Finishing resume for first job application.
Send letter to recommenders.
Working on method to derive LAI from hyperspectral. First idea is to extract shade using a linear unmixing method. This could be coupled with PROSAIL to derive a least squares estimation of LAI with inclusion of shade spectra (as identified from deep forest gaps).
Completed field LAI measurements using the new LAI-2000’s made available through Sue and Paul from IPIF. A total of 65 good points georeferenced to within 10 cm XYZ. Now I am using an offset IDL program to extract hyperspectral and LiDAR information for each pixel LAI location which I will input into a statistics program (JMP) to compare methods for LAI calculation, including simple vegetation indices and more advanced radiative transfer approaches. One of the main issues I am dealing with is how to minimize the shade resulting from canopy topography. Even ‘normalized’ indices, which use ratios to remove much of this effect, do not do a perfect job. First, I will look into how different spectral bands relate to canopy shade positions and see if it is possible to regress shade out of the spectra prior to running indices.
LAI variation 1-8 with all fixed parameters.
Dealing with a few floating underflow errors in prosail when sun or observer angle are greater than 85 degrees
Next step is to link this to a non-linear curve fitting code in order to model LAI out of our hyperspectral imagery.
http://www.physics.wisc.edu/~craigm/idl/
Of particular interest is his code:
MPFIT - Robust non-linear least squares curve fitting.
Example screenshot of PROSAIL module in IDL format.
coding offset calculations, then after converting differential stake lat long to utm coordinates will calculate utm zone 5 locations for all lai measurements collected so far (~40).
code snippet from preliminary version pre-testing
Laupahoehoe magnetic declination:
Today summary:
1. wake up, enter data, pack field gear
2. Set up LAI-2000 top of canopy unit, get all programmed and running correctly, wait for rain to stop and weather to be perfect.
3. Finally in the forest, all tripods set up and ready to collect data, then the unit stops working. I take it apart and realize nobody has ever changed the batteries (since 2008) and they have exploded in the interior..
4. Packed up all gear, drove to store bought new batteries, went home, cleaned unit, crossed fingers that it would turn on, and itd did. now its sitting in the dry box looking pretty and ready for field work trial #8 tomorrow…
Working on the LAI measurements. First step was to find an open location at the high elevation area, much easier than the low elevation much taller forest. For this we used a Suunto Inclinometer mounted on a tripod.
The low elevation required some more creativity to find an sufficiently open location, we’ll start collecting data with this system on Friday when we have finished the high elevation parcels.
Here is the remote open area LAI 2000 setup and collecting measurements every 15 seconds. I take basically the same system into the forest interior to collect LAI values, georeferenced using our Trupulse system.
Heading to field for LAI-2000 measurements. Today’s goal is to complete measurements on T1,2,3 and 4. We have 4 hours of 15 second remote LAI-2000 measurements possible in given the RAM capacity.
Required gear:
2 LAI-2000 units + wands;
1 TruPulse 360 for georeferencing + non-magnetic mounting bracket;
1. Tape for measuring height of Trupulse above differentially corrected stake;
3 tripods (one extra large) for mounting LAI-2000s and Trupulse;
2 reflectors for TruPulse;
1 Suunto Inclinometer for locating suitable location for remote top of canopy LAI-2000.
Day synopsis: well, went to the field in the morning and it was too sunny to collect LAI measurements, then went back in the afternoon and it was too rainy! Man, this equipment is way too picky for a tropical forest… Tomorrow I’m up and out of here at 5 am to take advantage of the increased diffuse radiation from 6-9, hopefully collect abundant LAI points!
Finished calibration of low and high elevation PAR sensors.
Calculated calibration coefficients for interior forest PAR sensors.
Glad to know that following re-calibration of sensors they have an average difference versus the LiCOR calibrated sensor of only 24 PPF with a max of 54 over the 3 hours of 1 minute data, covering 800-2400 PPF. Sweet!
Working to recalibrate top of canopy PAR sensors to tower mounted LiCOR sensor.
The high elevation is complicated. There is obvious diurnal and seasonal offsets to the sensor position (i.e., leans to the East and North, I think). I will model this out using julian day and fractional day.
The low elevation sensor is a much simpler situation requiring only a fractional change to the multiplier, maintaining the linear relationship between the two sensors. This one must be in a leveled position.
