Finally completed the image subset procedure. Sweet. This is the only way to deal with multiple 8 GB LiDAR images. The I/O is also rapid, good for running through spatio-temporal models.
mapping out the microclimate model flowchart, here is the version as is showing individual IDL components, most have been started, highlighted yellow ones have been completed.
Job applications and IDL programming. Working on the final program to read in small memory management sections of the 8 GB LiDAR image. Almost complete, just some final details with interleave formatting.
Car all packed and ready to go.
but it didn’t start….the starter was busted. had to have it towed to hilo for repairs.
Fortunately they managed to repair it within one day and by Saturday afternoon I was back in business and ready for field work on Monday!
Rebuilding the sensors following calibration.
Equipment ready to pack in the car.
In the field in the rain.
1 Sensor installed in gap on a small traverse. 2 others in the canopy. 1 more in the forest interior.
Coiling some of our 100s of meters of cable…so many crazy tangles today!
Sensors finally all up! Just as the sun is setting, at least the rain let up for us! Uploading the final programs and verifying the data collection. All is well, time to head home.
So the first approach didn’t work. The sensors were all over the place. This was due to the water temperature never stabilizing and different sections of the tub having different depths.
I resolved this using 60 lbs of ice and deeper water depth and removing the wooden supports to fit them all in the deep end.
Following a period of crazy readings while the ice cooled and the water stabilized, I obtained decent calibration points (from 13:27 on).
Calibration offsets calculated were very small.
not what i intended to do today, but here it is: my near final LAI map of my study transect..not an easy task! (due to field collection, and then serious sun/canopy topography cosine noise, fortunately was able to model it out and derive this…
400-1050 nm reflectance directional. Two lines, input and modeled are shown (hard to tell there are 2).
Changes in leaf angle cause major errors in the LAI estimation. This would need to be held constant if used ultimately. I will try to code an EVI at multiple spatial scales extractor, this should deal with the issues effecting the NDVI relationship…hopefully.
relationship for LAI of 8..good, with slight difference in low nm area. Input LAI of 8, output of 6.5. could be improved, but saturation is known to occur at high LAIs (leaf area index) for all vegetation indices so also would be happening within our hyperspectral dataset.
Here I used the pixel purity feature in ENVI to identify endmember shade fractions (green) versus an example of full sun canopy direction reflectance. The red curve would be modeled from PROSAIL while the green (shade) would remain as is (the average of multiple pure shade pixels). The issue with this approach is the reduction in the signal/noise ratio with increasing shade levels until parameterization becomes meaningless…
