7#,,- t+xm /_*_m__T4______ MODULE 2 - HOW RADAR IMAGING WORKS D) SCATTERING MECHANISMS Objectives: 1. Students will be able to draw and label a diagram showing an example of what happens to a transmitted pulse when it hits four types of surface. 2. Students will be able to predict what the brightness will be, (high, medium or low) for each type of surface. 3. Students will be able to list several examples of each type of scattering mechanism. Radar is reflected in the same manner as visible light. We refer to this reflection as scattering. The radar pulses sent out by an imaging radar are scattered upon contact with the earth's surface. The way in which the energy contained in the pulse is scattered is known as a scattering mechanism. An imaging radar of the type we are discussing can only measure the energy scattered back towards it, i.e. the backscatter. In this section, four different types of common scattering mechanism are introduced, including scattering from four types of surface: 1. Smooth surface 2. Rough surface 3. Double-bounce 4. Vegetation Layer Smooth Surface The first type of scatter we shall consider is from a smooth surface. As can be seen in the diagram below, most of the scatter from a smooth surface is in the forward direction, away from the radar. Only a very small fraction of the energy in the transmitted pulse is reflected back towards the radar.  Thus, backscatter from a smooth surface will tend to be very low (say less than -20dB) and will appear dark in the radar image. The exception to this is when the radar is pointing straight down, so that the smooth surface is at right angles to the radar look direction. In this case, almost all the energy in the pulse will be reflected back towards the radar, and the backscatter will be very high (white in the radar image). The types of surface that would demonstrate smooth surface scattering are smooth water, dry lake beds, some types of smooth rock and bare soil. The effect of scattering off a smooth surface is like a pool ball hitting the side of a pool table. The ball is "reflected" at the same angle it hit the side, but in the opposite direction. To demonstrate the same effect using visible light, use a flashlight and a flat piece of aluminum foil. Have the students stand above the aluminum foil and aim the flashlight at the foil. They will notice that most of the light is reflected in one direction, away from them. Rough Surface The second type of scatter we shall consider is from a rough surface. As seen in the diagram below, the scatter from a rough surface is in all directions. Some fraction of the energy in the transmitted pulse is reflected back towards the radar. In general, the rougher the surface is, the higher the backscatter. This can be understood if we consider that the rougher a surface is, the more likely it is that some segments (or facets) of it will be inclined at right angles to the radar look direction. Backscatter from a rough surface will tend to be greater than -20dB and will appear grey to white in the radar image. The types of surface that would demonstrate rough surface scattering are water blown by the wind or with waves, lava and other types of rock and bare soil with clumps.  To demonstrate the effect of scattering from a rough surface using visible light, use a flashlight and a crinkled piece of aluminum foil. Double-bounce The third type of scatter is from two surfaces, one flat on the ground (horizontal), the other upright (vertical). The reflected pulse hits both surfaces one after the other. This type of scattering is known as double-bounce. From the diagram below, most of the scatter for a double-bounce mechanism is back towards the radar, i.e. in the backscatter direction. Double-bounce backscatter will tend to be fairly high (HH, VV greater than -10dB) and will appear light grey to white in the radar image. Double-bounce scattering occurs commonly in urban areas, where there are plenty of vertical surfaces (the sides of buildings) and horizontal surfaces (sidewalks, streets). It can also occur frequently in nature, whenever there are upright vegetation stems (stalks, trunks) and a relatively smooth (and flat) surface underneath. Examples are flooded forests, forests where the vegetation lying on the ground has been removed, rice fields, corn fields, marshes and swamps.  