[GRASS-user] Extracting vegetation phenology from Landsat-based time series

[Markus Metz]

There is a new addon r.seasons which determines the number of seasons per
pixel and extracts start and end dates for the given number of seasons from
a time series.

The module is designed for noisy input, e.g.

time|value
0|0
1|0
2|1
3|0
4|1
5|1
6|1
7|0
8|1
9|1
10|1
11|1
12|0
13|0

with threshold 0.5 and minimum season length set to 3 would detect one
season. Core season start is at 4, end at 10. Full season start is at 2,
end at 10.

Markus M


On Wed, Nov 9, 2016 at 4:17 PM, Nikos Alexandris <nik at nikosalexandris.net>
wrote:

> Nikos Alexandris:
>
> Is the number of cycles per year in the Wiki as well?  Going through,
>>>> last time, I think I didn't grasp that.  Can you pin-point?  It's
>>>> exactly what we need at the moment.
>>>>
>>>
> Veronica Andreo:
>
> Yes, maybe it is not well explained, but it is here:
>>> https://grasswiki.osgeo.org/wiki/Temporal_data_processing#Fi
>>> lling_and_reconstructing_time_series_data_with_gaps_-_HANTS
>>>
>>
> and also in the manual under "NOTES"
>>> https://grass.osgeo.org/grass70/manuals/addons/r.hants.html
>>>
>>
> The idea is, once you are happy with the result of hants, you ask for
>>> amplitude outputs and with those amplitude maps (you'll have one map per
>>> frequency) you run r.series method=max_raster. This will give you the
>>> most
>>> important frequency (according to amplitude value of that frequency) in
>>> each pixel. If the result is that the most important frequency is 0, then
>>> you have one cycle per year (that, if you use a base period of one year
>>> of
>>> course).
>>>
>>
> Hope it is clearer now :)
>>>
>>
> Nikos:
>
> Yes. It was even before -- just me read through quickly.
>>
>> r.hants is awesome!
>>
>> We have 24 Landsat 8 derived EVI2 maps (2 for each month) for one year.
>> Clouds and Water surfaces removed. Another set of "relatively normalised
>> images, using `i.histo.match` is also ready.  We really need to "fix"
>> `i.histo.match` to crunch floats directly.  The resulting ranges worry
>> me a bit -- I just followed the "old" way as discussed some years ago in
>> a relevant thread (floats > integers > histo-matching > floats).
>>
>> Anyhow, I am testing the following:
>>
>> for NF in 4 5 6 7 8 ;do r.hants file=evi2_maps nf=$NF fet=0.05 dod=3
>> base_period=24 suffix=_hants_nf_$NF amplitude=amplitude_hants_nf_$NF
>> phase=phase_hants_nf_$NF ;done
>>
>> (thinking loudly... it would be super-nice to have `t.rast.hants`)
>>
>> Not sure about `dod`. Perhaps it should also follow a patten like 3 4 5
>> 6 and 7 for the above?
>>
>> for NF in 5 6 7 8 ; do t.rast.series evi2_hants_nf_$NF method=max_raster
>> output=dominant_frequencies_hants_nf_$NF --o
>>
>
> Just in case, the above *should* read:
>
> for NF in 4 5 6 7 8 ;do r.series input=`g.list rast
> pattern=amplitude*nf_${NF}* separator=comma` method=max_raster
> output=dominant_frequencies_hants_nf_${NF} --o ;done
>
>
> "Dominant" frequencies 0, 1, 2 and 3 appear to be within the areas
>> of our interest (agricultural surfaces).  Very good.  And convincing.
>> Yet I am learning to interpret this correctly. And the "phase" as well.
>> Reminds math studies, years ago.
>>
>
> It is a "good" thing to actually have similar dominant frequency "maps"
> for all of the experimented Number of Frequencies.
>
> A confirmation would be reassuring.
>
> Thanks, Nikos
>
> [rest deleted]
>
> _______________________________________________
> grass-user mailing list
> grass-user at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/grass-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20161128/148e4407/attachment-0001.html>