Potter T.D., Colman B.R. (co-chief editors). The handbook of weather, climate, and water: dynamics, climate physical meteorology, weather systems, and measurements

Подождите немного. Документ загружается.

6 SATELLITE REMOTE SENSING

Meteorologists use two basic methods of observing the atmosphere: in situ and

remote-sensing methods. In situ methods, for ‘‘in place,’’ measure the properties

of the air in contact with an instrument. Remote-sensing methods obtain information

without coming into physical contact with the region of the atmosphere being

measured. Remote sensing the atmosphere is the emphasis of this section. In

remote sensing, radiation measurements are used to infer the state of the atmosphere.

The inference of the atmospheric state is often referred to as the retrieval process (see

page 39 for discussion of retrieval process).

There are two basic types of remote sensing the atmosphere: active sensors and

passive sensors. Active remote-sensing instruments emit energy, such as a radio

wave or beam of light, into the atmosphere and then measure the energy that returns.

Passive remote-sensing instruments measure radiation emitted by the atmosphere,

Earth’s surface, or the sun. Much of what we observe about the atmosphere using

remote-sensing techniques results from how radiation interacts with molecules or

objects, such as water drops, suspended in the atmosphere. The previous chapter

discussed the principles on which remote-sensing methods are based, while this

section looks at some applications of passive remote sensing from satellites.

Remote Sensing the Surface

In remote sensing the surface of Earth from a satellite, we select spectral regions, or

channels, in which the atmosphere is transparent. These are called atmospheric

windows. In the solar spectrum, or the shortwave, very little absorption occurs in

the visible region; however, Rayleigh scattering is large. Rayleigh scattering is well

understood and can be handled via modeling as described in the previous

chapter. There are also windows in the near-infrared spectral regions where Rayleigh

scattering is smaller.

Analysis of the spectral reﬂectance of different surfaces generally shows a distinct

difference between the visible and near-IR regions. Observations in both the visible

(such as 0.58 to 0.68 mm) and near-infrared (such as 0.725 to 1.1 mm) are useful for

monitoring surface conditions. Vegetation regions generally have reﬂectances in the

near-infrared (NIR) that range from 20 to 40%, while visible reﬂectances generally

range from 5 to 15%. Soils also have a higher reﬂectivity in the NIR than in the

visible while the opposite is true for snow.

A common method of monitoring surface vegetation is through the normalized

difference vegetation index (NDVI):

NDVI ¼

NIR

 R

VIS

NIR

 R

VIS

ð28Þ

This has long been used to monitor the vegetation, and changes in vegetation, of

the entire Earth. NDVI for vegetation generally range from 0.3 to 0.8, with the larger

values representing ‘‘greener’’ surfaces. Bare soils range from about 0.2 to 0.3.

376 RADIATION IN THE ATMOSPHERE: OBSERVATIONS AND APPLICATIONS

Identifying snow cover is important for weather and hydrological forecasting. To

detect the presence of snow, recent satellite instruments include observations at 0.66

and 1.6 mm (Fig. 30). The atmosphere is transparent at both these wavelengths, while

snow is very reﬂective at 0.66 mm and not reﬂective at 1.6 mm. The normalized

difference snow index (NDSI),

NDSI ¼

0:66

 R

1:6

0:66

þ R

1:6

ð29Þ

is used to monitor the extent of snow cover. At visible wavelengths (e.g., 0.66 mm),

snow cover is just as bright as clouds and is therefore difﬁcult to distinguish from

cloud cover. However, at 1.6 mm, snow cover absorbs sunlight and therefore appears

much darker than clouds. This allows the effective discrimination between snow

cover and clouds. Values of NDSI <0.4 typically indicate the presence of snow.

Figure 30 demonstrates the ability to separate clouds from snow using observations

at these wavelengths.

Sea surface temperature (SST) is another surface property we are interested in

from a meteorological perspective. An IR window in the atmosphere for sensing

Figure 30 (see color insert) False color image from a combination of observations at 0.66,

1.64, and 2.14 mm are used to detect snow. In this false color images, land surfaces are green,

water surfaces are black, snow cover is red, and clouds are white. See ftp site for color image.

