Skip to content

v3.11.00

Latest

Choose a tag to compare

@davidfrantz davidfrantz released this 02 Sep 07:09

FORCE v. 3.11.00

Release: 02.09.2026

  • General changes

    • A lot of work has been done on improving the Docker builds and images.
      Many thanks to Peter Jonsson for his help with this! Most things won't be visible to the user, but should improve stability and security.
    • Due to some difficulties in installing the sf R-package, an older version of GDAL has been unknowingly resurfaced in the Docker container.
      Thanks to Benjamin Jakimow for reporting this issue.
      This has now been fixed.
    • The code has beeen updated to accomodate requirments of NumPy 2 and GDAL 3.11, which threw compiliation errors in some cases. Thanks to Peter Jonsson for reporting this issue.

FORCE L1AS

  • All download capabilities were removed from FORCE.
    Specifically, force-level1-landsat, force-level1-csd and force-lut-modis were removed.
    Please see #418 for more details.

FORCE L2PS

  • All download capabilities were removed from FORCE.
    Specifically, force-level1-landsat, force-level1-csd and force-lut-modis were removed.
    Please see #418 for more details.

  • More resampling methods were added, including Cubic Spline, Lanczos, Average, Mode, Max, Min, Median, Q1, Q3, Sum, and RMS.
    Feature requested by Vu-Dong Pham.

  • force-level2-report has a new option -t to choose the theme of the generated charts.
    Available themes are inherited from echarts4r - look in their documentation for more details. Plot generation library has been changed from plotly to echarts4r.

  • In force-level2 / force-l2ps, the setting of the grid origin has been refactored.
    It is now necessary to specify the grid origin as a vector of two values (e.g., "GRID_ORIGIN = -25 60"),
    instead of separate parameters for longitude and latitude.
    Additionally, it is now possible to specify the type of coordinates used for the grid origin (geographic or projected)
    via the new parameter GRID_ORIGIN_TYPE, which can take the values "GEO" or "MAP".
    This allows for more flexibility in defining the grid origin, especially when "clean" values in the map projection are desired.
    The first value in the GRID_ORIGIN vector is always interpreted as the x-coordinate (longitude or easting), and the second value as the y-coordinate (latitude or northing), regardless of the coordinate type.
    Thanks to Sebastian Hafner for highligting this

  • Fixed a bug, where null-value stripes appeared in nodata regions at the edges of the input images, when using the force-level2 / force-l2ps module with co-registration enabled.
    Thanks to Dominique Weber for reporting this issue.

  • fixed a bug, where a segmentation fault occurred when using the force-level2 / force-l2ps module with FILE_AOI and the vector given as Geopackage and when built againt GDAL 3.10 or higher. The issue was seemingly caused by a change deep down in GDAL.
    Thanks to Vu-Dong Pham for reporting this issue.

