2 nd E wiem Nimdie International Summer School. Satellite remote sensing applications in Meteorology. Weather and Climate Forecasting in Africa and its Application to Agriculture & Water Resource Management. 19 July – 31 July 2010 Kumasi , Ghana.
Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.
Weather and Climate Forecasting in Africa and its Application to Agriculture & Water Resource Management.
19 July – 31 July 2010
GizawMengistu, Dept. of Physics, Addis Ababa University, Ethiopia
2. Retrieval of meteorological parameters
3. Measurement of rainfall
4. Measurement of winds
5. Satellite image/signal
Observations depend on
Radiance is the amount of energy/per unit time/per area of a detector/per spectral interval/per solid angle
The radiance at the sensor is given by
LS j =[εjLjBB(T)+(1-εj)Ljsky]*t+Ljatm
Where εj is surface emissivity ,
LjBB is spectral radiance of a blackbody at the surface at temperature T,
Ljsky is spectral radiance incident upon the surface
from the Atmosphere, calculated using radiative transfer equation (e.g. MODTRAN etc),
Ljatm is spectral radiance emitted by the atmosphere,
again from Model,
t is spectral atmospheric transmission and
LS j is spectral radiance observed by the sensor.
After getting all the necessary data from RT model
as stated on the previous slide, radiance from the
Surface, Lj, is:
Water is a good approximation of a black body (grey body)
Quartz is not a good approximation of a black body (selective radiator)
Relative Emissivity (to the average of all channels, say
5 channels in this example) is given by
From laboratory measurement, relationship between minimum emissivity, max.-min. relative emissivity difference can be constructed i.e εmin=f(βmax- βmin). Therefore, the revised emissivity can be computed from
εj = βj(εmin/ βmin)
where the kernel is a simple exponential
and compare with step 1
3. Use direct linear inversion method
((AT.A)^-1AT.g) to recover f(x)
4. If result in (3) is not good, use constrained
inversion ((AT.A+gamma.H)^-1AT.g) where H
is constraining matrix (smoothing matrix)
Mengistu et. al., doi:10.1029/2004JD004856, 2004, doi:10.1029/2004JD005322, 2005
The choice of polar-orbiting vs geostationary platforms for precipitation estimation entails several tradeoffs with regard to temporal and spatial sampling and geographical coverage: a geostationary satellite positioned over the equator can provide high frequency (hourly or better) images of a portion of the tropics and middle latitudes, while a polar orbiter provides roughly twice-daily coverage of the entire globe. Polar orbiters also fly in a low Earth orbit which is more suitable for the deployment of microwave imagers on account of the latter's coarse angular resolution.
The choice of spectral band for observing precipitation also involves tradeoffs. Historically, infrared (IR) and visible (VIS) imagery have been widely available for the longest period of time, with high quality microwave (MW) imagery becoming widely available only after the launch of the SSM/I in 1987.
Advantages of the VIS and IR bands include high spatial resolution as well as the possibility of frequent temporal sampling from geostationary platforms. A major disadvantage is the indirectness of the relationship between cloud top albedo or temperature and surface precipitation rate.
The evidence to date suggests that VIS/IR methods produce highly smoothed depictions of instantaneous rainfall fields which become useful only when averaged over larger space and/or time scales, and then only when carefully calibrated for the region and season in question.
Almost all IR techniques are based on variations of the premise that precipitation is most likely to be associated with deep clouds and thus with cold cloud tops, as observed by an infrared imager. Visible cloud albedos are generally used, if at all, as supplemental information to discriminate cold clouds which are optically thin and presumably non-precipitating from those which are optically thick and therefore possibly precipitating. Of course, visible imagery is only usable during the time that the sun is high above the horizon.
IR-only methods are often preferred for the simple reason that their performance is less likely to be a strong function of the time of day and therefore less likely to introduce spurious day-night biases in estimated precipitation.
Because rainfall usually occupies only a small fraction of the cold or bright cloud area visible from space, VIS/IR algorithms tend to overestimate significantly actual rain area. To avoid systematic overestimates in temporal or spatial averages, this tendency is usually accounted for by assigning very low rain rates (empirically derived) to the area identified as precipitating in the instantaneous images.