6 SATELLITE REMOTE SENSING 377

surface temperature is the 10- to 12-mm region where absorption by water vapor is

weak. Figure 31 is a MODIS (moderate resolution imaging spectrometer) 11-mm

image that clearly indicates changes of SST in the vicinity of the Gulf Stream. Most

of the radiation in this band is emitted by the surface and transmitted through the

atmosphere. In a warm moist atmospher e the difference between the SST and the

brightness temperature at 11 mm (BT

) can approach 10



C. This difference is often

Figure 31 (see color insert) An 11-mm image from MODIS of Atlantic Ocean off east coast

of North America. See ftp site for color image.

378

RADIATION IN THE ATMOSPHERE: OBSERVATIONS AND APPLICATIONS

corrected for by making observations at more than one wavelength, such as 11, 12,

3.7, and 8.5 mm. Differences between these channels represent the total amount of

water vapor in the column. For example, the 12-mm channel has more absorption and

therefore (BT

7 BT

) is positive; the greater this difference the larger the water

vapor loading of the atmosphere. Observations at these wavelengths are used daily to

derive SST. The SST from satellite observations is typically determined from

a regression derived empirically using observations from drifting buoys.

Remote Sensing of Clouds

Clouds are generally characterized by higher reﬂectance and lower temperature than

the underlying Earth surface. As such, simple visible and infrared window threshold

approaches offer considerable skill in cloud detection. However, there are many

surface conditions when this characterization of clouds is inappropriate, most nota-

bly over snow and ice. Additionally, some cloud types such as thin cirrus, low stratus

at night, and small cumulus are difﬁcult to detect because of insufﬁcient contrast

with the surface radiance. Cloud edges cause further difﬁculty since the instrument

ﬁeld of view will not always be completely cloudy or clear. There are many

different methods of detecting clouds. In this section we review some of the more

common approaches.

The simplest cloud measurement technique is the threshold method in which an

equivalent blackbody temperature or a spectral reﬂectance threshold is selected that

distinguishes between cloud and noncloud in infrared or visible satellite images.

Information on cloud top temperature is obtained by comparing the observed bright-

ness temperature with an atmospheric temperature proﬁle—this approach usually

underestimates the cloud height. Using a visible or near-infrared reﬂectance thresh-

old works well for determining clear-sky ocean scenes that are free of sun glint.

Another straightforward approach employs two channels in combination. For

example, the split window technique makes use of observations near 11 and

12 mm to detect clouds over oceans. Cloud classiﬁcation is accomplished by

considering the 11-mm blackbody temperature and the difference between 11 and

12 mm. Clear scenes have warm temperatures and brightness temperature differences

that are negative, usually less than about 1



. Another simple two-channel tech-

nique uses visible and infrared observations. In this approach observed visible

reﬂectance and equivalent blackbody temperature are compiled, and observations

are then classiﬁed based on their relative brightness and temperature. For example,

clear sky oceans would be warm and dark while convective clouds would be cold

and bright. Classiﬁcation of the cloud is accomplished by either assigning thresholds

or by employing maximum-likelihood statistical techniques.

Once clear pixels can be determined, these obser vations can be combined with

cloudy observations to derive cloud properties. One such approach is the CO

slicing

technique. In this technique a cloud pressure function G(l

, l

, p) is deﬁned as an

expression involving a pair of differences of two column radiances and [I(l

clr

) and

I(l

cld

)], one clear and one cloud contaminated with a cloud at pressu re level p

.If

you express the observed column radiances in terms of differences, then the implied

