Home » Posts tagged 'remote sensing'
Tag Archives: remote sensing
With the most recent FAA UAV announcement my phone has been ringing with excited potential UAV users. Two points always comes up in the conversation. NDVI (normalized difference vegetation index) and image resolution. This blog will address the use of NDVI, resolution will come later. Before getting into the discussion, what NDVI is should be addressed. As described by Wikipedia, NDVI is a simple graphical indicator that can be used to analyze remote sensing measurements, typically but not necessarily from a space platform, and access whether the target being observed contains live green vegetation or not. NDVI is a mathematical function of the reflectance values of two wavelengths regions, near-infrared (NIR) and visable (commonly red).
The index NDVI has been tied to a great number of crop factors, the most important being biomass. Biomass being important as most things in the plant world impact biomass and biomass is related to yield. The most challenging issue with NDVI is it is highly correlated with biomass and a plants biomass is impacted by EVERYTHING!!!! Think about it, how many things can impact how a plant grows in a field.
The kicker that most do not know is that all NDVI’s values are not created equal. The source of the reflectance makes a big difference.
Measuring reflectance requires a light source, this is where the two forms of NDVI separate. Passive sensors measure reflectance using the sun (natural light) as a light source while active sensors measure the reflectance from a known light source (artificial light). The GreenSeeker is a good example of a active sensor, it emits its own light using LEDs in the sensor while satellite imagery is the classic passive sensor.
The challenge with passive remote sensing lies within the source of the light. Solar radiation and the amount of reflectance is impacted by atmospheric condition and sun angle to name a few things. That means without constant calibration, typically achieved through white plate measurements, the values are not consistent over time and space. This is the case whether the sensor is on a satellite or held held. In my research plots where I am collecting passive sensor data, so that I can measure all wavelength, I have found it necessary to collected a white plate calibration reading every 10 to 15 minutes of sensing. This is the only way I can remove the impacts of sun angle and cloud cover. When using the active sensors as long as the crop does not change the value is calibrated and repeatable.
What does this mean for those wanting to use NDVI collected from a passive sensor (satellite, plane, or UAV)? Not much if the user wants to distinguish or identify high biomass and low biomass areas. Passive NDVI is a great relative measurement for good and bad. However many who look at the measurements over time notice the values can change significantly from one day to the next. The best example I have for passive NDVI is a yield map with no legend. Even the magnitude of change between high and low is difficult to determine.
Passive NDVI in the hands of an agronomist or crop scout can be a great tool to identify zones of productivity. It becomes more complicated when decisions are made solely upon these values. One issue is this is a measure of plant biomass. It does nothing to tell us why the biomass production is different from one area to the next. That is why even with an active sensor OSU utilizes N-Rich Strips (N-Rich Strip Blog). The N-Rich Strip tells us if the difference is due to nitrogen or some other variable. We are also looking into utilizing P, K, and lime strips throughout fields. Again a good agronomist can utilize the passive NDVI data by directing sampling of the high and low biomass areas to identify the underling issues creating the differences.
OkState has been approached by many UAV companies to incorporate our nitrogen rate recommendation into their systems. This is an even greater challenge. Our sensor based nitrogen rate calculator (SBNRC blog) utilizes NDVI to predict yield based upon a model built over that last 20 years. That means to correctly work the NDVI must be calibrated and accurate to a minimum of 0.05 level (NDVI runs from 0.0 to 1.0). To date none have been able to provide a mechanism in which the NDVI could be calibrated well enough.
NDVI values collected with a passive sensor, regardless of the platform the sensor is on, has agronomic value. However its value is limited if the user is trying to make recommendations. As with any technology, to use NDVI you should have a goal in mind. It may be to identify zones or to make recommendations. Know the limitations of the technology, they all have limitations, and use the information accordingly.