In DavidDHofmann/floodmapr: Download and Classify Watermaps for the Okavango Delta

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

floodmapr

floodmapr is an R-package that allows you to download and classify MODIS MCD43A4 satellite imagery into binary maps of dryland and water cover. To be able to download data, you need to have an EarthData account (free). The classification algorithm is based on the publication of Wolski et al., 2017 and currently only applicable for the extent of the Okavango Delta (see Okavango Research Institute).

Installation

You can install the development version of floodmapr from GitHub with:

``` {r, eval=F}

install.packages("devtools")

devtools::install_github("DavidDHofmann/floodmapr")

## Workflow
The `floodmapr` package follows a three step process using the functions,
`modis_download()`, `modis_load()`, and `modis_classify()`.

1. `modis_download()` Download MCD43A4 satellite imagery for desired dates for
the extent of the Okavango Delta. The function automatically downloads,
extracts, stitches and reprojects (EPSG 4326) corresponding modis tiles and
stores them into a multiband raster. The output files are named according to the
input date (e.g. *2020-01-01.tif*). Note: while only band 7 will be used in the
later process, bands 1-7 are stored for completeness.

2. `modis_load()` Once you have downloaded the satellite imagery, load it into R
using `modis_load()`. This will only load band 7 into your current R session.

3. `modis_classify()` Classify the satellite image into a binary map of water
(valued 1) and dryland (valued 0). You can then use the resulting raster and
e.g. `plot()` it.

In addition to the above described main functions there are some additional
helper functions that allow you to do some preliminary analysis on MODIS MCD43A4
band 7.

- `modis_bimodal()` allows you to check if modis band 7 is bimodal. This is
  required to be able to classify the image. Bimodality can fail during some
  months of the year, when surface reflectance of water and dryland become more
  alike.

- `modis_specs()` plots the densities of water- and dryland-reflectance values.
  Again, this can be used to visualize how bimodal the distribution of these
  values is.

- `modis_precentiles()` percentile values of water- and dryland-reflecatances.

Finally, there is a function called `modis_watermask()`. This function can be
used to create a dynamic watermask that can ultimately be fed into
`modis_classify()`. In contrast to the static watermask used by `floodmapr` in
the background, the dynamic watermask uses previously classified foodmaps to
determine areas that were frequently flooded in the past. This way, the
algorithm is able to learn about potential long-term changes in the system.

## Example
Here is a complete example of the above outlined workflow

```r
load("/home/david/ownCloud/Dokumente/Bibliothek/Wissen/R-Scripts/EarthDataLogin.rds")

library(floodmapr)

# Download modis data (you need to provide your own username and password)
downloaded <- modis_download(
    dates     = c("2017-08-21", "2017-08-22")
  , outdir    = tempdir()
  , tmpdir    = tempdir()
  , username  = username
  , password  = password
  , overwrite = F
)

# Load data
loaded <- modis_load(downloaded[1])

# Do some checks
modis_specs(loaded)
modis_bimodal(loaded)
modis_percentiles(loaded)

# Classify data
classified <- modis_classify(loaded)

# Visualize classified image (0 = dryland, 1 = water, 2 = clouds)
plot(classified, col = c("white", "cornflowerblue"))

References

Wolski, Piotr, Mike Murray-Hudson, Kgalalelo Thito, und Lin Cassidy. „Keeping It Simple: Monitoring Flood Extent in Large Data-Poor Wetlands Using MODIS SWIR Data“. International Journal of Applied Earth Observation and Geoinformation 57 (May 2017): 224–34. https://doi.org/10.1016/j.jag.2017.01.005.

DavidDHofmann/floodmapr documentation built on Jan. 15, 2024, 9:32 p.m.