6 SATELLITE REMOTE SENSING 379

vertical integration need only go from the surface to the cloud top because the above

cloud components subtract; then by taking the ratio of two observed radiances, you

can remove the coefﬁcient of cloud amount that results from expanding these terms

into clear and cloudy portions. It is necessary to make an assumption that the cloud

is inﬁnitesimally thin, so that you need only work with a single transmittance func-

tion below the cloud. This method enables us to assign a quantitative cloud top

pressure to a given cloud element using observed radiances for the CO

spectral

bands. Deﬁning the radiance from a partly cloudy scene as

¼ FI

cld

þð1  FÞI

clr

ð30Þ

where F is the fractional cloud cover. The cloud radiance is given by

cld

¼ e

bcld

þð1  e

ÞI

clr

ð31Þ

where e is the cloud emissivity, and I

bcld

is the radiance from an opaque cloud. Thus

clr

¼ B

ðT

Þtlðp

Þþ

ðT

Þdt

ð32Þ

bcld

¼ B

ðT

Þtlðp

Þþ

ðT

Þdt

ð33Þ

where p

is the cloud top pressure. Integrating by parts (e.g.,

vvdx¼ uv 

uudx)

and subtracting the two terms yields

clr

 I

bcld

ðpÞdB

ðT

Þð34Þ

and

 I

clr

¼ Fe

ðpÞdB

ðT

Þð35Þ

If two wavelengths are chosen that are close to one another, then e

 e

, which

leads to

 I

clr

 I

clr

ðpÞdB

ðT

ðpÞdB

ðT

ð36Þ

In practice, the left-hand side is determined from observations, and the right-hand

side is calculated from a known temperature proﬁle and the proﬁles of atmospheric

transmittance for the spectral channels as a function of p

. The optimum cloud top

380 RADIATION IN THE ATMOSPHERE: OBSERVATIONS AND APPLICATIONS

pressure is determined when the difference between the right-hand and left-hand

sides of the equation are a minimum.

Once the cloud height has been determined, the effective cloud amount Z ¼Fe is

determined from a window channel observation.

Fe ¼

 I

clr

ðT

ÞI

clr

ð37Þ

The CO

slicing approach is good for detecting clouds in the upper troposphere

but fails for clouds in the lower troposphere.

Remote-Sensing Atmospheric Temperature Proﬁles

The retrieval of the atmospheric temperature and moisture proﬁle is often accom-

plished using spectral observations in the infrared. The appropriate equation for the

transfer of infrared radiation is

¼ B

ðT

sfc

ÞT

ð0Þþ

z¼0

ðTÞ

ðzÞ

dz ð38Þ

where I

is the observed radiance at wavelength l, T

sfc

is the surface temperature, T

is the transmitta nce, and B

(T) is the Planck function containing information on the

atmospheric temperature. The term

ðzÞ

¼ W ðzÞð39Þ

is referred to as the weighting function. The intensity measured by a satellite radio-

meter due to the emission from a layer in the atmosphere at location z, is determined

from the layer blackbody emission B

(T) weighted by the factor W(z). The weighting

function is of fundamenta l importance to vertical sounding the atmosphere from

satellite obser vations. The weighting distribution depends on the strength and distri-

bution of the absorbing gas.

To retrieve atmospheric temperature proﬁles, satellite radiometers make measure-

ments in the carbon dioxide absorption bands because carbon dioxide is relatively

uniformly mixed in the atmosphere, and thus the vertical distribution is known.

Observations are made at spectral regions across the carbon dioxide absorption

band, including weak and strong absorption regions. Figure 32 shows the weighting

functions in the 12- to 15-mm spectral region of the Geostat ionary Orbiting Environ-

mental Satellite (GOES) sounder radiometer. The 14.7-mm region is a strong absorp-

tion region of carbon dioxide and so the weighting function peaks in the stratosphere.

The weighting function at 14.7 mm is near zero for pressures greater than 500 mb,

indicating that radiance observations at this wavelength receive no contribution

from the lower atmosphere and surface. The 13.4-mm spectral region is a weak

absorption region and weighting function peaks at the surface, with only small

6 SATELLITE REMOTE SENSING 381

contributions in the stratosphere. The location of peak in the weighting function

indicates the region of the atm osphere that makes the largest contribution to the

radiance being measured. The width of the weighting function characterizes the

vertical resolution of the retrieval.