To demonstrate the effect of double-bounce scattering using visible light, use a flashlight and two flat pieces of aluminum foil, one on a table top, the other held vertically next to it. Vegetation Layer The fourth type of scatter is from a layer of randomly oriented scatterers, which is common in many types of vegetation. This type of scattering is more complicated than the other three. The radar pulse penetrates the vegetation layer, then is scattered after hitting one of the randomly oriented branches or leaves in the canopy. From the diagram below, the scatter for a vegetation layer is in all directions, with some in the backscatter direction. Backscatter from a vegetation layer will vary. Some vegetation layers will be fairly low in backscatter (HH, VV less than -10dB) and will appear dark grey in the radar image. Other vegetation layers will have fairly high backscatter (HH, VV a few dB above -10dB) and will appear light grey in the radar image. Most naturally occurring vegetation, such as forests, shrubs, and grassland have scattering patterns which are characteristic of vegetation layers, but vary depending on the wavelength. Forests tend to have fairly high backscatter at all radar wavelengths, because forests are made up of objects of all different sizes (from leaves to trunks). Shrubs are smaller than forests and tend to have fairly high backscatter at short and medium wavelengths (i.e. less than say 30cm). Short grass tends to have low backscatter at all wavelengths. Long grass may have high backscatter at shorter wavelengths (say, less than 10cm) but low backscatter at longer wavelengths. Scattering from a vegetation layer and double-bounce backscatter can often be seen together, for example in a radar image of a forest. We will see later how these two phenomena can be distinguished.  Radar Backscatter Characterisitics Of Scattering Mechanisms Smooth Surface Overall backscatter very low (HH, VV < -20dB), except when looking straight down Rough Surface Overall backscatter low to high (HH, VV > -20dB) VV > HH HV < -20dB Phase Difference within 30o of zero Correlation coefficient > 0.7 Double-bounce Overall backscatter high (HH, VV > -10dB) Phase Difference within 80o of +180o Vegetation Layer HH and VV within 1dB of each other HV > HH - 7dB or HV > VV - 7dB Correlation coefficient between 0 and 0.5 These radar backscatter characteristics can be compared with the mean values in the "Statistics" results generated by Macsigma0. The above "rules-of-thumb" should separate most different types of scatterers, but will be wrong in some cases (there is always an exception to every rule). Mapping Vegetation Type Given the characteristics of the scattering mechanisms described above, it is possible to estimate the dominant (strongest) type of scatter for each pixel in a radar image data set where all the polarizations are available. We might for example, classify a pixel as being one of three types: Surface Scatter Double-bounce Vegetation Layer Suppose that we have three frequencies of radar data available to us. We can classify data from each frequency separately into one of the above three types of backscatter. We may get a different answer for each frequency. It is then possible to use these different answers to classify the image into different vegetation or ground cover types. The NASA/JPL AIRSAR system is an imaging radar which produces radar image data at three frequencies: C-band (6cm wavelength); L-band (24cm wavelength); and P-band (68cm wavelength). At each frequency, images for all polarizations are generated. Three frequency AIRSAR data can be found on the SIR-CED CD-ROM in the SITES folder as files of