Week 5: Lecture 1

LIDAR and Remote Sensing of Vegetation

What is LiDAR?

Is it Active or Passive?

How does it work?

what is their end product?

-Light Detection and Ranging

- Active Remote Sensing

- the soruce of the EMR is the aircraft or sensor itself

- Laser altimetry: take a beam of light > bounce it off an object > measure the amount of time it takes to come back > what you get is a measure of distance 

- They can create DEMs! 

What is ICESAT? 

What does it stand for?

What was ICESAT's purpose?

why is it important?

what is it doing?

- main purpose: focusing on earth's frozen reaches 

- studying the chriosphere --> important for understnadin global climate nad understanding how man is creating changes

-interested in the overall ice inventory, esp polar icesheets; i.e. in greenland

-how is melt affecting sea levels?

what are the tradeoffs of the ICESAT mission?

-using laser altimetry to capture surface details -- icesat is 10times more accurate than the laser altimetry that was done on mars

- there is a complex interplay of atmosphere, oceans, and land

-TRADE OFFS:

-while the primary drivers of the mission were to study the icesheets, it can do other things too

-pick and choose battles: they chose chriospheric science; i.e. glaciology and sea ice

-the tradeoffs: have to do with the spacing of the orbits and how high latitude that the space craft can go.

-high latitude = reduce ht etemporal frequency for coverage at the equator!

-it is retreiving biomass informaion -- taking inventories of forest biomass; doing carbon budgets

-related to REDD

-its crippled!

-but its the only game in town

-they had to descope the mission bc if problems with the diodes that were generating the lasers; so to keep it alive, they cut down the scope of the mission to keep the mission alive. 

-launched in 2003.

how many times does the lidar beam from space come down a second?

what is the accuracy?

40 times per second!

For ICESAT, the accuracy is roughly 6in.

X, Y, and Z - long, lat, and elevation

(as well as some measures of the scattering properties of different materials)

1. they are fundamentally a small pulse of laser light so the spatial coverage is quite poor; dont get big blankets of data like with RADAR

(you can create a "swath" by spining the lazer conically--but its still only a few hundred of meters; depends on the height of the aircraft)

2. the closer to the ground, the narrower the swath and thus the narrower the footprint

So you're flying along > directing pings down > and measuring the 2-way return time to get elevation of the surface

1. you need a datum -- the geoellipsoid referencing the data

2. most importantly: the big weakness = precisely knowing the position of your plane b/c of the pitch and roll of the plane

-post processing! 

-the correction of the data is a huge part of the processing and the expense

-the aircrafts themselves are riggedby with all sorts of gyroscopes and onboardi ntertial GPS nav systems and little sensors on the wings that try to catalogue ever so slight changes in the wing alance so they can be corrected during post procesing.

-Conical Sampling: swirling the laser points -- you get things called "slivers"

-Slivers: gaps in data -- parts where the plane miss part of the ground (either didnt fly over twice or the laser cut out)

- Lidars dont create an image or picture! 

-Lidars are jsut spraying down these points and they tend not to be uniformly distributed in space or geographically in a uniform gridded way

SO: if you want a DEM --> you need to interpolate to tget the correct X,Y,Z data

-you can get 5-7cm heights depending on the flying height of your system. 

1. Discrete Return

2. WaveForm

- you just send down this pulse of light > record a number when it comes back 

- the idea is: you send down a small focused beam > when the first echo returns > you record that time 

- you know from your intertial navigation system where it is in space > and oyu record the X,Y--geographic location, and Z - elevation value - based on the two way return time after all the post processing

-most commerical systems collect discrete return; very high spatial res, veyr precise DEMs.

-Good for measuring streets and cement

-Think: a narrow pinpoint of light, slips through and bounces back off of the first, good, strong, reflector that it hits.

1. first return elevation: if you're spraying these laser points at a tree for example those bright values would mean shorter length of itme was set out there which translates o a higher z value on the ground.

2. intensity: you can also record the strength of the return which gives you the sense of how clean of a scatter or shape of a reflector that surface was.