Completed the LAI-2000 protocol and test data looks good. Will start collecting forest data on Monday through Friday. The Apogee PAR sensors were calibrated versus the LiCOR and then sent back to Apogee for recalibration (250$..argg). Now starting to work on calibration of the top of canopy PAR sensors, dealing with an obvious leveling issue with the high elevation sensor which can be modeled out.
See LAI-2000 Protocol below:
Broadbent, E. 09112010 LAI-2000 Protocol, Remote X/Y Units
Collection of LAI values using two LAI-2000 units in remote X/Y mode
Basic:
1) Have 2 LAI-2000 units and two wands,
2) Make sure lens are clean and that protective covers are in place,
3) Make sure units power up and batteries seems adequate.
4) Avoid direct sunlight when collecting data. Collect on overcast days or close to sunrise or sunset. Direct sunlight can cause major errors in LAI (up to 50%).
Validate stored settings in control unit:
5) Power up both units (‘on’) and press ‘setup’,
6) FCT-01 & FCT-02: Start with factory calibrations on both units,
7) Above Canopy Unit X (PCH-0922): 4023, 1254, 1000, 1012, 1493
8) Below Canopy Unit Y (PCH-0916): 4040, 1256, 1000, 1011, 1393
9) FCT-04: Set resolution to high on both units,
10) FCT-05: Make sure date and time is exact same on both units (press enter to leave the same, xxx indicates invalid entry),
11) FCT-06: Check that distances are set to default values. 1.008, 1.087, 1.270, 1.662, 2.670. Different values would be calculated if foliage density was being obtained. These are mid-ring angles.
12) FCT-07: Check Ring angle values. Should be: 7, 23, 38, 53, 68.
13) FCT-08: Turn all channels to ‘off’ (unless LiCOR PAR sensors are connected).
14) Press ‘file’,
15) Select ‘Clear RAM’ (if all data has been downloaded!) on both units.
16) Verify using ‘memory status’ that no files are in RAM / Press ‘Break’ to exit to main screen,
17) FCT-13: Verify that wand log key is set to FCT-14.
18) FCT-16: set ‘bad readings’ to Beep and Ignore,
19) Turn off units using FCT-09.
Unit inter-calibration procedure:
20) Bring to open location close to field site.
21) Use sensor X control unit.
22) Attach sensor X (above canopy; PCH-0922) to X port.
23) Attach sensor Y (below canopy; PCH-0916) to Y port.
24) FCT-11: Set to ‘2 sensor’ mode to compare values / ave = 0,
25) Press ‘break’ and use up/down and left/right keys to select X1 & Y1 for compare.
26) Level and place side by side with exact same view.
27) FCT-55: Check that log buttons are functioning properly on both units.
28) FCT-03: Access ‘Comp X’
29) Move out of view of units.
30) Press ‘enter’ to calculate new X calibration values.
31) Verify that both units are measuring the same for all rings using up/down and left/right keys,
32) Press ‘edit’ in FCT-03 under X values to see new values. Record these in notebook.
33) FCT-01: to see X calibration defaults. Change these to use new calculations.
34) Change Y sensor to Y control unit.
35) Select ‘break’ and verify that corresponding rings read the same when sensors are side by side.
Set up above X unit:
36) FCT-11: Select ‘Rmt Below Y’
37) Select ‘avg’ to 0 (i.e., unknown).
38) FCT-12: Set up prompts (most likely not necessary).
Set up above X unit:
39) Set up A (reference unit) in large gap, open road or TOC position,
40) FCT-11: Select ‘Rmt Above X’.
41) Select ‘log interval’ of 15 seconds.
42) Enter ‘start’ and ‘stop’ times.
43) Verify there is enough memory available by checking ‘Room for’ # of files.
a. T/F = time to fill one file.
b. Num Files = How many files will be created.
c. Room for = How many files will fit in available RAM memory.
d. File 1 time: when the first file is started to be made.
44) FCT-14: Start logging (when start time is reached).
45) Enter descriptive location information (plot and person):
46) Verify taking LAI measurements as it will beep every 15 seconds.
Data collection:
47) Set up at point, level on tripod, georeference locations (Azi (mag; degrees), inc (degrees), SD (m). Make clear if location is from or to LAI-2000 from GPS stake.
