38, 2012 
  
fier and 100 
(IQ). Mixed 
) data. 
d multisensoral 
‘the mean UHI 
ayer Perceptron 
ie models were 
sample of about 
in prior model 
mal (correlation 
).17/0.19 K for 
rk) and the 
17/0.19 K) data 
s (R: 0.27/0.26, 
parameters (R: 
tainly partly a 
predicted mean 
quality of the 
pect to the low 
tter models this 
%). This is not 
ses contributing 
fects are related 
temperature in 
lirectly to the 
ultispectral data 
| conditions and 
odel calibration. 
ral and height 
e same sensor. 
:d features with 
night indicate a 
le might not be 
are likely to be 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B8, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
  
Parameter set LR NN 
N| R MAE RMSE| R MAE RMSE 
shs 01.027 024 932 | 026. 0.26 0.33 
acp 2| 0.19 024 032 | 0.18 024 0.32 
tir 331.074 20.17: 022 | 081 019 023 
ndvi 331 0.66 0.19 025 | 070 0.21 02% 
ms 198] 0.75 017 022 \ 077 018 023 
R».25 711 0.71 3 0.13 023 | 073 0138 025 
all 2721 6.79 30.16 021] 0:32 .. 0.15 0.2 
  
  
Table 3. Results of the spatial empirical models (Linear 
Regression and Neuronal Network) of the mean UHI with 
different parameter/predictor sets. Correlation coefficient R, 
mean absolute error and root mean square error. 
Nevertheless, it can be stated, that multisensoral and 
multitemporal datasets have some potential for spatiotemporal 
modelling of the mean UHI. This underlines the results of 
Bechtel and Schmidt, who found strong correlations between 
Landsat data and a long-term mean UHI dataset derived from 
floristic proxy data (Bechtel and Schmidt, 2011). 
The performance of Linear Regression and Neuronal Network 
models was rather similar, which might be due to the chosen 
standard options for the neuronal network classifier (with only 
one hidden layer). First tests with more sophisticated networks 
showed better results (for instance R: 0.83, MAE: 0.14 K for tir 
with a 20|10|10 node network). 
5. CONCLUSIONS 
The presented results from Hamburg indicate that multisensoral 
and multitemporal data has potential for both, the classification 
of Local Climate Zones and the empirical modelling of the 
spatial distribution of the UHI. 
The classification results show that the data (especially 
multitemporal thermal and multitemporal spectral data) are 
functional for the purpose and that micro-climatic meaningful 
urban structures can be classified from different remote sensing 
datasets. Further, it provides some evidence for the relevance of 
the Local Climate Zone system from a remote sensing point of 
view. 
The empirical modelling results also underpin the urban 
climatologic relevance of the multitemporal tir und ms data. 
Although a certain correlation is obvious, since vegetation and 
surface energy balance play important roles in the distinction of 
urban climates, these good results with freely available Landsat 
data offer the prospect of a wide application. However, further 
investigations are needed and the large number and complexity 
of the involved processes limits the potential of empirical 
models. The incorporation of data from other sensors also 
slightly improved the empirical modelling results. 
6. REFERENCES 
Arnfield, A. J., 2003. Two decades of urban climate research: a 
review of turbulence, exchanges of energy and water, and the 
urban heat island. International Journal of Climatology, 23(1), 
pp. 1-26. 
Barsi, J. A., Schott, J. R., Palluconi, F. D., and Hook, S. J., 
2005. Validation of a Web-Based atmospheric correction tool 
for single thermal band instruments. Earth Observing Systems 
X, Proc. SPIE, 58820F, pp. 1-7. 
Bechtel, B., 2011. Multitemporal Landsat data for urban heat 
island assessment and classification of local climate zones. 
Urban Remote Sensing Event, JURSE, 2011 Joint, pp. 129-132. 
doi:10.1109/JURSE.2011.5764736. 
Bechtel, B., 2012. Robustness of Annual Cycle Parameters to 
characterize the Urban Thermal Landscapes. IEEE Geoscience 
and Remote Sensing Letters, doi 10.1109/LGRS.2012.2185034. 
Bechtel, B. and Daneke, C., 2012, Classification of local 
Climate Zones based on multiple Earth Observation Data. IEEE 
Journal of Selected Topics in Applied Earth Observations and 
Remote Sensing. doi 10.1109/JSTARS.2012.2189873. 
Bechtel, B., Langkamp, T., Ament, F., Bohner, J., Daneke, C., 
Gunzkofer, R., Leitl, B. et al, 2011. Towards an urban 
roughness parameterisation using interferometric SAR data 
taking the Metropolitan Region of Hamburg as an example. 
Meteorologische Zeitschrift, 20(1), pp. 29—37. 
Bechtel, B., and Schmidt, K., 2011. Floristic mapping data as a 
proxy for the mean urban heat island. Climate Research, 49, pp. 
45-58. doi:10.3354/cr01009 
Bouckaert, R. R., Frank, E., Hall, M., Kirkby, R., Reutemann, 
P. Seewald, A., and Scuse, D., 2009. WEKA manual for 
version 3-7-0. University of Waikato, Hamilton, New Zealand. 
Disponible en: http://ufpr. dl. sourceforge. 
net/project/weka/documentation/3.7. x/WekaManual-3-7-0. pdf. 
Burges, C. J. C., 1998. A tutorial on support vector machines 
for pattern recognition. Data mining and knowledge discovery, 
2(2), pp. 121-167. 
Chander, G., Markham, B. L., and Helder, D. L., 2009. 
Summary of current radiometric calibration coefficients of 
Landsat MSS, TM, ETM+ and EO-1 ALI Sensors. Remote 
Sensing of Environment, 113, pp. 893-903. 
Eliasson, I, 1992. Infrared thermography and urban 
temperature patterns. International Journal of Remote Sensing, 
13, pp. 869-879. 
Fabrizi, R., De Santis, A., and Gomez, A., 2011. Satellite and 
ground-based sensors for the Urban Heat Island analysis in the 
city of Madrid. Urban Remote Sensing Event, JURSE, 2011 
Joint. doi:10.1109/JURSE.2011.5764791 
Frey, C., and Parlow, E., 2009. Geometry effect on the 
estimation of band reflectance in an urban area. Theoretical and 
Applied Climatology, 96(3), pp. 395—406. 
Frey, C., Rigo, G., and Parlow, E., 2007. Urban radiation 
balance of two coastal cities in a hot and dry environment. 
International Journal of Remote Sensing, 28(12), pp. 2695— 
2712.