type Filename.STK. This data can be classified as to the dominant scattering mechanism as described above. Also on the SIR-CED CD-ROM can be found vegetation maps, generated from NASA/JPL AIRSAR data, using an algorithm developed by A. Freeman of JPL. The vegetation maps are stored on the CD-ROM as image files named Vegmpbyt.pic. Each pixel in a Vegmpbyt.pic file is classified as corresponding to an urban area (e.g. a city or town); a forest; medium vegetation, such as shrubs or a mature corn field; low vegattion, such as grassland, scrub or low-lying agricultural crops (e.g. wheat or barley); no vegetation (rocks, bare soil and water surfaces). Within the three vegetated classes (forest, low and medium vegetation), there is a further subdivision into, for example, forest which has a significant amount of double-bounce [Forest (D)] and forest which doesn't [Forest]. The clasification is based on a model of the scattering mechanisms expected from these different ground cover types at all three AIRSAR frequencies and was trained on AIRSAR data from a few typical scenes. Some pixels are unclassified and are designated as bright, medium or dark. These pixels do not match the models. Classification Rules for Vegetation Maps The vegetation maps found in the Vegmpbyt.pic files were generated following rules broadly along the lines of those described below: No Vegetation - surface scatter dominant at all three frequencies Low Vegetation - vegetation layer scatter dominant at C-band, but not at L-and P-band. Low Vegetation (D) - classified as low vegetation, but with a significant amount of double-bounce (implied by large HH-VV phase difference) at C-band. Medium Vegetation - vegetation layer scatter dominant at L-band, but not at P-band. Medium Vegetation (D) - classified as medium vegetation, but with a significant amount of double-bounce (implied by large HH-VV phase difference) at L-band. Forest - vegetation layer scatter dominant at L-band and P-band. Forest (D) - classified as forest, but with a significant amount of double-bounce (implied by large HH-VV phase difference) at P-band. Urban - Double-bounce scatter dominant at all three frequencies. Some pixels do not match any of the scattering mechanism models or the rules described above. These are unclassified pixels. To differentiate between these unclassified pixels, three levels are given: Bright - Unclassified pixels which have high backscatter at L-band and P-band. Medium- Unclassified pixels which have high backscatter at L-band but not at P-band. Dark- Unclassified pixels which have do not have high backscatter at L-band and P-band. WARNING - the vegetation maps in the Vegmpbyt.pic files are our best estimate of the ground cover class based on current models and understanding of radar backscatter. They may be wrong in some cases. ue5tD5fD5VC CddPro C Q*Y-t X QYt XP  "Af##### #######  C )e@)e,-.{/z{0yz1xy2wx3vw5uv6tu7st8rs:qr;pq<op=no>mn?jm@ej "@e## ###P "f##### #######  C eeggjjmmnnooppqqrrs st tu uv vw wxxyyzz{{~~e "e## ###P  "*###########(# C"*#####( Q, X 1i+ 0h," currentpoint /legm_top exch def /legm_left exch def " currentpoint '/legm_high exch legm_top sub 2 div def (/legm_wide exch legm_left sub 2 div def gsave newpath 0 setlinewidth &/save_matrix matrix currentmatrix def &/legm_wide legm_wide 2 mul 19 div def %/legm_high legm_high 2 mul 4 div def legm_left legm_top translate legm_wide legm_high scale 0 0 moveto 0 3 lineto 18 3 lineto save_matrix setmatrix clip gsave ddProd6dPro,"- grestore """-,-,-,"<;<;<;"KJKJKJ"ZYZYZY"ihihih"xwxwxw""""""""""ddPro grestore  CpddProP "#### C"#P "###########(# C"#####( A^-p H"W currentpoint /legm_top exch def /legm_left exch def "!