Thermal Signatures from GOES - The figure to the right is an example of the GOES-8 11-micrometer imager channel for May 6, 1997 at 10:30p EDT over the southeastern part of the United States. The calibrated radiance data has been converted to brightness temperature using the inverse Planck function. The image represents the emission of radiation from surface features and the attenuation and re-emission of energy from (primarily) atmosphere water vapor in this spectral band. Since the transmittance of the atmosphere is typically around 0.50-0.80 in this spectral region (Suggs et al. 1998), most of the signal comes from surface thermal emission (combination of land surface temperature and thermal emissivity) of the features over this region. Large scale thermal gradients are apparent in this image. The image has been enhanced and color-coded to bring out the thermal gradients in the image. An obvious warm region centered over the Atlanta metropolitan region is several degrees warmer than the surrounding area. This region of elevated temperatures in the GOES image corresponds to about a 1600 km2 area. Other thermal "hotspots" related to variations in land use are also apparent.

Retrieval of Land Surface Temperature - The GOES 11 micrometer image indicates the relative variations of the surface thermal structure but is also "contaminated" by atmospheric absorption (mainly water vapor) which limits its quantitative use. The Physical Split Window retrieval algorithm (or PSW for short) is used with the split-window channels of GOES (the 11 and 12 micrometer combination) to obtain the land surface temperature (LST) (the physical radiating temperature of the underlying surface) over the region. This procedure accounts for atmospheric absorption resulting from spatial variations in atmospheric temperature and moisture in order to retrieve a more precise estimate than using the 11 micrometer channel alone. For this study, the PSW algorithm is applied to single pixels of the GOES data with a time interpolated (between 1200 and 0000 UTC data from the closest rawinsonde station) atmospheric profile data as a guess for the retrieval algorithm. While this procedure is not perfect, because of constraints of the simplified radiative transfer equation (used in the retrieval procedure) and the effect of varying surface emissivity on the channel brightness temperatures which has been ignored (assumed to be 0.98 in each channel), it does produce much more accurate estimates of LST which can be used to evaluate temporal and spatial gradients in LST for a given region.

Diurnal Variability of LST - The figure at the left presents the area-averaged LST for the Spring of 1997 for the Atlanta urban region (solid red) and the corresponding rural region (solid blue) to the southeast of Atlanta (see US1 and RS1 in the Figure) at hourly intervals from May 4 at 10:45 UTC through May 5 at 1045 UTC. The urban region undergoes rapid surface heating throughout the morning hours consistent with the increase in insolation during this period of time. Maximum LST of around 299.0 K occurs between 1745 (1345 LDT) and 1845 UTC (1445 LDT) which is 1-2 hours after the peak of insolation. After this time, rapid cooling occurs until around sunset (0045 UTC) with the cooling rate decreases more gradually throughout the night. The minimum temperature of 276.3 K is reached at 1045 UTC on May 5. The amplitude of this diurnal cycle over the Atlanta urban region exceeds 25 K.
The diurnal cycle of LST over the Atlanta urban region is in contrast with a corresponding rural region (solid blue curve). The rural curve is very similar to the urban temperature curve for the first five times in the morning. The curves differ beginning at 1545 UTC when the surface heating of the rural region slows appreciably. A maximum LST value of 296.4 K for the rural region is reached at 1845 UTC when cooling begins to occur at fairly rapid pace until sunset, when more gradual cooling occurs throughout the night. A minimum of 274.9 K is reached at 1045 UTC on May 5. The difference in LST between these two regions for each time is indicated by the solid black line (click on image for better view). Despite both regions starting out with nearly the same LST, the minimum values 24 hours later are different with the rural region being cooler by nearly 1.0 K. The maximum LST difference between the urban and rural region occurs around the time of maximum temperature with a broad secondary maximum difference occurring during the night. Minimum differences in the urban and rural areas occur during the few hours after sunrise and just before sunset (2245 - 2345 UTC). This urban vs. rural contrast is known as the urban heat island effect. Traditionally calculated from contrasting air temperature differences between the two regions, the use of LST yields some interesting results. The greatest difference in the LST data around 1845 UTC and minimum just before sunset is in sharp contrast to our traditional measurement of the urban heat island from air temperature measurements which indicate a maximum difference around sunset. This is confirmed with the surface air temperature observations from reporting stations close to these regions of interest.
It is interesting to compare the rate of change of LST with time over these two regions (dashed lines in the figure above). Both regions show rapid heating with increasing solar angle for the first 4 hours with values exceeding 5K/hr at 1345 UTC After this time heating continues but at a slower rate until peak temperatures are reached around 1845 UTC. The heating rate for the rural region is suppressed from 1445 through 1845 UTC. It is presumed that the absorbed solar radiation incident on the rural region goes into latent heat of evaporation rather than sensible heat as in the urban region. Both curves also indicate substantial cooling throughout the late afternoon, evening and through the night. Maximum values nearly reach -4.0 K/hr around sunset. It is noteworthy that the urban cooling rate is significantly larger than for the rural region in the late afternoon approaching sunset (around 2345 UTC) and explains the minimum difference in LST between the urban and rural regions at this time. Both regions show similar cooling throughout the night until sunrise on the second day (May 5).

For an interesting look at the urban heat island effect using high resolution thermal imagery from aircraft scanners see the work of Luvall and Quattrochi.

Last updated on: November 2, 1999