Given different spectral observations, I

, we wish to solve for the temperature

proﬁle, given the distribution of absorbing gas. This solution is not straightforward

as the equation is nonlinear and the problem is underconstrained so that no unique

solution exists. To simplify the problem we consider the discrete form of the

radiative transfer equation

¼ðB

ðz ¼ 0Þþ

j¼1

i;j

i;j1

 T

i;j

ð40Þ

where i represents a spectral channel and j is the atmospheric level. Or

¼ðB

iri

ðz ¼ 0Þþ

j¼1

i;j

ð41Þ

Figure 32 GOES sounder channels used in retrieval of temperature proﬁles.

382

RADIATION IN THE ATMOSPHERE: OBSERVATIONS AND APPLICATIONS

The vector I

represents our spectral channel observations, K

is the discrete weight-

ing function elements and includes the surface term, and the unknowns are the vector

i,j

. Our problem is to invert this equation to solve for B

i,j

from which the tempera-

ture proﬁle follows. The problem is underconstrained: Given I

we cannot say

anything about the individual values of B

i,j

. We require a priori constraints, for

example, by reducing the number of layers over which the proﬁle is speciﬁed or

by specifying the representation of B

i,j

. The numerical methods used to retrieve

temperature from the radiance measurements are described elsewhere. Retrievals

often start with a ﬁrst-guess proﬁle, for example, from in situ radiosonde observa-

tions or a numerical forecast model. The ﬁrst-guess proﬁles are then adjusted until

calculated radiances match the observed radiances within some threshold. Because

the observed radiances arise from deep and overlapping layers, as indicated in

the weighting function, the retrieved temperature proﬁles a re for layers of the

atmosphere and do not resolve the sharp temperature gradients sometimes observed

in radiosonde measurements. However, the satellite provides better spatial and

temporal resolution and can be very useful for forecasting.

An example is monitoring the convective conditions of the atmosphere using the

lifted index. The lifted index is the temperature difference found by subtracting the

temperature of a parcel of air lifted from the surface to 500 mb from the existing

Figure 33 (see color insert) Sequence of GOES-derived lifted index on May 3, 1999. Red

regions indicate the potential areas of thunderstorm development. Strong tornadoes developed

in southwest Oklahoma before 22 UTC. The lifted index versus color code is given in the

legend. See ftp site for color image.

6 SATELLITE REMOTE SENSING 383

temperature at 500 mb. The lifted index numbers are related to thunderstorm

severity. Values less than or equal to 6 indicates conditions are very favorable

for development of thunderstorms with a high likelihood that if they occur, they

would be severe with high winds and hail. The satellite-observed lifted index on May

3, 1999, over the south central United States is shown in Figure 33. The time

sequence of the satellite-derived lifted index clearly shows (see the red region) the

region favorable for the development of severe thunderstorms.

Remote-Sensing Atmospheric Moisture

Once the temperature proﬁle of the atmosphere has been determined, infrared obser-

vations in water vapor absorption bands can be used to infer atmospheric water

content. GOES observations at water vapor absorption bands are routinely used

by the National Oceanic and Atmospheric Administration (NOAA) to derive the

vertically integrated water vapor, or precipitable water (PW). An example is shown

Figure 34 (see color insert) GOES-8-derived precipitable water on April 20, 1999. Blue

regions indicate regions of low PW and green relative high amounts of PW. The satellite

analysis clearly shows dry region south of Great Lakes. See ftp site for color image.

384

RADIATION IN THE ATMOSPHERE: OBSERVATIONS AND APPLICATIONS

in Figure 34. The GOES sequence of observations captures the large region of

relatively dry air with less than 20 mm of PW (blue enhancement) in the western

Great Lakes region on August 20, 1999. During the late morning and afternoon

hours, the GOES-derived PW shows an elongated moist region with PW greater than

20 mm (green enhancement) across Illinois.

This moisture distribution impacted the cloud for mation on this day. Cumulus

formation was suppressed on either side of this moist plume, as seen in the GOES

visible image. The GOES-derived PW remained over these cloud-free areas during

the entire day (see Fig. 35).

Figure 35 GOES-8-visible image on April 20, 1999 at 1715 UTC. Cumulus clouds are

absent in the regions of low PW. See ftp site for color image.

6 SATELLITE REMOTE SENSING 385