| Author | Message | ||
     Jim Burger (jburger) Member Username: jburger Post Number: 12 Registered: 11-2010 |
Hi Francisco, I will emphasize what others have already stated: the remote sensing community (both satellite and airborne) has been troubled by this for decades. Compared to flickering lights/power supplies in the lab, the sun is an extremely stable light source - it's the atmosphere that causes problems. Do a search for 'atmospheric corrections'. You have the advantage of having high resolution spectra - often remote sensing has only 15 - 20 broad bands. Some other issues that will cause spectral variations: Geographic orientation - a spectrum of the same material taken from the North, will look substantially different than one from the South. In addition to atmospheric humidity, water content of the plant will create differences. Where in the plant you measure - i.e. close to stem or leaf edge. And finally remember you are using a spot device that averages reflection from a large area - make sure ALL of the area is target of interest, not for example a leaf edge with 'ground' as 30% background, or even other plant material at another distance. Best wishes, Jim | ||
| Francisco Garcia (ssisscu) Junior Member Username: ssisscu Post Number: 7 Registered: 11-2011 |
Dear all, Thanks for all your inputs, quite interesting to see how many things I need to take care of ;) Brian, I am using this information in order to build a model that can separate two species by using in field spectra. At the same time a reduction of the whole set of features will be performed to select the most prominent ones that can help in their separation. In the future it could be used for analyzing remote sensed images, but it is not the main purpose right now. I think that the first thing to do is to try to shadow my plants when there are no uniform clouds that can diffuse the light. With this I can avoid problems due to the low solar angle here in Denmark (around 50� at this time of the year). Of course the spectra will be pre-processed before any analysis in order to remove baseline and slope effects. Brian, can I ask what is the main difference that you state in this question? "Will these spectra be used to analyze a remote sensing image or will the spectra be used to build a chemometric model or ...?" Thanks again, FRAN | ||
| Brian Curtiss (briancurtiss) New member Username: briancurtiss Post Number: 5 Registered: 6-2006 |
There is no question that a spectrum of a plant canopy will vary both with changes in the angular distribution of illumination as well as the viewing geometry. The impact on you study will depend on how you are using the spectra. Will these spectra be used to analyze a remote sensing image or will the spectra be used to build a chemometric model or ...? | ||
| Peter Tillmann (tillmann) Senior Member Username: tillmann Post Number: 29 Registered: 11-2001 |
Tony, sorry, my mistake. The scales are different: absoption and reflectance. Peter | ||
| Tony Davies (td) Moderator Username: td Post Number: 284 Registered: 1-2001 |
Peter, Have the "passive" spectra been referenced? The scales look very different. The satellite scanning community has done a lot of work on cloud and water vapour interference; perhaps that would be a place to look. best wishes, Tony | ||
| Peter Tillmann (tillmann) Senior Member Username: tillmann Post Number: 28 Registered: 11-2001 |
Tony, the water bands will not be "cancelled out" by an appropriate reference. There is only noise to watch. See the attached picture "handy.jpg" taken with reference from sunlight (passive measurement haad) and "handy2.jpg" taken with a 'regular measurement head' (with light source and measuring on contact). Same samples, same day. "Passive"  "Active"  | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 484 Registered: 9-2001 |
Francisco - one thought I had was for you to contact the meteorology department at a nearby (or several) universities. After all, who studies clouds and their behavior more than they do? Another possibility would be to not wait for a clear, cloudless day, but collect some data as often as you can manage. That would give you information about how much variation of energy there is under various different conditions, and the amount of variation under some specified conditions, that might turn out to be reproducible. Howard \o/ /_\ | ||
| Tony Davies (td) Moderator Username: td Post Number: 283 Registered: 1-2001 |
Hello Peter, I decided not to mention the water vapour noise because I think most of it will be cancelled by taking references. A partially cloudy day with rapid changes in light levels could be quite difficult. Best wishes, Tony | ||
| Peter Tillmann (tillmann) Senior Member Username: tillmann Post Number: 27 Registered: 11-2001 |
Dear Francisco, I assume you are using no light source except the sun. A "passive" measurement head. 1. The water in the air will absorb all energy at the appropriate wavelengths/wavenumbers. The system will therefore produce noise for these wavelengths only. 2. For the other wavelengths you will reference the system by the white calibration plate. The moment you will change the irradiation from the sun on the samples you will need a new reference. 3. Distances are crucial. Peter | ||
| Tony Davies (td) Moderator Username: td Post Number: 282 Registered: 1-2001 |
Hello Fran, Yes, Light from cloud is highly scattered. Given your location it might be best to avoid direct illumination. You could try shading your plants on those occasions when cloud is absent (Quite rare here in the UK -the land of the wettest drought!). Of course there will be variation from the thickness of the cloud but I think your standardisation will take care of most of this. The expert on scattering is Don Dahm and he might come in when he wakes up in the States. [Come in Don]. best wishes, Tony | ||
| Francisco Garcia (ssisscu) Junior Member Username: ssisscu Post Number: 6 Registered: 11-2011 |
Dear all, I am conducting some tests using a portable spectroradiometer in order to collect the spectra of several plants under field conditions. The spectroradiometer has a 4� lens mounted on it, and it is between 10 to 20 cm from the target. It gives me a sampling spot of about 1 cm2. I use a white coat for calibration and the reflectance of the target is relative to the radiance of the white calibration plate. I try to take the measurements when the day is clear sky without clouds and around noon. I am in Denmark and this ideal situation is given once in a while. The most common situation is a sky covered by isolated clouds that I assume might scatter a lot of the incoming light. I try to take reference spectra quite often in order to control the variations of incoming light. I would like to know how much can this situation affect the repeatability of my measurements from one day to another. And if i should allways aim for a sunny clear day (which is quite difficult here in Denmark) or I can take data in all other conditions (scattered clouds and completelly covered sky). Anyone can give me a recommendation in terms of sky conditions in order to have reliable readings? Thanks in advanced. Fran |