- these preserve not onyl a single number of hte timing of the return, but a whole sequence of numbers

-the echo gets smeared out over time -- and returned over a longer interval of time

-better for if oyu're trying otm easure the height of a forest 

-tries to capture the complexity in the structure of the earth's surface and typically has a larger footprint.

1. they tend to have a larger footprint

2. rather than having a small focused beam which might get through a part of a tree or not > wavefrom lidar floods a tree or whole larger area with light > rather than recording a single Z value in time, it records the timing of the whole mess that comes back

3. how the return might look: good return at the top of the canopy > not such a good return within the canopy > then a pretty good return when the signal hits the ground --> SO this whole shape is recorded. 

*** waveform is particularly important for people who are trying to use RS to estimate forest biomass or something or for REDD. 

- if you have the raw data - you have more control over what you extract from the data. 

- you want the points; b/c you can manipulate them to get different things

-let the company do all the gyroscopy correction (all the 0 and level 1 processing)

- the density of point clouds tend to increase over areas where the plane has overlapped

- the task becomes : taking a messy, not uniform gridded datset and processing the point clouds to create a higher levle derivative product that is mroe afforable and discernable.

Lidar Processing: TIN

what does TIN stand for?

what is TIN processing?

-TIN: Triangulated Irregular Network

-very standard procedure in ArcGIS

-a vector, not raster, problem

-take a bunch of irregularly spaced points > and makes it into something more regular and gridded

-takes points and triangulated them into a plane

-by triangulating points, you can see where hte points clouds were more diffuse and ares where the point clouds are more tied together

-once you get a uniform spatial field, you can either resample it into a uniform grid (raster), or leave it as a TIN (vector)

-if you like raster data sets, yo ucan super impose the grid saple over the square pixels and extract the valeus out of the uniform gridded product --> this is called resampling.

- there are lots of different types of resampling

-resampling just refers to any time you are changing the geographic coordinats and the dimensions and reassinging new numerical valeus through estimate or planar modeling of the point cloud

- you can either have a coarse grid sample and resample it "finer" OR you can have a finer product and resample it coarser. 

-LIDAR surfacing: if you want only the surface topography, you can subtract out the surface obstructions using a math algorithm

-i.e. you have to make some decisions " i'm only going to keep the lowest elevations and anytime there are marked dramatic jumps i will assign a standard deviation filter and just throw those points away" > those high elevation points are then removed.

-the denser the point cloud and more tightly focused those beams > the more effective this type of lidar surfacing becomes

-like everything remote sensing, getting ground truth/ground validation is important

-you may need to correct for the pitch and roll of your plane!

-overall, the precision is pretty good.

Back to IceSAT: 

-True or false: ICESAT is the first orbiting satellite radar.

- True or false: launched in 2003?

-what type of lidar system is it? Discrete? Waveform? different?

-what is the size of the footprint?

-true: icesat is the first orbitting satellite radar

-true: it was launched in 2003

- It does listen to a waveform BUT its a special type: the standard product doesnt preserve the entire thing -- it will make some algorith which listens to the shpae of the wve form cloud and select out the peaks of interest.

-the size of the footprint: 70

-unlikey to get a clean ground return of a forest canopy b/c of the large 70m footrpint -- its gunna hit a tree so the returns that you get, b/c of the height of the satellite adn the large footprint, are going to be diffuse and messy. 

ICESAT: 

-is icesat TRULY a global mapper AT ALL TIMES? 

-what else does ICESAT study?

-NO, there are gaps and diamonds near the equator; but in the higher latitutdes, you get more continuous coverage (remeber the tradeoffs)

-While the primary purpose was icesheets, you can also study clouds, vegetaiton, land topography, etc. 

What does MOLA stand for?

when was MOLA launched?

what was its purpose?

why is it interesting? (with relation to earth?)

-Mars orbiter Laser

-Launched in the late 90s

-Its purpose was to map the topography of mars

-It was the first planet (even before earth) to use a lidar satelite the map a planet

-Any shorter, you strat to get into eye safety problems and any longer than that, you're getting messed up by atmospheric absorption --> and if oy ugo out way longer, you'll start getting out of hte optical and not get a return as you get into the thermal.

