- fix dplyr
first
andlast
- add TIMESAT global model function and cutoff function.
- add the parameter
na.rm
incheck_input
. Ifna.rm
= true, missing and spike values will be discarded; otherwise, they will be interpolated by valid neihbours. - change
analytical
toFALSE
inPhenoDeriv.fFIT
,PhenoGu.fFIT
- fix
minpeakheight
error inPhenoKl
, which leads toPhenoKl
failed unexpectedly. - plateau line fitting in
PhenoGu
-
Fix typo error in curvefits' document.
-
Fix the error of
plot.fFITs
, due tofFIT
renamed tomodel
. -
Rename function
opt_season
toseason_input
-
In the case of none growing season division information,
find_season.peaks
will be error. Fixed now. -
Fix the bug of
init_param
in the approximately monotonous time-sereis (e.g. fast harvest leads to a very short senescence period, and hence approximately monotonic in the whole growing season). In previous version, the parameter boundary ofsos
might exceed that ofeos
. -
Improved the phenology extraction method
PhenoTrs
,PhenoDeriv
andPhenoGu
in the approximately monotonous time-sereis. In the previous versionNA
value will be return. In the current version, if the time series is monotonously increasing, EOS will be the mean value oft[n]
and peak date of season (POP
). ButPhenoKlos
has no method to avoidNA
values, because it used the strict mathematical solution to find the extreme values in the curve of curvature's change rate.
MAJOR updates to improve multi-GS phenology extraction
-
Fix the bug of
findpeaks
, which lead to sharp changed growing season failed to detect. -
Fix the bug of
PhenoKlos
, whereminpeakheight
not work in previous version. -
Fix the bug of
check_season_dt
, wherepeak
might be able to greater thanend
. -
Remove the parameter
check_season_dt
inremoveClosedExtreme
, which might eliminate good extreme values. -
add
get_pheno.rfit
to extract vegetation phenology from rough fitting directly.
-
Parameters of
season_mov
andcurvefits
are wrapped into options. Scripts of phenofit v2.0 will not work anymore. -
Add global options
-
Growing season division was improved. Rough fitting and growing season division are separated.
-
In the package dependency, plyr was replaced with dplyr.
-
Add
doubleLog.AG2
, which allows unbalanced background value in the first half and the second half. -
Add
lambda_vcurve
andlambda_cv_jl
to optimize Whittaker's parameterlambda
-
Add pkgdown, http://phenofit.top/.
-
Growing season division was further tested on FLUXNET daily GPP estimations.
-
Julia interface is ready to go, https://github.com/eco-hydro/phenofit.jl.
-
phenofit script was separated into a another repository, https://github.com/eco-hydro/phenofit-scripts.
- The rough fitting result returned by [season_mov()] and [season()] was renamed
from
whit
tofit
. - parameters of
season_mov
andcurvefits
are moved intooptions
.
- fix visualization (plot and legend) of multiple iteration
- shinyapp works like TIMESAT now,
phenofit_process
andphenofit_shiny
- fix shift error of
wSG
, and addsmooth_SG
andsmooth_wSG
- add weights updating function:
wKong
- support asymmetric background values in spring and autumn season.
- weights of rough fitting not passed to curve fitting now
- growing season dividing is not used.
- fix point shape error of
plot_input
- make_legend supports 4 quality categories (MOD13A1) and 6 quality categories (MOD09A1) now.
- wWHIT only adjusts time series of growing season.
- release in CRAN
shiny
appphenofit
released.- add
QC_flag
to the output ofcheck_input
.
- Add V-curve optimization in
season_mov
for Whittaker's parameter lambda. - Remove
check_ylu
andupper envelope
inwWHIT
. - Update
v-curve
.
- Melt parameters (i.e.
nptperyear
andsouth
) intoINPUT
.check_input
,season
,season_mov
andcurvefits
are impacted. - Add
adj.param
parameter toseason
, which determine whether to automatically adjust roughn curve fitting parameters.
- shiny app
check_season
online now. season
can export rough curve fitting result, even no peaks or trough found.
- fix
Init_param
In previous version, parameters are only rely on "good" (w >= 0.5) values. But when "good" values in a relative low ratio, wield parameters will be generated.
In this version, when "good" values ratio is lower than 40%, all values will be used to generate initial parameters. - Export weigths of every iteration for growing season dividing function,
(i.e.
wHANT
,sgfitw
andwhitsmw2
). And unified their weights updating strategy. doubleLog.zhang
is still not as stable as others.wTSM_cpp
iter parameter is ignored now.