Skip to content

5. Export results

STEP 5: Output export

This step allows you to output the results in the format desired by the user, for the desired period and frequency. If chosen, a confidence index corresponding to the current stress index of pixels detected as suffering from dieback, and within the forest mask and valid area mask is extracted from the stress index results of step 3 and exported. The pixels are also classified, based on the discretization of this confidence index using a list of thresholds. Pixels with only three anomalies are classified as the lowest class, because 3 anomalies are not considered enough to calculate a meaningful index.


The input parameters are :

  • data_directory: The path of the output folder in which the results will be written
  • start_date: Start date, in YYYY-MM-DD format
  • end_date : End date, in YYYY-MM-DD format
  • frequency : Frequency used to aggregate results. Can be "sentinel", in which case the periods used to aggregate results corresponds to the time between available SENTINEL dates, Can also be a frequency as accepted by the pandas.date_range function (e.g.: 'M' (every month), '3M' (every three months), '15D' (every 15 days)")
  • multiple_files : If True, one shapefile will be exported per period, where each polygon corresponds to the state of the area at the end of the period. Otherwise, only one shapefile is exported and the polygons contain the period when the first anomaly was detected.
  • conf_threshold_list : List of thresholds used as bins to discretize the confidence index into several classes, stress_index_mode must not be None in step 3.
  • conf_classes_list : List of classes names, if conf_threshold_list has n values, conf_classes_list must have n+1 values, stress_index_mode must not be None in step 3.


The outputs of this fifth step, in the folder data_directory/Results, are :

  • if multiple_files is False:
    • the shapefile periodic_results_dieback, whose polygons contain the time period when the first anomaly was detected for the areas suffering from dieback. The areas reached before start_date or after end_date are ignored. If conf_threshold_list and conf_classes_list are provided, the polygons also contain an anomaly intensity class discretized and vectorized from the stress index calculated in the 05_compute_confidence step. This class therefore contains the "final" state, calculated at the last available Sentinel-2 date. If bare ground is detected, the confidence index is not calculated and this final state becomes "Bare ground".
    • if soil_detection was True in the first step, the shapefile periodic_results_soil whose polygons contain the period when the first soil anomaly was detected for areas detected as bare soil/cut. Bare areas before start_date or after end_date are not shown.
    • A confidence_index raster, containing the continuous value of the confidence index which is also discretized in the vectorized results.
  • if multiple_files is True :
    • One shapefile per period whose name is the end date of the period (for example, with start_date = 2018-01-01, end_date = 2018-04-01 and frequency = "M", we will have the following files: 2018-01-30.shp, 2018-02-28.shp and 2018-03-31.shp. Each shapefile contains polygons corresponding to the stand state at the end of the period, even if the first anomalies occur before start_date. The status can be 'Anomaly', "Bare ground" and 'Bare ground after anomaly' if soil_detection was True in the first step, or simply "Anomaly" otherwise.

How to use

From a script

from fordead.steps.step5_export_results import export_results
    data_directory = <data_directory>,
    start_date = <start_date>,
    end_date = <end_date>,
    frequency= <frequency>,
    multiple_files = <multiple_files>,
    conf_threshold_list = <conf_threshold_list>,
    conf_classes_list = <conf_classes_list>

From the command line

fordead export_results [OPTIONS]

See detailed documentation on the site

How it works


Importing information on previous processes and deleting obsolete results if they exist

The information about the previous steps is imported (parameters, data paths, used dates...).

Functions used: TileInfo(), methods of the TileInfo class import_info()

Importing the results of the detection

The results of the previous steps are imported.

Functions used: import_soil_data(), import_dieback_data(), import_binary_raster()

Determining the periods for aggregating the results

The results will be aggregated according to the time period at which the first anomalies occur for both soil and dieback detection. These periods are determined from the frequency indicated by the parameter frequency, the start date start_date and the end date end_date. Periods before the first SENTINEL date used, or after the last one, if they exist, are ignored since they cannot correspond to any result.

Functions used: get_bins()

Conversion of first anomaly dates into number of days since 2015-01-01

The dates of first anomalies, stored as indexes of the dates used, are converted into number of days since a reference day "2015-01-01" corresponding to the launch of the first SENTINEL-2 satellite. Thus these dates can be compared with the limits of the previously determined periods.

Functions used: convert_dateindex_to_datenumber()

If exported in several files:

  • For each period, the algorithm checks if the pixel has a first anomaly before the end of the period. We thus obtain the information for each pixel "Healthy", or "Dieback" if export_soil is False, or "Healthy", "Dieback", "Cut", "Sanitary cut" otherwise.
  • This information is vectorized using only the area of interest (within the forest mask and with enough valid dates to model the vegetation index). Healthy pixels are also ignored.

    Functions used: get_state_at_date()

  • This vector is written for each of the periods using the period end date as the file name.

If exported to a single file:

  • Pixels are aggregated according to the period in which the first anomaly occurs.
  • The result is vectorized, ignoring the pixels with first anomalies outside the determined periods and the pixels outside the studied area (within the forest mask and having enough valid dates to model the vegetation index).
  • If conf_threshold_list and conf_classes_list are provided, the stress index of pixels currently detected as suffering from dieback is extracted from the multi-band stress_index raster stack, then dicretized and vectorized using conf_threshold_list and conf_classes_list parameters.
  • The periodic results are intersected by union with the vector thus obtained. The polygons detected as confirmed anomalies without a confidence class can only come from the bare ground detection if it is activated, so these polygons take the "Bare ground" class.
  • The resulting vector is written in a single vector file where polygons contain the period when the first anomaly was detected, and possibly the confidence class.
  • If computed, the continuous confidence index array is also written to a raster file.

    Functions used: get_periodic_results_as_shapefile(), union_confidence_class(), vectorizing_confidence_class

If export_soil is True, the same operation is performed using the bare soil detection results and the results are written to a second vector file.