Urban Viewscapes

Developing viewscape model for urban landscape using LiDAR and Immersive Virtual Environments

Payam Tabrizian, Perver Baran, Helena Mitasova & Ross K. Meentemeyer

Background

Viewscape modeling
Binary viewsheds (calculated from point locations using line of sight calculations and digital elevation models DEM, DSM)
Landcover viewshed, configuration metrics, composition metrics
Widely used in ecosystem services, landscape assessment, planning, infrastructure impact assessment, etc.

viewshed concept shown in profile

viewshed map computed on digital elevation model

let me start with a brief background of what is Viewscape modeling ? Viewscape modeling can be refered to as analysis of the configuration, and composition of human 3D visual domain on the Landscape using geographic information systems. - The process usually starts with delineating the viewshed to estimate to estimate how much of the visible 3D surface one can see from a given vantage point ? The viewshed is then combined with landcover to estimate the composition of the visible surface ? Landscape structure analysis and spatial statisics can also show the configuration of the space such as is it wide or narrow vista ? how far one can see ? how diverse is the visible landscape and so forth ? These information is widely used as the basis to characterise landscape visual charactristics, model their visual quality and preservation value and cultural ecosystem values.

Background

Focus on continental, regional and landscape, limited application for urban envioronments.
Focus on objective assessment, limited correspondence with human perceptions.
LiDAR and Immersive Virtual Environments (IVEs)

Study area

Dorothea Dix park, Raleigh, NC, 308 acres

LiDAR data

Airborne LiDAR (North Carolina), Acquired Jan 11, 2015 (leaf-off)

Digital surface model

0.5m Digital surface model (DSM)

Digital surface model

0.5m Digital surface model (DSM)

Tree obstruction error

Real-world situation

Representation of trees in DSM

Vegetation structure

Evergreen

Evergreen + dense understory

Deciduous stands

Trunk obstrcution modeling

DSM

Trunk obstrcution modeling

Landform analysis (Geomorphon)

Trunk obstrcution modeling

Extracted tree peaks

Trunk obstrcution modeling

DSM after trunck replacement

Viewscape before obstruction modeling

Viewscape after obstruction modeling

Panoramic taken from viewscape point

High resolution Landcover

Trees derived from lidar points

Ground cover derived from supervised classification

Roads and buildings derived from official vector data

Viewscape metric	references

Configuration metrics
Relief	Tveit et al.(2009), Dramstad et al.(2006)
Visible Horizontal surface	Tveit et al.(2009), Stamps(2005)
Viewdepth variation	Sahraoui(2016)
Extent (viewshed size)	Tveit et al.(2009), Dramstad et al.(2006)
Shannon diversity index (SDI)	Sahraoui(2016), Sang(2008), Ode et al.(2008)
Mean Shape Index (MSI)	Stamps(2005), Sahraoui(2016)
Edge density (ED)	Sang(2008), Ode et al.(2008), Fry et al.(2012)
Number of patches (Nump)	Sang(2008), Ode et al.(2008), Fry et al.(2012)
Depth	Ode et al.(2008), Fry et al.(2012), Tveit et al.(2009)

Composition metrics
% Visible landcover	Ode et al.(2008), Sahraoui(2016)

Composition metrics

Herbacous               0 %
Mixed forest          12 %
Evergreen forest    0 %
Deciduous forest     0 %
Grass land             17 %
Paved roads           48 %
Buildings                 25%

Herbacous               0 %
Mixed forest            0 %
Evergreen forest     0 %
Deciduous forest    18 %
Grass land             38 %
Paved roads           27 %
Buildings                17 %

Herbacous               0 %
Mixed forest            5 %
Evergreen forest     4 %
Deciduous forest    38 %
Grass land              32 %
Paved roads           14 %
Buildings                  5 %

Herbacous             65 %
Mixed forest          12 %
Evergreen forest    8 %
Deciduous forest    15 %
Grass land               0 %
Paved roads            0 %
Buildings                 0 %

Configuration metrics


Extent	185000 m2	50710	1072
Relief	7.12 m	4.76	2.11
Vdepth	17.3	29.90	5.29
SDI	1.11	1.518	1.449
MSI	39.1	63.48	19.7
Nump	3700	4168	642
Depth	538 m	1293	305

IVE image acquisition method

Image aquisition Stiching and editing Cube mapping and wrapping

IVE survey

item	question
Perceived Visual access	How well can you see all parts of this setting without having your view blocked or interfered with?
Perceived Complexity	I perceive this environments as . . . Simple=0, Complex=10
Perceived Naturalness	I perceive this environment as … Not natural = 0 , Natural =10
Aesthetic Preference	I like this environment

Results

Response variable	R2 adjusted	Significant independent variable
Perceived Visual access	0.64	Extent ↑*, Depth↑, Relief↓*, Vdepth var↓, Building↓, Paved↑ , Deciduous↑, Building↓* , Nump↓***
Perceived Complexity	0.42	SDI * ↑, Relief ↑, Depth ↓, ED↑, Nump↑, Building↑
Perceived Naturalness	0.62	Relief ↑*, Deciduous ↑, Mixed↑*, herbaceous↑, Building↓, Nump↑
Aesthetic Preference	0.54	Relief ↑*, Extent ↑, Depth↑*, SDI↑, herbaceous↓, Deciduous↑ Buildings↓ Nump↓

Generalized linear models for four response variables. Best model fit was determined by step-wise regression.
↑ : Positive association
↓ : Nagative association
*p<0.05, ** p <0.01, *** p <0.001
Variables: Vdepth_var = viewdepth variation, Nump = patch number, ED = edge density, MSI= mean shape index, SDI= shannon diversity index.

Conclusion

Viewscapes modelled using LiDAR data can be potentially a usefull method for modeling landscape perceptions at site-scale and for complex environment
IVE perception can be used to generate perceptions maps for the site (regression coeffiecient)
Method can be used to create high-precision viewscape models for larger regions (HPC).