Application of Remote Sensing Techniques in the Long Term Ecological Research

Organizer: Wei Wu -- SUNY-ESF, Luquillo LTER

September 21, 2003

The workshop has one presentation by Wei Wu (Spatial modeling of the probability of clouds and evapotranspiration in the LUQ using TM data). Other representatives come from ARC, BNZ, LUQ, MCM, University of Washington, and research institute for humanity and nature in Japan. The workshop not only discussed some technique issues related with application of remote sensing data but also management of remote sensing data in the LTER network.

During the discussion, we focused on the following questions:

1) What projects are going on that involve the application of remote sensing data in their sites?

The availability of satellite and aerial photography time-series data offer ecologists new opportunities to examine scale dependent relationships across time and space. We shared our experience to derive environmental variables and model ecological phenomena using remote sensing data. The research going on in different sites are related with monitoring soil moisture, vegetation (LAI) and algae change, classifying land use types, modeling glacier retreating, evapotranspiration and probability of cloud cover from RADAR, aerial photos and TM images.

2) What satellite images or aerial photos have high potential to assist long term ecological research?

We realized there is always trade off between spectral, temporal and spatial resolution. Like TM image, it has high spatial resolution (30 m* 30 m) but low temporal resolution (one scene every nine days), while MODIS has high temporal resolution (two scenes a day) and low spatial resolution (250 m, 1km). We must choose the suitable remote sensing data according to the research objective. We focused on the application of the data obtained from Ikonos satellite and hyperion sensor on EO-1 satellite. Ikonos is the first of the next generation of high spatial resolution satellites. Ikonos data records 4 channels of multispectral data at 4 metre resolution and one panchromatic channel with 1 metre resolution.  This means that Ikonos was the first commercial satellite to deliver near photographic high resolution satellite imagery of anywhere in the world. The customers can ask about data for specific area. Ikonos will guarantee to provide image with cloud cover less than 10% for the area of interest. It is very useful in land use classification, vegetation evaluation, updating existing maps, managing storm water runoff, exploration natural resource, etc. The Hyperion sensor onboard the satellite is the first hyperspectral sensor on an Earth observation satellite. It covers the complete spectral range from 0.4 to 2.5 µm in 220 bands. Such comprehensive spectral resolution permits very detailed land cover classifications or identifications to be performed.

We also discussed the importance of incorporating the remote sensing data into computer models to get more accurate and spatial results.

3) Some issues related with validation

It is very important that we get field data to validate the derived ecological variables from remote sensing data. Validation involves sampling techniques (random, systematic, stratified), scaling up (scalar or geostatistic technique) and accuracy assessment (user’s, producer’s and overall accuracy, Kappa index). We recognized LTER network provides us an opportunity to do cross-site comparison (a type of scale-up).

4) Data sharing

The limitation of propagating the use of remote sensing data is that they are usually expensive to purchase. Not every LTER site has a helicopter like in Toolik Lake so that the scientists could decide the spatial resolution and when they are going to obtain the images. It is not feasible to ask the network office to purchase specific remote sensing image since different scientists have different requirements on time, space and spatial and spectral resolution. We need a good mechanism to manage remote sensing data to make best use of the money we spent on purchasing. We discussed two ways: establish a pool of money for each site so the scientists of each site can ask the information manager of their own site to purchase remote sensing images and the images (.jpg files) may be put on the site webpage; in order to obtain the data, a request form must be filled out. An alternative is that the scientists purchase the scenes and make them online via network office so they could be available to the other scientists and students, of course, when the papers based on the data get published, the authors need to cite the source of the data and give acknowledgement for that.

Added notes:

The scientists who want to incorporate remote sensing data in their research seldom go to the website of LTER network office that focuses on network-wide data - http://www.lternet.edu/technology/sdw. They usually go to their data manager of their sites to obtain images.