48) FCT-14: For Plot: input traverse #, stake #, point #, height (L or H). i.e., T1S1P1H
49) Collect 5 points, one every 5 seconds, per location.
50) For each location collect a set of points at 1 m (H-HIGH) or 0.2 m (L-LOW).
51) Each location requires FCT-14 command again to collect new information. Press enter to log data point or button on wand (pressing log on keyboard starts a new log file).
Data collection (using the FV-2200 software):
1) Plug in unit X (A; above) to computer via USB RS-232 adaptor.
2) Verify that:
a. FCT-31: Baud=4800; Data bits=8; Parity=None; XON/XOFF = NO
b. FCT-33: Format: Standard; Print Obs: Yes
3) Turn on FV-2200 software and select acquire.
4) When prompted select: FCT-32 enter page: Start:1, End: 1000.
5) Open in new view named “LAI Elevation Traverse (s) Raw Date”, and save to .txt file in folder named as above in LAI-2000 data folder,
6) Repeat for unit Y (below), clearing print view between imports using ‘select all’ and ‘delete’, and opening in pre-existing view named as above, also saving as .txt file as above,
7) Strip all ? data points (using ‘strip’ button) from all files,
8) Select all B files and then select import, select A file, use interpolate A,
9) Recompute using band filters (leaving only 1 and 2), use A interpolation, and horizontal canopy model, and skip records with transmittance > 1,
10) Verify that each file looks acceptable, remove strange data points if found, and recompute if necessary,
11) Deselect all files and save as LAI-2000 format file with name as “LAI Elevation Traverse (s) Computed Date”,
12) Copy calculated final LAI values for each location and transfer to Excel with georeferencing coordinates, date, misc information, etc...
Notes misc:
Ø Sin(@)*height = 0.5*horizontal view necessary at that height.
Ø The distance from the sensor to the nearest leaf over it should be at least four times the leaf width.
o 1 ring = 0-13 degrees
o 2 rings = 16-28 degrees
o 3 rings = 32-43 degrees
o 4 rings = 47-58 degrees
Ø Use 2 rings minimum for each unit.
Ø LAI is dimensionless but can be thought of as {m2 leaf area (one sided) / m2 ground area}
Ø Press up arrow to enter letters
Ø See page 4-8 of the manual to get an estimate equation of the ground area covered by the sensor at different ring angles. A=f(pi)H^2
Ø Compute foliage density? Using 5-3, changing the dist factor for the first two rings. The dist factor would be obtained from the lidar data.
Ø If the unit locks up open the back case (4 screws) and unplug for 10 seconds. If this fails you can do a hard reset which will delete all data stored located on the processor board (approximately under the setup key).
Ø FV-2200 software is freely available for download for LiCOR ftp site, search their website to identify path location.
TOC Sensor calibration.
1) All dates converted to Julian style day and fraction day.
Trends in high elev sensor versus JD and FD (fractional day).
Calculate the new multipliers linear relationship with JD and FD.
New high elevation PAR values after fractional day noise removed.
Next step is to pull out any seasonal (NS offset) differences. Then finalize the calibration and apply to all high elevation PAR data.
Original data relationship looked like this (1 to 1):
Following diurnal correction no significant seasonal correction is found, therefore only the diurnal correction will be used. This correction is:
LiCOR calibrated PAR = (1.3974048 - 0.8695949 * :Fraction_Day) * Raw High PAR
PAR calibration ongoing. Setup measures every 30 seconds in high resolution, the Apogee sensors are single ended measurements while the LiCOR is differential. Measurements are within range, but updating the multipliers will definitely be necessary.
Pre correction mid elevation tower PAR sensor data. Post-correction the values are perfect.
Replaced HOBO Robot 4 with new one.
temp/rh calibration good
insanity for par calibration. the 4 apogee sensors were too far off. we’ve developed a calibration method versus a new LiCOR par sensor which is set up and will be run tomorrow, then sending the apogee units back to check for any hardware malfunctions.
we’ve also developed a method to calibrate all the top of canopy data as well as interior forest par sensors to new LiCOR par sensors recently calibrated. phew…a mind-boggle to work through!
preliminary calibration work in the morning prior to discovering that something was seriously not right…..
