Temporal compositing is a general technique used in optical remote sensing. Daily observations are composited into the temporal domain to generate proper images, e.g. to acquire cloud-free images.

With this application you can generate composites on-demand, starting from PROBA-V synthesis images (S1/S5 products). It is important to note that PROBA-V S10 products are generated starting from PROBA-V segments while the N-daily compositor starts from S1/S5 TOC or TOA products. As such, an S10 product generated by the N-daily compositor will differ from an S10 product from the PROBA-V Product Distribution Portal.

To request the generation of a composite product:

  • select an area of interest by drawing it on the map or by specifying the lat/lon coordinates.
  • select the start and end date (both inclusive) for the composite product. For composites with a sliding window, select the sliding window tab. For 100m and 300 m products, check the data policy.
  • select a composite algorithm.
  • select the requested resolution (1 km, 300 m or 100 m).
  • select the bands you want in the output product.
  • select the input product type: TOC (Top Of Canopy) or TOA (Top Of Atmosphere).

You can also specify a job description which will make it easier for you to identify the started job in the job overview. In case you want to store these inputs as your user preferences, click the "Save" button Save preferences button. To restore the pre-defined preferences again, click the "Restore" button Restore preferences button.

To start the processing by clicking the "Launch processing" button.

Launching a process

After launching the process, it will become available in the job overview. On the left you see a job overview and the input parameters which were used to start the job. On the right you can track the status of the job and the different steps in the processing workflow. When the processing is finished, viewing and FTP URLs will become available in the results panel. You will also receive an email explaining how to download the results.

Using the drop-down menu on the top, you can filter on the status of the job:

  • Running: job is actually running on the cluster.
  • Completed: job has finished successfully; the results should become available soon.
  • Failed: job has failed.
  • Killed: job has been killed.
  • Pending: job is waiting for data to become available (see subscriptions).
  • Suspended: job is currently suspended and can be resumed using the "Resume" button.

Running jobs can be killed by clicking the "Kill" button Kill job button. Running jobs can also be suspended Suspend job button and resumed Resume job button afterwards.

As shown in the image below, the N-daily compositor processing workflow contains the following steps:

  • CalculateComposite: create the composite products as requested.
  • PublishResults: composite results are being made available for viewing and downloading via FTP. An e-mail will be sent when the products are available for download.

Composite algorithms

As already explained, PROBA-V synthesis (S1/S5) products are used as input for the compositing algorithms.

For all algorithms the quality of the pixel is derived from the status map available in the synthesis input product:

  • cloud-free observations are preferred over ice/snow observations, which in turn are preferred over shadow observations and cloudy observations.
  • observations with a good pixel quality indicator for all bands are preferred over observations of less quality. Observations with good NIR and RED samples are favoured. Also, observations with good BLUE samples are preferred over observations with good SWIR samples.

The following compositing algorithms are available:

  • Maximum value: this algorithm selects the observation with the best quality as described above; when two observations have the same quality, the one with the highest reflectance value is selected.
  • Mean value: this algorithm selects the observation with the best quality as described above; when two observations have the same quality, the mean value of the reflectance values is used in the composite output. Reference: C. Vancutsem , J.‐F. Pekel , P. Bogaert and P. Defourny (2007) Mean Compositing, an alternative strategy for producing temporal syntheses. Concepts and performance assessment for SPOT VEGETATION time series, International Journal of Remote Sensing, 28:22, 5123-5141, DOI: 10.1080/01431160701253212,  http://dx.doi.org/10.1080/01431160701253212.
  • Maximum NDVI: this algorithm selects the observation with the best quality as described above; when two observations have the same quality the observation with the highest NDVI value is preferred. This algorithm is similar but not identical to the compositing algorithm used to create PROBA-V S10 products. Besides using other input data (S1/S5 synthesis products vs PROBA-V segments) an important difference with the PROBA-V S10 compositing algorithm is that also for TOC input products the TOA NDVI product value is used to decide which pixel needs to be selected.

Composites with sliding window

You can create simple composites and composites with a sliding window.

  1. a simple composite creates a composite product for the specified period (start date to end date inclusive).
  2. a composite with a sliding window allows you to create each x days (the composite frequency) a composite product containing y days (the composite span). For example, if you want a composite product of 5 days each 7 days within May 2015 you will get 4 composite products:
  • 01-05-2015 until 05-05-2015 inclusive
  • 08-05-2015 until 12-05-2015 inclusive
  • 15-05-2015 until 19-05-2015 inclusive
  • 22-05-2015 until 26-05-2015 inclusive

When creating a composite with a sliding window, you can choose to align the composites to a calendar month, which can be used to create composites of e.g. a dekad, which are aligned to a calendar month, similar to the composites of the PROBA-V S10 systematic synthesis processing.

Subscriptions

By specifying an end date in the future the composite products will be generated when the required synthesis input products become available. When creating a composite with a sliding window and an end date in the future, the composite products for which the data is available will be generated and delivered. The other composite products will be generated and delivered when the input synthesis products needed to create a composite product become available. This will lead to multiple job runs in the job overview: the first job run will deliver the composite products which could be generated after submitting the process. When new data becomes available, additional job runs will be started when enough data is available to create an additional composite product.

Results

You will receive an email when the data is available for download via FTP. Per composite product, a subdirectory composite_<startDate>_<endDate> is created. The dates (both inclusive) are formatted as yyyyMMdd, e.g. composite_20150501_20150510. This directory contains tiled GeoTIFF files per output band and also a GDAL VRT file (extension .vrt) referencing all GeoTIFF files per output band. These GeoTIFF tiles are aligned to the PROBA-V synthesis tiles.

The filename convention for the GeoTIFF files is PROBAV_PoD_S<nrDaysInComposite>_<startDate>_<resolution>_<tileIndex>_<outputBand>.tiff, e.g. PROBAV_PoD_S10_TOC_20150501_1KM_X18Y03_NDVI.tiff.

We recommend to download the results using an FTP client (e.g. FileZilla) instead of a web browser.

The results can also be viewed in a web browser.

PROBA-V 300 m and 100 m data policy

The PROBA-V data policy is explained at http://proba-v.vgt.vito.be/content/data-policy. All PROBA-V 1 km data are freely available to all users, as well as 300 m and 100 m data older than one month. According to your access rights, composite products requiring 300 m or 100 m input data of the last month will possibly only become available when you are allowed to access all data needed to create a composite product. This could lead to composite products being delivered one month later than expected.

POD REST API

This service is a process which is exposed through the MEP Processing On-Demand REST API and can therefore be used in processing chains. A description of the REST interface can be found here. A valid MEP OpenAM token is required in order to access this service.