NAME

KEYWORDS

SYNOPSIS

DESCRIPTION

The option nf, number of frequencies, should be carefully chosen. Different numbers of frequencies should be tested first on a small test region before running the module on the full region. As a rule of thumb, the number of frequencies should be at least estimated periodicity + 3, e.g. for NDVI with an annual cycle (one peak per year), the number of frequencies should be at least 4 when analysing one year. If two peaks are assumed per year, the number of frequencies should be at least 5 when analysing one year.

The number of frequencies should also not be too large. With a large number of frequencies, outliers can no longer be identified because the fit is "too good", outliers can be represented by the estimate. The number of frequencies should also be smaller than n input maps / 2 if missing values should be reconstructed.

NOTES

If the range option is given, any values which fall outside that range will be treated as if they were NULL. The range parameter can be set to low,high thresholds: values outside of this range are treated as NULL.

The low,high thresholds are floating point, so use -inf or inf for a single threshold (e.g., range=0,inf to ignore negative values, or range=-inf,-200.4 to ignore values above -200.4).

The maximum number of raster maps that can be processed is given by the per-user limit of the operating system. For example, the soft limits for users are typically 1024. The soft limit can be changed with e.g. ulimit -n 4096 (UNIX-based operating systems) but not higher than the hard limit. If it is too low, you can as superuser add an entry in

/etc/security/limits.conf
# <domain>      <type>  <item>         <value>
your_username  hard    nofile          4096

cat /proc/sys/fs/file-max

Use the file option to analyze large amount of raster maps without hitting open files limit and the size limit of command line arguments. The computation is slower than the input option method. For every single row in the output map(s) all input maps are opened and closed. The amount of RAM will rise linear with the number of specified input maps. The input and file options are mutually exclusive. Input is a text file with a new line separated list of raster map names.

For the usage of the option fet see Roerink et al. (2000). Its value is relative to the value range of the variable being considered.

EXAMPLES

Precipitation data example

# set computational region to one of the maps
g.region raster=2004_03_tempmean -p

# note: color table is different from standard "celsius"
g.gui.animation rast=`g.list type=raster pattern="*tempmean" sep=comma`

Since HANTS is CPU intensive, we test for now at lower resolution:

g.region -p res=5000

# HANTS: Harmonic analysis of the 156 input maps...
# just wildly guessing the parameters for a test run:

# generate and check list of input maps (the order matters!)
g.list type=raster pattern="20??_??_tempmean" output=tempmean.csv

r.hants file=tempmean.csv \
  nf=6 fet=0.1 dod=5 delta=0.1 base_period=12

# assign reasonable color tables for temperature
for map in `g.list type=raster pattern="*tempmean_hants"` ; do
    r.colors $map color=celsius
done

# assign degree Celsius color table
r.colors 2000_06_tempmean_hants color=celsius

# verify with one of the 156 results (still at reduced resolution):
r.mapcalc "2000_06_tempmean_diff = 2000_06_tempmean - 2000_06_tempmean_hants"

r.colors 2000_06_tempmean_diff color=differences
d.mon wx0
d.rast 2000_06_tempmean_hants
d.rast 2000_06_tempmean_diff


r.univar 2000_06_tempmean_diff -g
n=5066
null_cells=5040
cells=10106
min=-0.0899336115228095
max=0.359362050140941
range=0.449295661663751
mean=0.188579838052468
...

# see HANTS time series as animation
g.gui.animation rast=`g.list type=raster pattern="*tempmean_hants" sep=comma`


# Check HANTS behaviour in a given point 
east=740830
north=168832

for map in `g.list rast pat="20??_??_tempmean"` ; do   
  r.what map=$map coordinates=$east,$north >> time_series_orig.csv
done

for map in `g.list rast pat="*tempmean_hants"` ; do 
  r.what map=$map coordinates=$east,$north >> time_series_hants.csv 
done

# merge files:
echo "east|north|temp_orig|temp_hants" > time_series_final.csv

paste -d'|' time_series_orig.csv time_series_hants.csv | \
      cut -d'|' -f1,2,4,8 >> time_series_final.csv

# Resulting CSV file: 'time_series_final.csv'

Using t.* modules to check for basic statistics

# create spatio temporal data set with hants output maps
t.create type=strds temporaltype=absolute  output=tempmean_hants \
  title="Mean Temperature HANTS" description="Mean Temperature reconstructed with HANTS"

# register maps in the strds
t.register -i type=raster input=tempmean_hants \
  maps=`g.list raster pattern=*tempmean_hants sep=,` start="2000-01-01" \
  increment="1 months"

# getting general info of the strds (including max and min of the whole series)
t.info type=strds input=tempmean_hants

# getting statistics for each map in the series
t.rast.univar -h tempmean_hants > stats_hants.txt

SEE ALSO

REFERENCES

Roerink, G. J., Menenti, M. and Verhoef, W., 2000. Reconstructing cloudfree NDVI composites using Fourier analysis of time series. International Journal of Remote Sensing, 21 (9), 1911-1917. DOI: 10.1080/014311600209814

AUTHOR

Last changed: $Date: 2015-05-25 23:13:52 +0200 (Mon, 25 May 2015) $

SOURCE CODE

Available at: r.hants source code (history)

Main index | Raster index | Topics index | Keywords index | Graphical index | Full index

© 2003-2017 GRASS Development Team, GRASS GIS 7.2.2svn Reference Manual