The GOES Precipitation Index (GPI) is one of the simplest and most widely used IR indices of precipitation in the tropics and subtropics. The GPI is computed by simply taking the fraction of pixels within a region whose IR brightness temperatures are less than some threshold, T0, and multiplying that fraction by a constant rain rate, RQ. Most commonly, T0 is taken to be 235°K and RQ is taken to be 3.0 mm/hr. However, other values for these parameters may be more appropriate in some cases, depending on location, season, spatial averaging scale and other factors.
The Negri-Adler-Wetzel Technique (NAWT) technique
NAWT assigns rain rates to "cloudy pixels" based on a threshold brightness temperature, T0, which was originally taken to be 253 °K but has been modified in more recent versions of the algorithm. Of the area defined as cloud, the coldest 10% is assigned a rain rate R10 and the next coldest 40% is assigned a lower rain rate R40. Values for R10and R40 were originally specified as 8 and 2 mm/hr, respectively, but have been adjusted slightly in more recent applications.
RAINSAT is a supervised classification algorithm which is trained to identify areas of precipitation from a combination of VIS and IR imagery. At night, visible imagery is unavailable and RAINSAT reverts to apure IR technique. RAINSAT and its relatives are among the few VIS/IR algorithms that are used operationally in middle latitudes.
Recently, a similar approach by using a multivariate classification scheme and raingauge data to estimate daily mean areal precipitation is proposed.
The factors to be considered in comparing methods include the following:
- the type of regression model employed (linear, non-linear? multivariate)
- the inteval between images (slots); the time averaging period
- the space averaging scale; the threshold temperature adopted
- data treatment (e.g. linear or temperature weighted accumulation) additional data incorporated (e.g. water vapour Channel, visible Channel or contemporary surface
- localization of calibration (time or space varying TIR features, variation with geographic location, time of year, character of season, topography and local storm climatology.
Comparison of actual and estimated rainfall at different rainfall ranges, July 1995 for the whole country
Comparison of observed and estimated over western Ethiopia for the period June to September 1994.
Comparison of observed and estimated over northeastern Ethiopia for the period June to September 1994.
The IR information is separated into three classes, based on
temperature thresholds, for improving quantitative rainfall
estimation for cold convective clouds, middle layer clouds and
warm coastal clouds.
The METEOSAT spin scan radiometer operates in three
0.5 - 0.9 μm (visible band - VIS)
5.7 - 7.1 μm (infra-red water vapour absorption band - WV)
10.5 - 12.5 μm (thermal infra-red band - IR)
The amount of radiation absorbed by water vapour is dependent on the amount of moisture in the radiation's path and the wavelength of the radiation. Increased amounts of moisture, or water content, in the radiation’s path lead to more absorption of the radiation emitted from lower layers. Therefore, if the air temperature decreases with height, higher moisture content result in colder brightness temperature. On a 6.7µm image the coldest temperatures correspond to high cloud tops, whilst the warmest are observed over lower altitude areas when the air is very dry through a deep layer in the atmosphere. For the 6.7µm water vapour channel, the radiation values may also be converted to brightness temperatures. A difference exists between WV (6.7µm) brightness temperature and that of the standard IR (11µm) channel. This is attributed to the absorption and re-radiation by water vapour above the earth's surface or clouds. It is this difference that allows a distinction to be drawn between cirrus and moist updraft regions.
100mb - 250mb
250mb - 400mb
400mb - 700mb
100mb - 400mb
400mb - 700mb
i) Visible imagery − bright white but diffuse. Thick anvils may cast shadows on lower
surfaces whereas thin anvils are often translucent to lower features.
ii) IR imagery − bright white patches, usually coldest (whitest) cloud, except when
active thunderstorms are present.
d) Cirrocumulus − cumuliform ice crystal clouds formed by upward vertical motions in the
upper troposphere. May precede rapidly developing cyclone.
i) Visible imagery − thin patches of clouds, gray to white, usually in advance of a
cyclone. Individual elements often below the resolution of geostationary sensors.
ii) IR imagery − similar to cirrostratus, white to gray clouds subject to contamination.