_ currentpoint '/legm_high exch legm_top sub 2 div def (/legm_wide exch legm_left sub 2 div def gsave newpath 0 setlinewidth &/save_matrix matrix currentmatrix def %/legm_wide legm_wide 2 mul 8 div def %/legm_high legm_high 2 mul 8 div def legm_left legm_top translate legm_wide legm_high scale 0 0 moveto 4 1 lineto 7 2 lineto 7 5 lineto 4 7 lineto 0 7 lineto 0 0 lineto save_matrix setmatrix clip gsave ddProd6dPro4W ^WYY]]^]^ W]"1o grestore "WTTT"X"Y"ZUUUUUU"[***"\"]ddPro grestore  Cp*W ^W[^^ [ WWWddProP "W############ C######""W currentpoint /legm_top exch def /legm_left exch def "*_ currentpoint '/legm_high exch legm_top sub 2 div def (/legm_wide exch legm_left sub 2 div def gsave newpath 0 setlinewidth &/save_matrix matrix currentmatrix def %/legm_wide legm_wide 2 mul 8 div def %/legm_high legm_high 2 mul 8 div def legm_left legm_top translate legm_wide legm_high scale 0 0 moveto 4 1 lineto 7 2 lineto 7 5 lineto 4 6 lineto 0 7 lineto 0 0 lineto save_matrix setmatrix clip gsave ddProd6dPro4"W)^"W\#\]$]^(Z^)WZ":o grestore ""WTTT""X""Y""ZUUUUUU""[***""\""]ddPro grestore  Cp*"W)^"W"[$^'^([)W"W"WddProP ""W############ C######" \ currentpoint /legm_top exch def /legm_left exch def "#o currentpoint '/legm_high exch legm_top sub 2 div def (/legm_wide exch legm_left sub 2 div def ""^ currentpoint )/legm_radius exch legm_top sub 2 div def Llegm_left sub 2 div dup legm_radius gt {/legm_radius exch def} {pop} ifelse gsave newpath 0 setlinewidth &/save_matrix matrix currentmatrix def 9legm_wide legm_left add legm_high legm_top add translate Alegm_wide legm_left add neg legm_high legm_top add neg translate ,/legm_high legm_high 2 mul legm_top add def -/legm_wide legm_wide 2 mul legm_left add def *legm_left legm_radius add legm_top moveto Hlegm_wide legm_top legm_wide legm_top legm_radius add legm_radius arcto 4 {pop} repeat Jlegm_wide legm_high legm_wide legm_radius sub legm_high legm_radius arcto 4 {pop} repeat Jlegm_left legm_high legm_left legm_high legm_radius sub legm_radius arcto 4 {pop} repeat Hlegm_left legm_top legm_left legm_radius add legm_top legm_radius arcto 4 {pop} repeat save_matrix setmatrix clip gsave ddProd6dPro- \#o"8t grestore " U "U"V&f&f&f"V323232"W???"XLLL"XYYY"Yfefefe"Zs2s2s2"Z"[ˌˌ"\"\eee"]111"^"^"_ٗٗٗ"`ddd"`111"a"bddPro grestore  C @ \#o H"* currentpoint /legm_top exch def /legm_left exch def ". currentpoint '/legm_high exch legm_top sub 2 div def (/legm_wide exch legm_left sub 2 div def gsave newpath 0 setlinewidth &/save_matrix matrix currentmatrix def &/legm_wide legm_wide 2 mul 19 div def %/legm_high legm_high 2 mul 4 div def legm_left legm_top translate legm_wide legm_high scale 0 3 moveto 0 0 lineto 18 0 lineto save_matrix setmatrix clip gsave ddProd6dPro4*.*+,-."B grestore "*"*"*-,-,-,"*<;<;<;"*KJKJKJ"*ZYZYZY"*ihihih"*xwxwxw"*"*"*"*"*"*"*"*"*"*ddPro grestore  Cp*..**.ddProP ".#### C".#"f"-f1!81"$8P "!########  C !####P "########  C$####P "########  C$####P ""########  C$"$"#$####P "$######  C %&####P "'########  C %'####P # ###################"####  CT&"$%&"################" c"8c" Y currentpoint /legm_top exch def /legm_left exch def "#p currentpoint '/legm_high exch legm_top sub 2 div def (/legm_wide exch legm_left sub 2 div def gsave newpath 0 setlinewidth &/save_matrix matrix currentmatrix def 9legm_wide legm_left add legm_high legm_top add translate */x_proportion legm_wide legm_high div def x_proportion 1 scale 0 0 legm_high 0 360 arc save_matrix setmatrix clip gsave ddProd6dPro4 X#q [n!X[nq"X[nq#[n"8q grestore " Y " ZEEE" ["""" \..." ]:.:.:." ^EEE" _QsQsQs" `]]]" ahhh" bt\t\t\" c" d" eDDD" f" g" h---" i" jrrr" k" l" m[[[" n" oddPro grestore  C P Y#p X1!81!#81"$8 1 8 1(* 8P "*o########  C (*n1*o.no1n ####"*|P "o########  C 4mmmnnoo ####"|P "########  C ( ####"P "+########  C (+1+.1 ####"+ 1u${ 8 1z$ 8 1~$ 8 1$ 8 1$ 8 1$ 8 1$ 8 1$ 8 1$ 8 1#u){ 8 1#z) 8 1#~) 8 1#) 8 1#) 8 1#) 8 1#) 8 1#) 8 1#) 8 1ux 8 1wz 8 1z} 8 1| 8 1~ 8 1 8 1 8 1 8 1 8 1 8 1 8 1 8 1 8 1 8 1 8 1 8 1 8 1 8 1)u+ 8 1)+ 8 1)+ 8Q!XQ"'X UUUU1 # 80/.P@  0@0@@ `@@ pP@  @ `@?@  ?00  ddPro8 XG,  Helvetica .+[Sր) P#) A@ ) C\`) E) B ) O) R ) N`) EA)  %)I `)M) A ) G) I)N) G)  z`)R) A) D ) A`) R) - )C ddProH '= (7S) I)R) -)C) E@) D .S S S SddPro S *d dPro qquwutwqz"-+KFddPro d dPro qEQEQNKE"H"""z"u"o"j"d"_"Y"T"NddPro @@!