|
Reflectance modeling results
|
|
Our native and nonnative species are separable according to the analysis. How is separation possible? If you compare the two lines on the graph above ("reflectance spectra"), you can see that strawberry guava reflected more light than ‘ōhi‘a. Notice that the difference between the two species is greater in some regions of the electromagnetic curve than in others. Visible light is 400 - 700 nm, near infrared is about 700 - 900 nm, shortwave infrared starts at about 1400 nm and goes to the end of the graph. These differences in reflection aid in our ability to find certain species. To see the results for all six species, click here. To see the regions of the electromagnetic curve, click here. |
|
The graph above shows the accuracy that two sensors have in detecting a target species in a simulated forest containing the other five species. Strawberry guava detection has the highest accuracy. This means that if a forest is composed of only koa, lama, ‘ōhi‘a, coffee, strawberry guava and christmas berry, satellite data can separate the strawberry guava with the greatest accuracy. This implies that we should be able to use remote sensing to map strawberry guava with very high accuracy. Note that all detection accuracies are above 50%. There is quite a difference in accuracy between the Landsat ETM+ sensor and the IKONOS sensor. In fact, the Landsat ETM+ is more accurate at detecting all species than the IKONOS. Why might this be true? To see the results for other sensors, click here. |
|
Our analysis shows that certain sensors are more accurate than others. The overall detection accuracy for Landsat ETM+ and IKONOS are given in the figure above. In general, multispectral sensors that collect shortwave infrared data (like Landsat ETM+) have higher accuracies than sensors that don't (like IKONOS). Remember that the model we used is one possible simplification of the real world. For example, it only analyzes reflectance (spectral information), so things like the sensor's spatial resolution are not taken into account. To see the results for other sensors, click here. To see the characteristics of our studied sensors, click here. |
|
