520
2 DETERMINATION OF PHYSICAL SURFACE TEMPERATURE
2.1 NOAA and Meteosat
A first approach uses brightness temperatures derived from thermal infrared channels erf
Meteosat (cf. Hiltbrunner, 1991) and NOAA-AVHRR (Kidwell et al., 1986) as an estimation of the physical
temperature of the land surface. Fig. la shows the correlation between the geolocated infrared brightness
temperatures derived from NOAA channel 4 and ground-measurements of grass-temperature provided by the
Swiss Operational Meterological Network (ANETZ) for May 3,4:30 UTC. The infrared brightness temperature
scenes where visually classified as cloud-free but not subject to any atmospheric correction or cloud separation.
Differences between estimated and measured temperatures up to 14°C occur. In the case of Meteosat (Fig lb),
even differences of 18°C are observed. This may be attributed to thin cirrus clouds within certain Meteosat
pixels or to the fact that the Meteosat pixel with a size of about 6 x 8 km cover a wide variety of different
altitude levels and therefore different temperature values. A transformation of the temperature to potential
temperatures showed no significant improvement
ANETZ Tgroee [C]
ANETZ Tgroee [C]
Fig. 1 (a, left): Comparison between grass-temperature measurements from 70 ANETZ stations and NOAA
Channel 4 brightness temperatures for May 3, 1989, 13:00 UTC. (b, right): Same for Meteosat thermal
infrared channel, 4:30 UTC. Temperatures are in degrees Celsius.
For our purposes, the observed differences between measured and estimated temperatures are
too high. While Cooper & Asrar (1989) reported that only the McClain et al. (1983) split window model gave
satisfying atmospheric corrections of the NOAA temperatures, Sugita and Brutsaert (1993) concluded that an
atmospheric corrections of the NOAA thermal infrared data using split window or linearized profile method
did not result in a significant improvement of the temperature estimation. Therefore, no atmospheric correction
was applied.
22 Interpolation of in-situ measurements
A second approach consists in the spatial interpolation of ground-based measurements of
physical air-temperatures provided by ANETZ using the so-called Kriging interpolation. Kriging is a statistical
interpolation method which takes the anisotropic autocorrelation of the spatial distribution of a geophysical
phenomenon like the surface temperature into account (Ishida & Kawashima, 1993; Lam, 1983). We applied
this method to 25 ANETZ-stations located at an altitude between 316 and 779 m.a.s. distributed over the
Central Plains, which are located between the Jura mountains and the Swiss Alps (Fig. 2). The resulting
temperature field is strongly dependent on the local temperature values of the meteorological stations used as
tie-points. This method may be appropriate for large-scaled studies of the temperature-field, where the
variations of the temperature within some kilometers are less important For our purposes, the interpolated
temperature does not sufficiently reflect the true variability of the different surface types.
