Environmental Engineering Reference
In-Depth Information
than building count, it appears that building count is less sensitive
to building detection errors than is building area, and building
area is less sensitive than building volume. Since both building
area and building volume measurements heavily depend on lidar
measurements, they are more severely affected by detection and
measurement errors. This is the case especially for SF residential
buildings located in densely vegetated areas. Moreover, the mod-
els that distinguish between dwelling types (Models 2, 4, and 6)
tend to have larger errors than their counterpart that consider
residentialbuildingsaltogether(Models1,3,and5).Thisresult
suggests that the best strategy is to use the minimal amount of
information: residential building counts.
remote sensing-derived parameters, it was also apparent that the
most important parameters for population estimation, namely
residential subtype land uses and building volume, were also
the most difficult to accurately extract from remote sensing.
Based on results reported herein, future lidar-based population
estimation should focus on improving building detection meth-
ods first, particularly in reconstructing the 3-d building shape
more accurately, and then on improving land use classification
methods.
Acknowledgments
This study was supported by grants to LeWang from the National
Science Foundation (BCS-0822489, DEB-0810933), and from
National Key Basic Research and Development Program, China
(2006CB701304). The authors are thankful to three Co-PIs of the
NSF project (BCS-0822489) Dr. Changshan Wu at University of
Wisconsin at Milwaukee, Dr. Peter Rogerson at SUNY-Buffalo,
and Dr. Frederick Day at Texas State University-San Marcos.
Ms.TiantianFengandMr.BenjaminD.Kamphausarethanked
for their help on developing building detection and land use
classification.
Discussion and conclusions
Small-area population estimation is an important task that has
received considerable attention by the remote sensing commu-
nity in the past four decades. The wealth of related studies
reveals that the notion of living space had been considered a key
linkage between population and remote sensing measurements.
Unfortunately, a formal definition for this important variable
has proved difficult due in part to the relatively coarse spatial
resolution of the remote sensing data used for population esti-
mation. The advent of fine spatial resolution satellite images (1
to 5 m) coupled with lidar measurements opened new oppor-
tunities for considering the 3-d nature of living space in urban
environments and for improving small-area population estima-
tions. In the study reported here, we tested the potential of high
spatial resolution lidar measurements coupled with automated
and semiautomated techniques for building extraction and land
use classification. We compared seven linear models for small-
area population estimations, each of which is parameterized in
terms of one, two or three explanatory variables representing
building statistics on a per-block basis (count, area, and volume)
at one of two land use classification levels (residential or SF/MF).
These explanatory variables were meant to more closely repre-
sent the living space because the great majority of population
lives inside buildings. Interestingly, when considering other geo-
metric characteristics of building, such as perimeter, shape and
height, their contribution to the regression was not significant
(Model 7).
At the model fitting stage, it was observed that the incorpora-
tion of either the fine-level land use information or the volume
information led to higher correlation coefficients. At a validation
stage, however, the differential improvement achieved by includ-
ing building volume appeared not as important as that of using
fine land use information. Presumably, this was due to errors
introduced by the method used for calculating the height infor-
mation from lidar data. At the estimation stage, we first replaced
the reference building layer by the detected buildings and tested
the effect on the population estimation errors. The original trend
observed during the fitting state was totally inverted; suggesting
that the incorporation of finer land use or building volume (or
even building area) did not improve the population estima-
tions. The sensitivity of each model to the errors in the land
use information further favored the simplest model based on
counts of residential buildings. While the reason for this inver-
sion appeared to be the violation of the fundamental assumption
that the high-quality calibration sample was representative of the
References
Aplin, P., Atkinson, P. and Curran, P. (1999) Fine spatial resolu-
tion simulated satellite sensor imagery for land cover mapping
in the United Kingdom.
Remote Sensing of Environment
,
68
(3),
206-216.
Benn, H. (1995) TCRP synthesis of transit practice 10: Bus route
evaluation standards.
Transportation Research Board
,National
Research Council, Washington, DC.
Chen, K. (2002) An approach to linking remotely sensed data
and areal census data.
International Journal of Remote Sensing
23
, 37-48.
City of Austin (2009a) City of Austin demographics. Online.
Available http://www.ci.austin.tx.us/demographics/ (accessed
16 November 2010).
City of Austin (2009b) City of Austin GIS data sets. Online. Avail-
able ftp://ftp.ci.austin.tx.us/GIS-Data/Regional/coa_gis.html.
(Last accessed Jan. 16, 2011)
Congalton, R.G., 1991. A review of assessing the accuracy of
classifications of remotely sensed data, Remote Sensing of
Environment, 3735-3746.
Congalton, R., and K. Green, 1999. Assessing the Accuracy of
Remotely Sensed Data: Principles and Practices, CRC/Lewis
Press, Boca Raton, Florida, 137 p.
Cowen, D.J. and Jensen J.R. (1998) Extraction and modeling of
urban attributes using remote sensing technology. In
People
and Pixels: Linking Remote Sensing and Social Science
(ed. D.
Liverman,E.F.Moran,R.R.Rindfuss,andP.C.Stern),National
Academy Press, Washington DC.
Downs, A. (1992) Stuck in traffic: coping with peak-hour traffic
congestion. Brookings Institution, Washington, DC, Lincoln
Institute of Land Policy, Cambridge, MA.
Search WWH ::
Custom Search