A basic Level 3?

[Antguz]

Hi,

When you see the force-higher-level submodules all of them have a purpose (e.g., time series, clear sky observations, temporal aggregation, etc). Likewise, all of them do a qai screening in their .prm file. But what about if a user, like myself, just wants a Level 3 product with similar properties like Level 2 (e.g., dates, band names) filtered by a given qai screening procedure? Is that possible?

I was expecting to do so using the Level 3 force-higher-level submodules (i.e., force-parameter L3.prm LEVEL3 > force-higher-level L3.prm). However, for reasons that I do not know why, it requires arguments from submodules BAP and PAC (enclosed describe bellow). Thus, the outputs are far from being Level 2-qai screened.

I was not expecting to see that. Is this ok? If so, how I can do a basic Level 3 product (i.e., Level 2-qai screened)?

Thanks in advance!

Regards.

AG.

# Best Available Pixel (BAP) compositing
# ------------------------------------------------------------------------
# This parameter specifies the target year for compositing.
# Type: Integer. Valid values: [1900,2100]
YEAR_TARGET = 2021
# This parameter specifies thenumber of bracketing years (target year +- 
# bracketing years), i.e. the compositing period. A value of 2 would result
# in a five-year compositing period.
# Type: Integer. Valid values: [0,100]
YEAR_NUM = 1
# This parameter is a tradeoff parameter that balances the inter- and intra-
# annual selection. Lower values (e.g. 0.75) favor data from the target year.
# Higher values favor data that was acquired close to the target DOY (regard-
# less of the year).
# Type: Float. Valid values: [0,...
Y_FACTOR = 0.75
# These parameters specify the function values used for fitting the DOY
# scoring functions. The function type is automatically chosen from the 
# given values, i.e.
#   Gaussian           s0 < s1 > s2
#   Descending sigmoid s0 > s1 > s2
#   Ascending sigmoid  s0 < s1 < s2
# Type: Float list, 3 values. Valid values: ]0,1[
DOY_SCORE = 0.01 0.99 0.01
# These parameters specify the DOYs used for fitting the DOY scoring
# functions in case of the static compositing. They are not used for the 
# phenology-adaptive compositing. However, in each case, the target date
# appearing in the file name is derived from these values. The target date
# is the value with highest score (see last parameter).
# Typically, the DOYs are in order, e.g. p0 = 60, p1 = 90, p2 = 120.
# However, the DOY scoring can also extend between the years (i.e. around 
# the turn of the year). If p0 > p1: p0 is from previous year, e.g. p0 = 330,
# p1 = 30, p2 = 90. If p2 < p1: p2 is from next year, e.g. p0 = 300, p1 = 330,
# p2 = 30.
# Type: Integer list, 3 values. Valid values: [1,365]
DOY_STATIC = 120 180 240
# This parameter specifies whether all available data within the requested time
# frame are used – or only from the season of interest. If FALSE, the composites 
# only consider data for the period, in which the intra-annual score is higher 
# than 0.01. If there is no clear-sky data within this period, data gaps are 
# possible. If TRUE, all data from the requested years are used, thus the risk
# of having data gaps is lower. However, it is possible that data from unwanted
# parts of the year are selected.
# Type: Logical. Valid values: {TRUE,FALSE}
OFF_SEASON = FALSE
# This parameter controls the strength of the DOY score.
# 0 disables the use of this score.
# Type: Float. Valid values: [0,1]
SCORE_DOY_WEIGHT = 1.0
# This parameter controls the strength of the Year score.
# 0 disables the use of this score.
# Type: Float. Valid values: [0,1]
SCORE_YEAR_WEIGHT = 1.0
# This parameter controls the strength of the cloud distance score.
# 0 disables the use of this score.
# Type: Float. Valid values: [0,1]
SCORE_CLOUD_WEIGHT = 0.4
# This parameter controls the strength of the haze score.
# 0 disables the use of this score.
# Type: Float. Valid values: [0,1]
SCORE_HAZE_WEIGHT = 0.3
# This parameter controls the strength of the correlation score.
# 0 disables the use of this score.
# Type: Float. Valid values: [0,1]
SCORE_CORREL_WEIGHT = 0.0
# This parameter controls the strength of the view zenith score.
# 0 disables the use of this score.
# Type: Float. Valid values: [0,1]
SCORE_VZEN_WEIGHT = 0.0
# This parameter indicates the distance (to the next cloud or cloud shadow) after 
# which the sky is assumed to be clear (cloud score approaches 1.0). The distance
# needs to be given in meters.
# Type: Float. Valid values: [1,...
DREQ = 3000
# This parameter indicates the view zenith angle at which the view zenith score
# approaches 0.0. The angle needs to be given in degree. 
# Type: Float. Valid values: [1,90]
VREQ = 7.5
# Output the composite?
# Type: Logical. Valid values: {TRUE,FALSE}
OUTPUT_BAP = TRUE
# Output the compositing information?
# Type: Logical. Valid values: {TRUE,FALSE}
OUTPUT_INF = TRUE
# Output the compositing scores?
# Type: Logical. Valid values: {TRUE,FALSE}
OUTPUT_SCR = FALSE
# Output quicklook of the composite?
# Type: Logical. Valid values: {TRUE,FALSE}
OUTPUT_OVV = FALSE

