegetation density and spectral traits for the quantification of vegetation heterogeneity. Both indicators are derived from RapidEye data, thus featuring a continuous spatial resolution of 6.5 m. Using these indicators as predictors, we model breeding bird habitats using a random forest (RF) classifier for the city of Leipzig, Germany using a single EO image.
SDMs are trained for the breeding sites of 44 urban bird species, featuring medium to very high accuracies (59–90%). Analysing similarities between the models regarding variable importance of single predictors allows species groups to be determined based on their preferences and dependencies regarding the amount of vegetation and its spatial and structural heterogeneity. When combining the SDMs, models of urban bird species richness can be derived.
The combination of high-resolution EO data paired with the RF machine learning technique creates very detailed insights into the ecology of the urban avifauna, opening up opportunities of optimising greenspace management schemes or urban development in densifying cities concerning overall bird species richness or single species under threat of local extinction.
The deployed concepts encompass vegetation fractions for determining vegetation density and spectral traits for the quantification of vegetation heterogeneity. Both indicators are derived from RapidEye data, thus featuring a continuous spatial resolution of 6.5 m. Using these indicators as predictors, we model breeding bird habitats using a random forest (RF) classifier for the city of Leipzig, Germany using a single EO image.
SDMs are trained for the breeding sites of 44 urban bird species, featuring medium to very high accuracies (59–90%). Analysing similarities between the models regarding variable importance of single predictors allows species groups to be determined based on their preferences and dependencies regarding the amount of vegetation and its spatial and structural heterogeneity. When combining the SDMs, models of urban bird species richness can be derived.
The combination of high-resolution EO data paired with the RF machine learning technique creates very detailed insights into the ecology of the urban avifauna, opening up opportunities of optimising greenspace management schemes or urban development in densifying cities concerning overall bird species richness or single species under threat of local extinction.
2019
Haase, Dagmar; Jänicke, Clemens; Wellmann, Thilo
Front and back yard green analysis with subpixel vegetation fractions from earth observation data in a city Journal Article
In: Landscape and Urban Planning, vol. 182, pp. 44–54, 2019.
Abstract | Links | BibTeX | Tags: Leipzig, Private green, RapidEye, Remote Sensing, Spectral unmixing
@article{haase2019front,
title = {Front and back yard green analysis with subpixel vegetation fractions from earth observation data in a city},
author = {Dagmar Haase and Clemens Jänicke and Thilo Wellmann},
url = {https://thilowellmann.de/wp/wp-content/uploads/2018/11/HaaseJänickeWellmann_FrontBackyardGreen_AcceptedManuscript.pdf},
doi = {10.1016/j.landurbplan.2018.10.010},
year = {2019},
date = {2019-01-01},
urldate = {2019-01-01},
journal = {Landscape and Urban Planning},
volume = {182},
pages = {44--54},
publisher = {Elsevier},
abstract = {This paper introduces a novel approach to green space availability in cities that includes the thus-far mostly neglected urban front and backyard green space around residential buildings on privately owned ground. To quantify the full spatial scope of urban green space, we calculated subpixel vegetation fractions from RapidEye remote-sensing data for the entire city with a spectral unmixing technique that enabled us to model the extent of urban vegetation with a high degree of confidence (MAE 7%, R2 0.92). We then applied a new ‘urban front and back yard green space derivation algorithm’, namely, a masking of the fractional vegetation data using GIS vector data of land cover, in order to delineate the front and backyard greenspace of residential houses in a city with an accuracy of 96%. Combining these two approaches, we can calculate the area of urban front and back yard green space for the entire city (including different residential structure types) and compare this data to the area of public (parks, urban forests) and semi-public (allotment gardens) green spaces that have been used for prevailing per capita green space availability analyses. The new method is exemplified at the city of Leipzig, Germany, which provides different residential structures concerning house types and the surrounding green that are characteristic of many European cities. Key findings include that the total amount of urban front and back yard green space is almost 2000 ha, which is ∼40% of the amount of public green space (4768 ha). In 15 out of the 63 total districts, there is more front and backyard than public green space, which highlights the importance of these urban front and back yard green space for the analysis of urban livelihoods and a tool for detailed ecosystem services-oriented urban planning.},
keywords = {Leipzig, Private green, RapidEye, Remote Sensing, Spectral unmixing},
pubstate = {published},
tppubtype = {article}
}
This paper introduces a novel approach to green space availability in cities that includes the thus-far mostly neglected urban front and backyard green space around residential buildings on privately owned ground. To quantify the full spatial scope of urban green space, we calculated subpixel vegetation fractions from RapidEye remote-sensing data for the entire city with a spectral unmixing technique that enabled us to model the extent of urban vegetation with a high degree of confidence (MAE 7%, R2 0.92). We then applied a new ‘urban front and back yard green space derivation algorithm’, namely, a masking of the fractional vegetation data using GIS vector data of land cover, in order to delineate the front and backyard greenspace of residential houses in a city with an accuracy of 96%. Combining these two approaches, we can calculate the area of urban front and back yard green space for the entire city (including different residential structure types) and compare this data to the area of public (parks, urban forests) and semi-public (allotment gardens) green spaces that have been used for prevailing per capita green space availability analyses. The new method is exemplified at the city of Leipzig, Germany, which provides different residential structures concerning house types and the surrounding green that are characteristic of many European cities. Key findings include that the total amount of urban front and back yard green space is almost 2000 ha, which is ∼40% of the amount of public green space (4768 ha). In 15 out of the 63 total districts, there is more front and backyard than public green space, which highlights the importance of these urban front and back yard green space for the analysis of urban livelihoods and a tool for detailed ecosystem services-oriented urban planning.