  • FORCE HLPS

    • in force-higher-level, LEVEL3 module:
      It is now possible to use custom cutoff values for seasonal. cloud, haze, and view zenith scores when
      selecting/weighting observations for compositing.
      Note that parameters USE_CLOUDY, USE_HAZY, USE_OFF_SEASON, and USE_HIGH_VZEN have been refactored,
      and are now replaced by CLOUDY_CUTOFF, HAZY_CUTOFF, OFF_SEASON_CUTOFF, and VZEN_CUTOFF respectively.
      These parameters now take float values instead of logical values.
      This allows users to set a threshold (between 0 and 1) below which observations are considered as cloudy, hazy, off-season, or with high view zenith angle, respectively.
      These are flat-out ignored for the compositing process.
      Setting these parameters to 0.0 will allow all observations to be considered.
      Thanks to Oleg Zheleznyy for proposing and contributing to this feature.
    • in force-higher-level, there are up to three additional paramaters, which allow fine-tuning
      the behaviour in case of reading errors in the input data. This is usually only relevant when
      the input data cube is not properly maintained, e.g., when some files are missing or corrupted.
      These parameters are READ_ERROR_MASK, READ_ERROR_PRIMARY and READ_ERROR_SECONDARY.
      Usable values (depending on context) are STOP, SKIP, and YOLO. STOP will stop the program with an error message, SKIP will skip the problematic block of data and continue with the rest. YOLO will ignore any errors and tries to proceed. Obviously, use this with caution. All
      option will print warning/error messages, so make sure to check your logs/stdout. YOLO is only
      available for ARD-type input data. Note that those parameters only act on reading errors, not on
      empty tiles. Feature requested by Vu-Dong Pham.
    • in force-higher-level, TSA and CSO modules:
      The new parameter ADAPTIVE_RANGE allows to use an adaptive date range for filtering images on a per-pixel basis.
      It comes with a set of other new parameters to control the behaviour of this adaptive date range.
      Technically, it works like the continuous fields already implemented in FORCE, e.g. in the CFIMP module or the phenology-adaptive compositing.
      You require two input images (cubed, Int16), which hold start and end dates for each pixel.
      Multiple filtering windows may be given (as multiband images).
      The dates need to be given as "Year x 365 + DOY", i.e., the same format as used internally in FORCE, with an optional offset (see parameter ADAPTIVE_START).
      Note that this option is just a filtering routine.
      If you provide multiple windows, you will not receive different products for each window!
      Also note that this option might badly interact with other functionality, e.g., deriving phenology metrics.
      Please also note, that this filtering option (contrary to filters like DATE_RANGE) does not reduce input as we need all the data in RAM to determine the adaptive date range for each pixel.
      Feature requested by Oleg Zheleznyy.
    • In FORCE v. 3.9, the default reation options for GTiff files have been updated to use a gridded tile layout.
      This choice has resulted in a decoupling of internal file layout (blocks) and higher level processing units (chunks).
      In the case of working with masks, and relatively small chunks, this has resulted in data artefacts (within the mask-disabled pixels - so not super critical), where
      blocks of data have 0-values (instead of the proper nodata value) in the output files.
      More details can be found here: #430.
      This behaviour can be mitigated by setting the CHUNK_SIZE paramater to the full tile size.
      When using the chunking mechanic, a new fix is now implemented, which fully initializes the output image with the proper nodata value, before writing any data to it.
      In addition, there is a new parameter OUTPUT_INITIALIZE (default: TRUE), which allows to disable this behaviour, if desired.
      Thanks to Simon Spengler for reporting this issue.
    • Some output drivers do not allow updating existing files. Among them, COG.
      If built against GDAL 3.11 or higher, FORCE will now stop with an error message if chunking is used and the driver does not allow updating existing files.
      Either use a different driver, or disable chunking by setting the CHUNK_SIZE parameter to the full tile size.
      For GDAL versions < 3.11, FORCE will still try to write the output file, but will fail with a GDAL error message. Note that earlier GDAL versions are missing a proper check for this, which is why FORCE cannot stop with an error message in this case.
      The provided Docker images are built against new GDAL releases, so this should not be an issue for most users.
    • Added a new spectral index to the TSA submodule, i.e., DWSI (Disease Water Stress Index)
  • FORCE AUX

    • force-tile-finder now allows to input either geographic (lon/lat) or projected map (x/y) coordinates.
      The coordinate type can be specified via the new -t option (values: "geo" or "map").
      The program still outputs both coordinate types for the given point, as well as the corresponding tile indices and pixel indices within that tile.

    • new program: force-runtime-data, which provides information about the FORCE runtime data installation.
      This includes the current location of this data,
      as well as available sensors and indices.
      More features will be added in the future once the runtime data expands.

      .. code-block:: console

      $ force-runtime-data -p # show the installation path of the runtime data
      $ force-runtime-data -s # show available sensors and their band names
      $ force-runtime-data -x # show available indices and their required bands

    • force-map-accuracy is now re-written in C. This avoids to rely on the sf package
      which caused some trouble in the Docker image as it caused a GDAL version conflict.
      The output format has been changed to markdown for nicer rendering.
      Apart from this, there is not really a change in usage.
      Not that it really matters, but it is also faster now.