Tuesday, June 14, 2016

Value of urban views in a bay city: Hedonic analysis with the spatial multilevel additive regression (SMAR) model

Highlights
• Non-linear effects of open view, ocean view, and green view, are estimated.
• The spatial multilevel additive model is applied for a hedonic analysis.
• Not only poor green view, but also excessive green view has a negative impact.
• How to evaluate 3D view to trees with remotely sensed data is demonstrated.

Abstract
This paper attempts to assess the value of urban views in a bay city (Yokohama), Japan. Firstly, three types of views, open view (goodness of visibility), green view (visibility of open space), and ocean view (visibility of ocean), were quantified employing the viewshed analysis implemented on the GIS with airborne LiDAR data and 0.5 m × 0.5 m high resolution aerial photos. Secondly, hedonic analyses were conducted to test the capitalization of value of those views into condominium prices using the spatial multilevel additive regression (SMAR) model, where possible non-linearity, multilevel structure of condominiums (unit-building), and spatial dependence were considered. This study implies that “very nice” open view (in terms of the amount of visibility) and ocean view may have a positive premium, whereas “slightly nice” open and ocean views may not. Also, a “moderate amount” of green view may raise condominium prices, but “poor” and “too much” green view may reduce condominium prices. These results indicate that the effects of views are indeed non-linear, and therefore it may be misleading to interpret the results obtained by linear models as existing studies have done....


File:Minato Mirai In Blue.jpg
https://en.wikipedia.org/wiki/File:Minato_Mirai_In_Blue.jpg
Our viewshed analysis was conducted by employing the DSM and DTM shown in Fig. 2 and Fig. 3, respectively. The DSM describes the height of the surface defined as the sum of the height of the ground and the height of the objects on it, and the DTM describes the height of the ground surface only. They are created from airborne LiDAR data through GIS data processing. Their spatial resolution (mesh block size) is approximately 0.5 m × 0.5 m. Using these data, we applied the viewshed analysis to evaluate open view, green view, and ocean view. The latter two were evaluated by counting the number of mesh blocks of green and ocean which were visible from each unit. Open view was evaluated by counting the number of mesh blocks of visible DSM from each unit.
...
Table 4. Parameter estimation results.


MR

SMR

SMAR


Coef.t-value
Coef.t-value
Coef.t-value
Intercept3.8020021.51***2.7920012.61***3.1900015.85***
Area1.12100421.19***1.12000421.26***1.11800422.86***
Open view0.025817.03***0.026157.13***


Green view0.003200.90
0.002970.84



Ocean view0.002234.89***0.002264.96***


Floor0.0097144.95***0.0097044.87***


Num. dev.0.012551.05
0.005820.51
0.006670.60
Major dev.0.047643.99***0.042913.77***0.049824.44***
SRC−0.02254−1.52
−0.02365−1.68.−0.02262−1.63
WRC0.031080.76
0.035470.92
0.035580.94
Station0.008091.25
−0.00378−0.58
−0.00073−0.12
Central distance0.008450.85
−0.06025−2.19*−0.06400−2.96**
Green−0.01583−1.18
0.064303.67***0.063383.95***
Park−0.00846−1.83.−0.00625−1.35
−0.00897−2.00*
Ocean−0.02005−2.97**0.013200.95
0.012771.05
C1 res.−0.03670−1.93.−0.01818−1.00
−0.02054−1.16
C1 high0.044372.07*0.040971.97*0.029941.48
Semi Ind.−0.05483−2.58**−0.05243−2.58**−0.04733−2.37*
The estimates of the view variables (“Open view”; “Green view”; and “Ocean view”) in the MR and SMR models are rather similar. The results indicate that “Open view” and “Ocean view” are positively significant at the 0.1% level whereas “Green view” is also positive but does not have a statistically significant influence. While these estimates are interpretable, the non-linear effects in the SMAR model, plotted in Fig. 11 (x-axis: values of regressors: zq,i-j; y-axis: the estimated non-linear effects of each regressor: the estimates of f(zq,i-j)), are highly nonlinear. In other words, there is a possible danger in interpretation of these linear estimates
...
Fig. 11


