Creating a Raster Stack from Hyperspectral Imagery in HDF5 Format in R

In this tutorial, we will learn how to create multi (3) band images from hyperspectral data. We will also learn how to perform some basic raster calculations (known as raster math in the GIS world).

About Hyperspectral Data

We often want to generate a 3 band image from multi or hyperspectral data. The most commonly recognized band combination is RGB which stands for Red, Green and Blue. RGB images are just like the images that your camera takes. But there are other band combinations that are useful too. For example, near infrared images emphasize vegetation and help us classify or identify where vegetation is located on the ground.

A portion of the SJER field site using red, green and blue (example dataset bands 14,9,4; bands 58,34,19 in the full NEON dataset). Here is the same section of SJER but with other bands highlighted to create a colored infrared image – near infrared, green and blue (example dataset bands 22, 9, 4; bands 90, 34, 19 in the full NEON dataset).

Create a Raster Stack in R

In the previous activity, we exported a single band of the NEON Reflectance data from a HDF5 file. In this activity, we will create a full color image using 3 (red, green and blue - RGB) bands. We will follow many of the steps we followed in the Intro to Working with Hyperspectral Remote Sensing Data in HDF5 Format in R tutorial. These steps included loading required packages, reading in our file and viewing the file structure.

# Load required packages
library(raster)
library(rhdf5)

# set working directory to ensure R can find the file we wish to import and where
# we want to save our files. Be sure to move the download into your working directory!
wd <- "~/Documents/data/" # This will depend on your local environment
setwd(wd)

# create path to file name
f <- paste0(wd,"NEON_hyperspectral_tutorial_example_subset.h5")


# View HDF5 file structure 
View(h5ls(f,all=T))

To spatially locate our raster data, we need a few key attributes:

• The coordinate reference system
• The spatial extent of the raster

We'll begin by grabbing these key attributes from the H5 file.

# define coordinate reference system from the EPSG code provided in the HDF5 file
myEPSG <- h5read(f,"/SJER/Reflectance/Metadata/Coordinate_System/EPSG Code" )
myCRS <- crs(paste0("+init=epsg:",myEPSG))

# get the Reflectance_Data attributes
reflInfo <- h5readAttributes(f,"/SJER/Reflectance/Reflectance_Data" )

# Grab the UTM coordinates of the spatial extent
xMin <- reflInfo$Spatial_Extent_meters[1]
xMax <- reflInfo$Spatial_Extent_meters[2]
yMin <- reflInfo$Spatial_Extent_meters[3]
yMax <- reflInfo$Spatial_Extent_meters[4]

# define the extent (left, right, top, bottom)
rasExt <- extent(xMin,xMax,yMin,yMax)

# view the extent to make sure that it looks right
rasExt

## class      : Extent 
## xmin       : 257500 
## xmax       : 258000 
## ymin       : 4112500 
## ymax       : 4113000

# Finally, define the no data value for later
myNoDataValue <- as.integer(reflInfo$Data_Ignore_Value)
myNoDataValue

## [1] -9999

Next, we'll write a function that will perform the processing that we did step by step in the Intro to Working with Hyperspectral Remote Sensing Data in HDF5 Format in R. This will allow us to process multiple bands in bulk.

The function band2Rast slices a band of data from the HDF5 file, and extracts the reflectance. It them converts the data to a matrix, converts it to a raster and returns a spatially corrected raster for the specified band.

The function requires the following variables:

• file: the file
• band: the band number we wish to extract
• noDataValue: the noDataValue for the raster
• extent: a raster Extent object .
• crs: the Coordinate Reference System for the raster

The function output is a spatially referenced, R raster object.