# Phenology Adaptive Compositing (PAC)
# ------------------------------------------------------------------------
# This parameter defines whether the phenology-adpative compositing (TRUE)
# or the static compositing (FALSE) should be used. In case of the static 
# version, the target DOYs are derived from DOY_STATIC. In case of the 
# PAC, the target DOYs are retrived from the files given by LSP_FILE
# Type: Logical. Valid values: {TRUE,FALSE}
LSP_DO = FALSE
# Land Surface Phenology datapool (parent directory of tiled LSP)
# Type: full directory path
DIR_LSP = NULL
# Basenames of the LSP datasets that are used as compositing targets 
# (analogously to DOY_STATIC). Each file should be a multi-band image
# wherein the bands represent different years. The number of bands, and
# the corresponding years, need to be the same for all files.
# Type: List with basenames of 3 files
BASE_LSP = NULL
# This parameter defines year, which corresponds to he 1st band of the
# LSP.
# Type: Integer. Valid values: [1900,2100]
LSP_1ST_YEAR = 2000
# This parameter specifies the starting point of the LSP values. 
# Internally, the data are represented as ‘Year x 365 + DOY’. Thus, LSP_START
# is an offset, which must be given as ‘Year x 365 + DOY’. If the values are 
# provided in this format, use LSP_START = 1. If the LSP values would be pro-
# vided relative to January 1 2000, use LSP_START = 730001, i.e. 2000*365+1. 
# Leap years are not taken into account and each year consists of 365 days.
# Type: Integer. Valid values: [1,2100*365]
LSP_START = 2000
# This parameter is a threshold in days. If the inter-annual variability of the 
# LSP (of a given pixel) exceeds this value, the long-term average LSP is used 
# instead of the yearly values. The value should be between 0 (long-term average
# is used for all pixels) and 365 (long-term average is never used).
# Type: Integer list. Valid values: [0,365]
LSP_THRESHOLD = 182
# This parameter defines the nodata value for the LSP.
# Type: Integer. Valid values: [-32768,32767]
LSP_NODATA = -9999

[davidfrantz]

Hi @Antguz,

Level 3 products are temporal aggregations of lower-level products.
If I understand correctly, you are looking for Level 2 products, in which the pixels with a specific quality are set to nodata, correct?

There is no functionality to exactly do this.
The main reason is: by doing this, you are merely creating another physical copy of the data, instead of dynamically working with the data and applying the quality layers on-the-fly.

I believe the closest thing would be to:

• use force-higher-level with the TSA submodule
• INDEX = BLUE GREEN RED NIR SWIR1 SWIR2 (this is for Landsat)
• OUTPUT_TSS = TRUE to output the filtered time series of these bands

This will give you one file for each Index. These are filtered according to the dates etc, and the quality screening will be applied too.

You can then optionally use force-stack in another step to create a 4D-datastructure.

Alternatively, if you prefer single-band files, you can use OUTPUT_EXPLODE = TRUE.

Hope this helps somewhat,
David

[Antguz]

Hi @davidfrantz,

Thanks for your explanation. It helps!

AG.