Fig. 11 indicates that the effect of “Open view” is non-linear, that is, f(Open view) increases non-linearly with the increase of the value of “Open view”. We can see that if the value of “Open view” is less than around 12, f(Open view) is constantly negative, but after around 12, the effect of f(Open view) becomes positive and it increases rapidly with the increase of the value of “Open view”. It means that very nice view (in terms of the amount of visibility) may be capitalized into condominium prices, but slightly nice view may not have any positive impacts. Such information could be useful for condominium developers and/or urban designers. With regard to “Ocean view”, its effect is fairly linear except for the low-value regions where ocean is not visible. Hence the results suggest that “Ocean view”, if the ocean is visible, may linearly increase the value of condominiums.

With regard to the coefficient estimates of variables other than view, the estimates for “Area”, “Floor” and “Major dev.” are positive and statistically significant at the 0.1% level and “Semi-Ind”, “HH pop.”, “Ind. 2 ratio” are negative and statistically significant at the 0.1% level, whereas the levels of significance of the other variables are fairly different among the models. The positive sign of “Floor” demonstrates that some factors other than visibility inflate the prices of higher floors. These factors would include fresh air (Kei, Wing, Yung, & Chung, 2006), better ventilation, much sunlight, better security, and lower humidity, which is important in Japan. Furthermore, high-rise units are often penthouse units, which have a positive premium (Conroy, Narwold, & Sandy, 2013). Thus, the positive sign of “Floor” is reasonable.
...
Some earlier studies measured visibility using a dummy variable, which takes one if a focused object is visible and zero if it is not visible (e.g., Benson, Hansen, Schwartz, & Smersh, 1998; McLeod, 1984). Other studies evaluated visibility based on field investigations. For instance, Tyrvainen and Miettinen (2000) used a field investigation to obtain the visibility information from the window of a unit of a condominium, whereas Luttik (2000) extracted such information from maps, complemented by field investigations....
...
Recent studies have employed more sophisticated visibility evaluation approaches (Yamagata et al., 2015): the isovist analysis, which has been developed mainly in architectural and urban studies (Benedikt, 1979), and the viewshed analysis, which has been developed mainly in landscape studies (Lynch, 1976). Isovist is defined as “the set of all points visible from a given vantage point in space and with respect to an environment” (Benedikt, 1979). A conventional isovist analysis evaluates views in a two dimensional (2D) space, and therefore one of the limitations of the conventional isovist analysis is ignorance of the third dimension (i.e., height)—capturing only a 2D horizontal slice of human perception (Yu, Han, & Chai, 2007; Yang, Putra, & Li, 2007). Note that, recently, some extensions that cope with this problem have been proposed (e.g., Bhatia, Chalup, & Ostwald, 2013; Morello & Ratti, 2009). On the other hand, the viewshed analysis tries to quantify a three dimensional (3D) view by examining whether each cell in a 3D raster is visible or not from an observation point.
...
by Yoshiki Yamagata 1, Daisuke Murakami 1, Takahiro Yoshida 1, Hajime Seya 2, Sho Kuroda 1
1. Center for Global Environmental Research, National Institute for Environmental Studies, 16-2, Onogawa, Tsukuba, Ibaraki Pref. 305-8506, Japan
2. Graduate School for International Development and Cooperation, Hiroshima University, 1-5-1, Kagamiyama, Higashi-hiroshima, Hiroshima Pref. 739-8529, Japan
Landscape and Urban Planning via Elsevier Science Direct www.ScienceDirect.com
Volume 151; July, 2016; Pages 89–102
Keywords: View; Viewshed analysis; Hedonic analysis; Spatial multilevel additive model; LiDAR data

No comments:

Post a Comment