# file: the hdf file
# band: the band you want to process
# returns: a matrix containing the reflectance data for the specific band

band2Raster <- function(file, band, noDataValue, extent, CRS){
    # first, read in the raster
    out <- h5read(file,"/SJER/Reflectance/Reflectance_Data",index=list(band,NULL,NULL))
	  # Convert from array to matrix
	  out <- (out[1,,])
	  # transpose data to fix flipped row and column order 
    # depending upon how your data are formatted you might not have to perform this
    # step.
	  out <- t(out)
    # assign data ignore values to NA
    # note, you might chose to assign values of 15000 to NA
    out[out == myNoDataValue] <- NA
	  
    # turn the out object into a raster
    outr <- raster(out,crs=CRS)
   
    # assign the extents to the raster
    extent(outr) <- extent
   
    # return the raster object
    return(outr)
}

Now that the function is created, we can create our list of rasters. The list specifies which bands (or dimensions in our hyperspectral dataset) we want to include in our raster stack. Let's start with a typical RGB (red, green, blue) combination. We will use bands 14, 9, and 4 (bands 58, 34, and 19 in a full NEON hyperspectral dataset).

# create a list of the bands we want in our stack
rgb <- list(14,9,4) #list(58,34,19) when using full NEON hyperspectral dataset

# lapply tells R to apply the function to each element in the list
rgb_rast <- lapply(rgb,FUN=band2Raster, file = f,
                   noDataValue=myNoDataValue, 
                   extent=rasExt,
                   CRS=myCRS)

# check out the properties or rgb_rast
# note that it displays properties of 3 rasters.
rgb_rast

## [[1]]
## class      : RasterLayer 
## dimensions : 500, 500, 250000  (nrow, ncol, ncell)
## resolution : 1, 1  (x, y)
## extent     : 257500, 258000, 4112500, 4113000  (xmin, xmax, ymin, ymax)
## crs        : +init=epsg:32611 +proj=utm +zone=11 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0 
## source     : memory
## names      : layer 
## values     : 0, 9418  (min, max)
## 
## 
## [[2]]
## class      : RasterLayer 
## dimensions : 500, 500, 250000  (nrow, ncol, ncell)
## resolution : 1, 1  (x, y)
## extent     : 257500, 258000, 4112500, 4113000  (xmin, xmax, ymin, ymax)
## crs        : +init=epsg:32611 +proj=utm +zone=11 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0 
## source     : memory
## names      : layer 
## values     : 0, 9210  (min, max)
## 
## 
## [[3]]
## class      : RasterLayer 
## dimensions : 500, 500, 250000  (nrow, ncol, ncell)
## resolution : 1, 1  (x, y)
## extent     : 257500, 258000, 4112500, 4113000  (xmin, xmax, ymin, ymax)
## crs        : +init=epsg:32611 +proj=utm +zone=11 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0 
## source     : memory
## names      : layer 
## values     : 0, 9704  (min, max)

# finally, create a raster stack from our list of rasters
rgbStack <- stack(rgb_rast)

In the code chunk above, we used the lapply() function, which is a powerful, flexible way to apply a function (in this case, our band2Raster() fucntion) multiple times. You can learn more about lapply() here.

NOTE: We are using the raster stack object in R to store several rasters that are of the same CRS and extent. This is a popular and convenient way to organize co-incident rasters.

Next, add the names of the bands to the raster so we can easily keep track of the bands in the list.

# Create a list of band names
bandNames <- paste("Band_",unlist(rgb),sep="")

# set the rasterStack's names equal to the list of bandNames created above
names(rgbStack) <- bandNames

# check properties of the raster list - note the band names
rgbStack

## class      : RasterStack 
## dimensions : 500, 500, 250000, 3  (nrow, ncol, ncell, nlayers)
## resolution : 1, 1  (x, y)
## extent     : 257500, 258000, 4112500, 4113000  (xmin, xmax, ymin, ymax)
## crs        : +init=epsg:32611 +proj=utm +zone=11 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0 
## names      : Band_14, Band_9, Band_4 
## min values :       0,      0,      0 
## max values :    9418,   9210,   9704