2018
Wellmann, Thilo; Haase, Dagmar; Knapp, Sonja; Salbach, Christoph; Selsam, Peter; Lausch, Angela
Urban land use intensity assessment: The potential of spatio-temporal spectral traits with remote sensing Journal Article
In: Ecological Indicators, vol. 85, pp. 190–203, 2018.
Abstract | Links | BibTeX | Tags: GLCM, Hemeroby, Human-use-intensity, NDVI, RapidEye, Remote Sensing, Spectral trait variations, Spectral traits, Urban land-use-intensity
@article{wellmann2018urban,
title = {Urban land use intensity assessment: The potential of spatio-temporal spectral traits with remote sensing},
author = {Thilo Wellmann and Dagmar Haase and Sonja Knapp and Christoph Salbach and Peter Selsam and Angela Lausch},
url = {https://thilowellmann.de/wp/wp-content/uploads/2022/03/Wellmann_etal_Manuscript_U-LUI_2018.pdf},
doi = {10.1016/j.ecolind.2017.10.029},
year = {2018},
date = {2018-01-01},
urldate = {2018-01-01},
journal = {Ecological Indicators},
volume = {85},
pages = {190--203},
publisher = {Elsevier},
abstract = {By adding attributes of space and time to the spectral traits (ST) concept we developed a completely new way of quantifying and assessing land use intensity and the hemeroby of urban landscapes. Calculating spectral traits variations (STV) from remote sensing data and regressing STV against hemeroby, we show how to estimate human land use intensity and the degree of hemeroby for large spatial areas with a dense temporal resolution for an urban case study. We found a linear statistical significant relationship (p=0.01) between the annual amplitude in spectral trait variations and the degree of hemeroby. It was thereof possible to separate the different types of land use cover according to their degree of hemeroby and land use intensity, respectively. Moreover, since the concept of plant traits is a functional framework in which each trait can be assigned to one or more ecosystem functions, the assessment of STV is a promising step towards assessing the diversity of spectral traits in an ecosystem as a proxy of functional diversity.},
keywords = {GLCM, Hemeroby, Human-use-intensity, NDVI, RapidEye, Remote Sensing, Spectral trait variations, Spectral traits, Urban land-use-intensity},
pubstate = {published},
tppubtype = {article}
}
By adding attributes of space and time to the spectral traits (ST) concept we developed a completely new way of quantifying and assessing land use intensity and the hemeroby of urban landscapes. Calculating spectral traits variations (STV) from remote sensing data and regressing STV against hemeroby, we show how to estimate human land use intensity and the degree of hemeroby for large spatial areas with a dense temporal resolution for an urban case study. We found a linear statistical significant relationship (p=0.01) between the annual amplitude in spectral trait variations and the degree of hemeroby. It was thereof possible to separate the different types of land use cover according to their degree of hemeroby and land use intensity, respectively. Moreover, since the concept of plant traits is a functional framework in which each trait can be assigned to one or more ecosystem functions, the assessment of STV is a promising step towards assessing the diversity of spectral traits in an ecosystem as a proxy of functional diversity.