(' UUUUQ7" X@@@ UUUUq 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'#  p 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'# @@@"('`72Z@@@p6,   !$'* ,@@@" ,('`/"Z@@@p600 / - *'$"   @@@"(' A$' H@@@ q('#'(#  p('#'(# @@@" (' ""Q"-( X@@@ ""qf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$# pf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$#@ ddPro0w ~S,  Helvetica .\#+Smooth SurfacefB6P(PP PddProP"# # # # ########## #### ### # ################ ################### P"####################### ################################## ###d dPro qot}}utrvoy}"+*KFddPro d dPro qo{o{yvo""""yddPro d dPro qJUWbJURbT_W]JU"}"}"w"q~"kw"eq"_j"XdddPro d dPro q}lz}l}zzy}l"""ddPro d dPro qcrcrpoc"""wddPro d dPro qxx|yx""""}ddPro ddPro0r P,  Helvetica .\#( Rough Surface@@ P P(' UUUUQ7" X@@@ UUUUq 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'#  p 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'# @@@"('`72Z@@@p6,   !$'* ,@@@" ,('`/"Z@@@p600 / - *'$"   @@@"(' A$' H@@@ q('#'(#  p('#'(# @@@" (' ""Q"-( X@@@ ""qf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$# pf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$#@d dPro q^m^klm^"}"}"uddPro m . . p. .ddPro .  0`1v 8 `0 1l 8ddPro 1w ddPro0O t,  Helvetica .\#+w ddProddPro .1m ddPro0X m( ddProd dPro  . q"P6KFddPro d dPro q""""ddPro d dPro q<\Ii<\CiFfId<\"""|"v"p"j"d"^"Wy"Qr"KlddPro @@#(' UUUUQ7" X@@@ UUUUq 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'#  p 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'# @@@"('`72Z@@@p6,   !$'* ,@@@" ,('`/"Z@@@p600 / - *'$"   @@@"(' A$' H@@@ q('#'(#  p('#'(# @@@" (' ""Q"-( X@@@ ""qf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$# pf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$#@ ddPro0{$ z.\#( Double-bouncere  ddPro "q>""} """;""" " "nb#1"""|"p"{m "{ ""ug"h"""{ "qC>"I "yD"Q"L"B"x;"Z"N k+k"k+"k4"k="kF"kO"kX"ka"kj"ks"k|"k"k"k"k"k"k"k"k"k"k"k"k"k"k"k"k  -"-"6"?"H"Q"Z"c"l"u"~""""""""""""""""d dPro qpv~~vvsxp{~",,KFddPro d dPro qYifuYiaucsfpYi"w"w"q"kz"dtddPro d dPro qR`R`_^R"z"z"r"i"addPro d dPro q_j_jhe_"y"y"t"n"hddPro d dPro q" """"ddPro d dPro q" """"ddPro d dPro qwwy}w"""""}ddPro d dPro qYhYfghY"""z"q"hddPro ddPro0p /,  Helvetica .\#(2Vegetation Layer@@  (' UUUUQ7" X@@@ UUUUq 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'#  p 2-!$ % & ) + - - - - -,+*(%"! #&)-./122 1 1 / - + )'# @@@"('`72Z@@@p6,   !$'* ,@@@" ,('`/"Z@@@p600 / - *'$"   @@@"(' A$' H@@@ q('#'(#  p('#'(# @@@" (' ""Q"-( X@@@ ""qf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$# pf!!,*$##$"&!'!)"*#*$*%*'((')'+%+$,#,"+!+!)!(!'"%#$#@ The following are approximate characteristics of backscatter for the four different surface types:is a summary ofsare %&@AQP[:;<N M^,-_`b |}`d !G!U!!"""-":">"?%%%%&7&D&z&&&'k @p @y @j @ax@ $$ @,P'k'|''(`(f(()+)0))*7*>****+7+>+\+h++, RSTUVW `ghv @@B&'@AQR[^_"89:<NO~ { | JKM^_󖹏{t0!0!0 !0!0 !0!0!00!0!0!0!00!00 !0x!0!0!0!D0,_+,./^_abd  OWbA`Ķtj !0$@ !0$@ 0$@ !0$@ !0$@ !0$@ !08$@ !08$@ 08$@!00!0!0!0!0!0) wx!9!:%%%%%&6&7&y&z&&'j'k''(_(`(()*)+)l)m*6*7****+5+6+7,VӽӲӧӜӜӑӜӑӜӑӜӜӜӑ !0 !08$@ !0 !0 0 !0 !0 !0 !0 !0!00 !0$@2  ; or when the surface is tilted towards the radar.ess Surfaces may be rough at different scale lengths. Comparing the scale length with the radar wavelength we can get an idea whether the surface will appear rough to the radar and thus whether it will give high or low backscatter. For example, if we look at a ploughed field, we may find that the ploughed furrows are about 6cm high. This is one measure of the scale of the roughness for that field. With a roughness scale of 6cm, the field will appear rough (i.e. bright in the radar image) at C-band (6cm wavelength) and X-band (3cm wavelength), but not at L-band (24cm wavelength). wavelengthwavelengthwavelengthwavelengthwavelengthfirst made to workwavelength-@,; k'}- ; !O !g#*+&-2-@ !! !!@!!!@!!!@!!!'kLM_NOPHH(FG(HH(d'@=/RBH/TonyF:Mod3cont.txt, -:LaserWriter ChicagoNew YorkGenevaMonacoPalatinoTimes HelveticaCourierSymbol+d+d+d( ' ; < = &HW")+9:; !0? & 0 p z !!"&"0&&+Z+d-- RSTUVW  `ghv <C @$)i+4 SectionD.txtSIR-CED Tony Freeman Tony Freeman