# scale the data as specified in the reflInfo$Scale Factor
rgbStack <- rgbStack/as.integer(reflInfo$Scale_Factor)

# plot one raster in the stack to make sure things look OK.
plot(rgbStack$Band_14, main="Band 14")

We can play with the color ramps too if we want:

# change the colors of our raster 
myCol <- terrain.colors(25)
image(rgbStack$Band_14, main="Band 14", col=myCol)

# adjust the zlims or the stretch of the image
myCol <- terrain.colors(25)
image(rgbStack$Band_14, main="Band 14", col=myCol, zlim = c(0,.5))

# try a different color palette
myCol <- topo.colors(15, alpha = 1)
image(rgbStack$Band_14, main="Band 14", col=myCol, zlim=c(0,.5))

The plotRGB function allows you to combine three bands to create an image.

# create a 3 band RGB image
plotRGB(rgbStack,
        r=1,g=2,b=3,
        stretch = "lin")

A note about image stretching: Notice that we use the argument stretch="lin" in this plotting function, which automatically stretches the brightness values for us to produce a natural-looking image.

Once you've created your raster, you can export it as a GeoTIFF. You can bring this GeoTIFF into any GIS program!

# write out final raster	
# note: if you set overwrite to TRUE, then you will overwite or lose the older
# version of the tif file! Keep this in mind.
writeRaster(rgbStack, file=paste0(wd,"NEON_hyperspectral_tutorial_example_RGB_stack_image.tif"), format="GTiff", overwrite=TRUE)

Raster Math - Creating NDVI and other Vegetation Indices in R

In this last part, we will calculate some vegetation indices using raster math in R! We will start by creating NDVI or Normalized Difference Vegetation Index.

About NDVI

NDVI is a ratio between the near infrared (NIR) portion of the electromagnetic spectrum and the red portion of the spectrum. Please keep in mind that there are different ways to aggregate bands when using hyperspectral data. This example is using individual bands to perform the NDVI calculation. Using individual bands is not necessarily the best way to calculate NDVI from hyperspectral data!

# Calculate NDVI
# select bands to use in calculation (red, NIR)
ndvi_bands <- c(16,24) #bands c(58,90) in full NEON hyperspectral dataset

# create raster list and then a stack using those two bands
ndvi_rast <- lapply(ndvi_bands,FUN=band2Raster, file = f,
                   noDataValue=myNoDataValue, 
                   extent=rasExt, CRS=myCRS)
ndvi_stack <- stack(ndvi_rast)

# make the names pretty
bandNDVINames <- paste("Band_",unlist(ndvi_bands),sep="")
names(ndvi_stack) <- bandNDVINames

# view the properties of the new raster stack
ndvi_stack

## class      : RasterStack 
## dimensions : 500, 500, 250000, 2  (nrow, ncol, ncell, nlayers)
## resolution : 1, 1  (x, y)
## extent     : 257500, 258000, 4112500, 4113000  (xmin, xmax, ymin, ymax)
## crs        : +init=epsg:32611 +proj=utm +zone=11 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0 
## names      : Band_16, Band_24 
## min values :       0,       0 
## max values :    9386,    9424

#calculate NDVI
NDVI <- function(x) {
	  (x[,2]-x[,1])/(x[,2]+x[,1])
}
ndvi_calc <- calc(ndvi_stack,NDVI)
plot(ndvi_calc, main="NDVI for the NEON SJER Field Site")

# Now, play with breaks and colors to create a meaningful map
# add a color map with 4 colors
myCol <- rev(terrain.colors(4)) # use the 'rev()' function to put green as the highest NDVI value
# add breaks to the colormap, including lowest and highest values (4 breaks = 3 segments)
brk <- c(0, .25, .5, .75, 1)

# plot the image using breaks
plot(ndvi_calc, main="NDVI for the NEON SJER Field Site", col=myCol, breaks=brk)