EO with python

Question 1

Why learn python?

Answer 1

for new algorithms or data formats that don’t exist yet, if files are too large for the GUI based system to hold at once (code can carefully manage memory) and it lets us manage many files (looping command line functions over input files)

Question 2

what is python?

Answer 2

a “high level” language written for “hacking”
+ Many of the hardware controls are hidden from you, but can be called if the user wants.
* It is an “interpreted language”
* Slower to run than lower level languages.

Question 3

Why are we in a data age?

Answer 3

More sensors generating more data than ever before, more of it is freely available
* New techniques to make use of that data, informing decisions and making new discoveries
this is why we require programming skills

Question 4

benefits of using python in the data age

Answer 4

• Test new ideas
• Produce custom applications
• Not have to pay expensive license fees

Question 5

How many NASA product levels are there?

Answer 5

5

Question 6

Level 0

Answer 6

direct from the sensor, rawest data, usually encode file describing voltages (only for mission team)
couldn’t load into Arc

Question 7

Level 1

Answer 7

Raw data
Has been encoded but still in terms of instrument voltage read out.
Can be used by the science community, not easy.
Colour distortion, clouds

Question 8

Level 2

Answer 8

derived geophysical variables as seen by the satellite.
After algorithms have been applied to level 1 instrument data to estimate land surface properties (surface reflectance, NDVI, DTM, DSM, CHM, backscatter, elevation, etc… Derived for the swath of the satellite.

Question 9

Level 3

Answer 9

level 2 data that has been gridded.
Swath of the satellite has been gridded onto some common raster layer.

Question 10

Level 4

Answer 10

modelled product from level 2 or 3. This is when models estimate properties the satellite is not directly sensitive to. Ie. landcover class, biomass

Question 11

How do you estimate biomass

Answer 11

Alometric equations using Diameter Breast Height (DBH), tree height and species. These are equations people have made for estimating biomass for a wide range of tree species.

Question 12

What is needed if you want to upscale from field plots to remotely sensed data?

Answer 12

To upscale, we must have data covering a large area which is sensitive to our
parameter of interest

Question 13

Definition of sensitivity in the context of upscaling biomass estimations

Answer 13

• The variable changes at a predictable rate with the variable of interest
• It can “explain the variance” of our parameter of interest

Question 14

What is sensitive to biomass?

Answer 14

• Tree diameter (total in a plot is the “basal area”)
• Tree height
• Forest density (canopy cover)
• A combination of the two

Question 15

What remote sensing method?

Answer 15

Passive optical - density through measuring greenness
Lidar - tree height and cover
Radar - backscatter and decoherence could be used

Question 16

If you have height from remote sensing, how do you get biomass?

Answer 16

prelate height to biomass (measured in the field). Start by testing to see if there is a relationship by plotting two datasets together (either strong positive, strong negative or no clear relationship)

Question 17

what correlations coefficient assesses strength of correlation?

Answer 17

pearson’s (-1 to +1)

Question 18

what is a good positive correlation?

Answer 18

more than or equal to 0.6

Question 19

List the techniques used to determine which lidar derived metrics are sensitive to biomass

Answer 19

1. Height of Median Energy (HOME), also know as relative height 50 (RH50) - the height at which half of the returns are above and below - created for full waveform lidar (link in one note)
2. mean canopy height MCH - within a pixel
   for discrete return data link in one note

Question 20

What is ABGD

Answer 20

Above Ground Biomass Density

Question 21

What does ABGD corelate to well??

Answer 21

ABGD

Question 22

equation for Canopy Height Map (CHM)

Answer 22

CHM=DSM-DTM
then average values over the area of interest

Question 23

How did the luton Lidar metric for biomass work?

Answer 23

• Create a CHM
• Lay over a coarser-resolution grid
• Calculate mean CHM per coarse grid-cell
• Relate MCH to plot level biomass (ground data)
• Compare MCH to plot biomass
• Apply model to entire remotely sensed dataset to produce a map

Question 24

what happens to RH50 with taller trees

Answer 24

it increases

Question 25

what happens to RH50 with denser forest

Answer 25

it increases

Question 26

how does cover influence MCH

Answer 26

lower cover, more zeroes, bring down MCH

Question 27

In Liar you can download data from Scottish government in 3 forms

Answer 27

point cloud (raw data)
DTM and DSM

Question 28

What needs to be considered when data collecting in the field (ground calibration considerations)?

Answer 28

cover useful lidar and biomass metrics (highs, mediums and lows), collect sufficient data to create a reliable model and estimate accuracy/ uncertainty of the product

Question 29

How do you guarantee good ground calibration is achieved

Answer 29

We can use statistical techniques to make sure this happens - random stratified sampling - picking high medium and low forests - 5 plots in each - ideally add plots until error metric changes but that is time consuming…