-RADAR in the microwave wavelengths

-it is a measure ofthe naturally emitted microwavelengths from the earth itself as opposed to spraying the world with microwaves

-very interestinf for broad sale mapping of sea ice --> the reaon why do know that sea ice has been shrinking so dramatically during summer is because of passive microwave systems

-passive microwave systems have been up there since the 70s but nooned cared about sea ice; but now, in looking back, we can see that the sea ice is going down

Passive Microwave systems:

1. Based on theenergy of a quantum -- would you expect the eerngy content of passive microwaves to be high or low?

2. if the fundamental energy of that radaition is low, would yo uexpect the spatial resolutions you collect to be high or low?

1. LOW!

2. LOW! --> need to collect over a great area to get a bunch of wimpy photons to get a good enough SNR; the best spatial res is at 5km. 

Remote Sensing of Vegetation:

-What are the 6 significant things about vegetation: (based on the slides)

- True or false: Remote Sensing tasks for vegetation are limited

1. covers about 70% of the land surface

2. all life stems from photosynthesis

3. food is provided by plants

4. they are a source of food and energy

5. ecosystems

6. global change

-FALSE: there are myriad of uses and the tasks are wide and varied.

Remote Sensing of Vegetation:

What are the 3 aspects of Vegetation Remote Sensing We will be looking at?

1. Spectral Characteristics

2. Phenological Characteristics

3. Vegetation Indices 

Spectral Characteristics:

1. True or False:In the optical wavelengths, REFLECTANCE peaks NIR?

2. In what wavelengths are there dips when looking at vegetation?

1. FALSE: in the optical wavelengths, REFLECTANCE peaks in the GREEN

2. Dips in the red and the blue becuase they are absorbing these wavelengths!

Spectral Characteristics:

1. Does all vegetation look the same in RS?

2. If you had to design a cheap satellite and you wanted to distinguish between real parks and astroturf parks, what band would you equip your satellite with?

1. NO! even veg within the same species has different spectral response functions based on how the health of veg waxes and wanes and moisture

2. You'd want to use the NIR because that will give you the most reflectance!

-based on the health/phenology of the leaf

-there is a physical reason for this: the changing colors in leaves are being driven by PIGMENTS

-Those pigments are based on Chlorophyll A and B

-As the leaf goes through its cycle and goes from the time when the cells are full of chlorophyll to the time that the chloropyll goes away.

-Scale up!

-Higher resolution data tends to be "noisier" than low resolution data 

-When you go to a coarser spatial resolution, there is a lot more averaging going on with respect to spectral characteristics

-i.e. when you're looking at Fall colors from MODIS (4.5km spatial res) - it gets easier b/c you can see a bigger picture.

1. If you look out beyond from the narrow range of the visible and go out into the NIR - you start to see dips
2. this has to do with the water content!
3. remember: water is a big absorber - think back to the atmospheric bands and GHGs (water is the strongest GHG!)
4. if a material will absorb water --> it will display these absorption characteristics in the spectral reflectance of materials.

1. The absorption of the blue and red is owing to the presence of Chlorophyll A and B in the leaf pallisade cells in the upperside of the cell structure.
2. but as you move out in the NIR which is a slightly longer wavelength > you start to penetrate a little bit and scatters strongly off the spongy inside of the cells.
3. the chlorophyll absorption is taking place here - this is where the green color is coming out and is being revealed because of the strong absorption in the blue and red and the strong scattering of the NIR in the middle layer of the cells! cool!
4. Veg is super bright in the NIR b/c in the NIR, there is strong scattering in the bottom cells. 

increase! Because of the water itself

i.e. just think of a lawn -- if its wet and healthy, its dark, but if its dry and dead, its not. 

-From very dry at the top > as you add more water > the overall refelctance in the infrared channels start brightening and especially in the dips

Its really hard! 

-BUT if you could assume that things are not moving and you go back to the same place over time > then you could say that "while i dont know what plant is there, I can probably attribute any changes in the NIR to be a result of the water content"

• They really want to see what types of plants are there, not just moisture differences
• they want to map ecosystems, biomes, separate crops - without going to the field exclusively to do it
• Creating classifications 

• A classification is a gridded dataset where the digital number behind every pixel set is not a reflectance but rather a sign of each crop. 
• vegetation remote sensors like to work on their classificaitons - requires field validations
• veg classifications are hard with low spectral res satellite systems, but with AVIRIS it gets much easier.(hyperspectral!)