3) Low Clouds − composed of water droplets. Wintertime conditions and vertical growth may allow glaciation.
a) Cumulus − similar to detached cauliflower-like clouds with sharp outlines. Often, a
region of unorganized cumulus (“popcorn”) forms over landmasses during fair weather. Cumulus clusters whose edges are clearly visible are referred to as “open cell” cumuli.
i) Visible imagery − scattered individual elements are often below the resolution of
geostationary sensors and appear as gray areas due to contamination. Large individual elements and groups of broken cumulus appear as bright white blobs of clouds.
ii) IR imagery − only large areas show due to contamination, appearing as dark gray
b) Towering Cumulus − cumulus of moderate or strong vertical extent.
i) Visible imagery − similar to cumulus but elements are larger, so are more likely to be distinguishable as bright white blobs.
ii) IR imagery − similar to cumulus, but appearing as lighter gray blobs.
c) Cumulonimbus − cumulus of strong vertical development with or without cirrus anvils. Vary greatly in size and shape depending on storm intensity and environment. If upper level winds are weak, mature thunderstorms are circular cirrus clouds often with cirrus plumes (filaments) streaming out nearly symmetrically in all directions with occasionally lumpy, penetrating tops (indicated by shadows on visible imagery). Stronger winds aloft blow the cirrus anvil downstream and create a diffuse downwind boundary with a sharp, smooth upwind boundary. In region of vigorous thunderstorms, cirrus anvils may merge into cirrus canopies. The active cells are indicated on visible imagery by their lumpy penetrating tops. Much of the cirrus in the ITCZ is actually decaying cirrus anvils.
i) Visible imagery − bright white cellular shape covered with diffuse thin cirrus and often a lumpy penetrating top.
ii) IR imagery − bright white, smooth cellular shape. Enhancement techniques help identify the maximum cloud tops by relating cloud top temperatures to height.
(2) Vegetation − snow-covered forest regions appear gray and mottled, rather than white. Snow-covered short grass regions appear white and smooth. The brightness of snow covered grass decreases with increasing height of the grass and decreasing snow depth.
(3) Snow Depth and Age − generally, the deeper and newer the snow, the whiter it appears. Rain on snow makes it appear grayer.
(4) Sun Angle − brightness of snow decreases rapidly when the sun angle drops below 45°.
(5) Cloud cover − since snow does not move, looping imagery may assist in distinguishing snow cover from clouds.
(6) Fracture Lines − ice can sometimes be distinguished by its location (water bodies) and the presence of dark fracture lines. Ice may also look chunky.
(7) Wind may blow snow around, and the peaks and depressions may add texture to an otherwise smooth sheet, helping to distinguish snow from fog and snow from ice.
ii) IR imagery − detection of ice, snow cover, and low clouds is very difficult without simultaneous visible imagery. Snow may appear as a patch colder (lighter) than bare ground.
b) Surface Variation
i) Visible imagery − differences in reflectivity among land cover types (e.g. grass and forest) may be apparent as variations in shades of gray. Some highly reflective sandy
areas (e.g. White Sands, NM) may be seen as white to gray unmoving blobs that could be confused for snow. Shadowing in mountainous areas may give the landscape texture.
ii) IR imagery − differences in land cover may produce differences in surface heating that may be apparent as variations in shades of gray. Typically, these are large areas (e.g. California’s Central Valley, the Great Salt Lake) that are significantly warmer or colder than their surroundings.
3) Enhanced Cumulus − area of cumulus congestus, towering cumulus, or cumulonimbus clouds. Associated with fronts, PVA, or orography, and appear as very bright dots in a field of otherwise uniform open cell cumulus.
4) Sea Breeze/Land Breeze − a nearly continuous band of cumulus clouds that tend to parallel the coastline. Sea breezes are most often found inland during the late afternoon, and land breezes offshore during the early morning. A cloud-free region along and off the coastline indicates the subsidence portion of a sea breeze cell.
5) Ship Tracks − Long, narrow, stratocumulus cloud plumes that form in the wake of ships when the winds are light, and there is a subsidence inversion capping the rising air.