Using VHR NICFI Imagery

Norway’s International Climate and Forest Initiative (NICFI) provides very-high resolution (VHR) imagery for tropical areas around the world. While VHR imagery is a useful tool, this does not necessarily mean it is the best fit for a given analysis goal. I will be assessing whether NICFI imagery is a good fit for examining my previous study area in Cameroon.

Figure 1. Landsat 8 and NICFI imagery of a forested area. In the Landsat 8 imagery, the area appears to have a consistent tree density, but the NICFI imagery shows variation in density and cover.

As imagery taken to research forest cover, it is no surprise that forests are salient features on the NICFI imagery. Figure 1 shows a comparison between NICFI and Landsat 8 imagery from the same month (but different years). The high spatial resolution of NICFI imagery compared to Landsat 8 and Sentinel 2 imagery make borders between land cover types more discernible. Urban areas are also easier to identify on NICFI imagery due to the high spatial resolution, as seen in the comparison in Figure 2. 

Figure 2. Landsat 8 and NICFI imagery of an area with mixed land cover, including a central urban area. The urban area appears lighter or “shinier” in the NICFI image, which is more clearly identifiable as urban than the flat Landsat 8 image.

Another way to measure forest cover is with normalized indices, such as NDVI. However, the NICFI imagery only provides four bands (R,G,B, and NIR) while both Landsat 8 and Sentinel 2 imagery have additional NIR and SWIR bands which can be used to calculate more indices and visualize different spectral characteristics of the images.

Figure 3. A NICFI image that allows for examination of Lake Maga but has a gap in the study area.

These images were both taken in August but from different years. NICFI provides imagery by month or biannual composites, but it is hard to search for specific images without choosing a specific month. The NICFI holds less attainable information in each pixel (with the only property being “cadence”, or when the image was taken) and thus cannot be filtered by other factors such as cloud cover or location. NICFI images have already been processed to remove cloud cover and atmospheric interference, but there are still gaps and imperfections in the imagery. An example of gaps is shown in Figure 3 where in order to have an image that shows Lake Maga, visibility of the study area has to be sacrificed, whereas in Figure 4, there is an anomaly over Lake Maga but good visibility within the study area. 

Figure 4. An anomaly over Lake Maga, perhaps due to incomplete cloud removal before the creation of the basemap composite.

NICFI imagery is a good option for analyses of tree cover where monthly intervals of imagery are sufficient. For example, NICFI imagery could be used to expand the Hansen’s data set of forest cover or map agreement with the Hansen’s data set (as was done in the supervised classification of forest cover in Cameroon). NICFI may not be helpful for multi-class land cover evaluation using a variety of indices. The spatial extent of the NICFI basemap eliminates the need to mosaic together multiple images for a large region of study, but gaps and anomalies still exist in this imagery.

Code for the comparison of NICFI, Landsat 8, and Sentinel 2 images: https://code.earthengine.google.com/149b6ce2862a40d1d030770d3c52c809