True! AVIRIS is hyperspectral - this allows you to start doing better classificaitons of veg type.

-the goal is to find the right band to measure and map vegetation; In these types of maps a DN is assigned to a crop type. 

Temporal (Phenological) Characteristics:

1. T/F: Plants have a different spectral response based on the time of year and changes in health.
2. T/F: you dont have to correct for the phenological changes in your spectral signature throughout the year. 
3. T/F: In non-water limited systems like wetlands, spectral responses dont vary that much.

1. True
2. False - you need to correct for these changes or else your results will be erroneous
3. False - even these systems have plants that have their own phenological changes/cycles. 

- Decrease! 

Temporal (phenological) Characteristics:

imagine that same field and keep the moisture content consistent > and furthermore, lets imagine plants have grown from seedlings up to big plants, covering up the dirt field > do you expect the red reflectance as viewed from above to increase, decrease, or stay the same?

Temporal (phenological) Characteristics:

1. As soil gets drier, the red reflectance will be a)_________; and the NIR will b)_________
2. As canopy closure occurs more and more, the red reflectance will c)________; and the NIR will d)____________

1. a) higher b) decrease
2. c) decrease d) increase

Temporal (phenological) Characteristics:

Describe what is happening in the chart with the distribution of reflectance valeus in BOTH the red and NIR for a spt on the ground as it goes through its phenological cycle

-(We looking at two processes: moisture and canopy closure)

• Looking at a bare field... as you moisten the field (go from dry to wet and wet to dry) > its going to get brighter in the infrared and NIR and darker in the red; back and forth along this line in response to moistre content
• as the canopy closes or opens (as the field grows) > the canopy is going to get brighter and brigher in the NIR/IR but darker in the red and vice versa. 
• The graph that these two processes are combined: lots moisture and canopy closure in spectral space for each pixel  - essentially it is showing a migration of two values in two different bands through time/a planting cycle.
• the graph encapsualtes a range of values that one can expct based on changes in phenology of plants and moisture content for the red and NIR bands.

In-Situ Measurements:

What is a Leaf Area Index (LAI) Ceptomenter and what does it do?

-A LAI Ceptometer measure the amount of canopy closure -- the amount of blockage of light create by a plant

-you stick your wand out under the sun and measure the sunlight; then stick it under the plants ; then measure the difference between the two

- the result is a funciton of the density of the canopy

In-Situ Measurements: 

What is the purpose of taking a Total Dryweight biomass measurement with respect to RS applications?

• SO what a scientist can do is go out, cut down some plants > weigh them > dry them > then weigh them again
• this gives you the actual biomass for a number of points within an image
• you can, after collecting a bunch of field validaiton points > create a regression relationship and apply a biomass weight to every pixel to spit out values
• this is the heart and soul of validation campaigns.
• Its an empirical model based on ground truth -- way more useful than reflectance alone
• the most important applicaiton is applying your model to areas that you cant necessarily get out too or cover in a lifetime.

Vegetation Indices:

 What is a Vegetation index?

What do they try to do (2)?

• The first thing they try to do:
• they are a standardized metric that takes the bands that you have and applies some band math to them
• essentially takes one band and divides it by another or a number of other combos
• its smart band math: where yo uchoose bands that are probably sensitive to different vegetaiotn properties which is most likely NIR and red.
• The other thing they try to do:
• they try to correct/normalize the effect of other stuff that affects the reflectance (i.e. how low the sun is in the sky, viewing angles, etc)
• things that they try to normalize out are things like atmospheric influences (i.e. is it cloudy -- if NIR is sensitive to moisture content, than the wetter it is, the lower the NIR)
• OR: if its really sunny out

Vegetation Indices:

With respect to vegetation indices...If its really sunny out, does your NIR reflectance increase, decrease, or stay the same? 

• Its going to increase (get brighter) b/c there is more light
• this is a huge problem b/c if you're trying to look at variations in brightness values
• vegetation indices try to get around this, they try to bypass this, but cant completely solve it

• A veg index is a dimensionless composition of reflectance (usually NIR and red) indicating abundance and/or health of green vegetation
• they are:
• 1. attempt to maximize sensitivity to vegetation parameters
• 2. attempt to reduce external effects by normalizing (i.e. sun/viewing angles, atmospheric influences; canopy background variations)

1. "The Simple Ratio" aka RVI
2. "Normalized Difference Vegetation Index" aka NDVI
3. "Moisture Vegetation Index" aka MSI
4. "Soil Adjusted Vegetation Index" aka SAVI

• aka RVI
• simple ratio of: infrared/red ratio veg index
• this is the first true veg index
• this takes advantage of the inverse relationship between chlorophyll absorption and red radiatn energy and increased reflectance of NIR energy for healtjy plant canopies
• capitalizes on both trends in a single number
• just a simple ratio
• a new map is created from the values of one band divide by the other
• so what you've done is leveraged opposing trends to make this super sensitive
• you're capitalizing on both trends in a single number

the Simple Vegetation Index:

If you have more plants, would you expect the NIR (numerator) to increase, decrease, or stay the same?

Increase! 

- so the more plants, the higher the numerator!

the Simple Vegetation Index:

If you have lots of plants, do you expect the red band (denominator) to increase, decrease, or stay the same?

the Simple Vegetation Index:

Does this accomodate for the brightness of the sun?

The Normalized Difference Vegetation Index:

What are the benefits?

• aka NDVI
• just one step up from the simple ratio
• (okin hates this!)
• 1. Simple
• 2. Low input: its useful because it doesnt have very big demands with regards to input; only needs two channels (you can use satellites from back in the 70s so you can build a time series from way back)
• 2.5. can be used with limited datasets

The Normalized Difference Vegetation Index:

Describe it:

• (NIR - red)/(NIR + Red)
• this takes the difference fothe two channels and divides it by the sum of both channels
• it is an attempt to normalize events
• i.e. so in a lush green field, you'd expect the NIR to increase and the red to decrease > SO: you're grwoing the numerator

The normalized Difference Vegetation Index:

What does dividing by the sum of both channels do?

• The reason for doing this is to do a better job for normalizing the illumination contrasts
• its not perfect, but it helps to divide by the sum of both channels bc on a bright day, both numbers will go up and the NDVi number will get corrected (suppressed a bit)
• on a dark day, this number will get small and puff up the number a bit

The normalized Difference Vegetation Index:

Summary of NDVI

• Has provided a method of estimating the net primary production over varying biome types
• identidying ecoregions
• monitoring phenologicalpatterns of the earth's surface
• and assessing the length of the growing season

The normalized Difference Vegetation Index:

At what scale does the NDVI work?

What is NDVI most interesting at looking at?

The normalized Difference Vegetation Index:

NDVI is a dimensionless index - T/F?

True: Its just a dimensionless index - ranging from -1 to 1

• 1 being very high NDVI - being very green
• -1 being very low NDVI - not green

The Moisture Vegetation Index:

what is unique about this?

• aka MSI
• It doesnt use the visible band at all -- just two NIR channels
• uses the NIR and midIR channels
• (NIR)/(midNIR) = MSI

The Moisture Vegetaion Index:

What satellite is thes based on?

In general, What do  the more sophisticated indices try to take into account?

what is the general trend with these indices?

• They try to take into account the soil properties in some way
• if any soil is shining through at all, the type of soil you have is going to make a big difference on the ratios and reflectance
• the trend with these indices is to try to make them more realistic and add terms to them that will get you more accurate representations of what is really going on

Soil Adjusted Vegetation Index (SAVI):

Describe this index

• recent indices have been developed to take advantage of improved hyperspectral ssytesm
• these improved indices incorporate:
• 1. a soil adjustment factor and/or 2. a blue band for atmospheric normalization.
• SAVI introduces L which is a soil calibration factor to the NDVI equation to minimize the soil background influences resulting from soil-plant spectral interactions. 
• SAVI looks like:
• [(1+L)(NIR - Red)]/(NIR + red + L)

The Soil adjusted Vegetation Index:

What does the L in SAVI do?

Case Study:

"Is climate warming enhancing warming-enhanced plant growth at high latitudes?"

Who did this study?