Sample Data

This page describes datasets that we provide for users to evaluate. It is not suitable for publication, and we have not completed QA/QC, but is a good place to start exploring as the datasets are limited in size and do not require access to our filesystem or databases.

Two Sorghum BAP experiments were selected for inclusion the sample data: Season 4 (summer 2017) and Season 6 (summer 2018). In order to provide a representative sample across the growing season, four general periods were targeted: 

  • planting (early April),
  • ~week 4 (mid May)
  • ~week 8 (mid June)
  • harvest (late July/August)

The selected sensors’ inventory was reviewed and dates were selected for each sensor. When possible, the same date ranges were selected across sensors, however due to the nature of the Scanalyzer schedule and periodic maintenance not every sensor has data for the same dates. Additionally, Season 4 has not been completely reprocessed with all of the new pipelines developed in 2018 so there are some data products such as plot-level 3d point cloud LAS data that are unavailable for Season 4. In general, dates were selected to provide the best coverage for each sensor for the periods above but cross-sensor comparison will not always be possible.

To reduce the data volume of the sample dataset, data was filtered where possible to two particular plot ranges (20 and 30) in the field, each consisting of 16 columns totaling 32 plots per day. For some sensors such as the environmental logger data (applying to the entire field) and the PSII data (where plot filtering was not yet supported) an entire field’s worth of data was included for the selected dates. One particular sensor, the hyperspectral VNIR sensor, is especially large (individual files exceed 75 GB) so only a single dataset was included for each day.

For more information about the processing steps involved in the generation of these products, our open-source code is made available onhttps://github.com/terraref in the various “extractors” repositories. For example, theps2_png directory was created using the psii2png extractor contained here:https://github.com/terraref/extractors-multispectral/tree/master/psii2png. The TERRA-REF documentation is constantly being expanded as well.

Our first example datasets have been curated and are available for download in the Sample Data (2019) space in our Sensor Data Portal; versions of the same data clipped to the plot level are also available in the Sample Data 2019 Plots space.

We are also developing introductory tutorials that teach users how to access data using R and Python. Please be patient as we publish datasets, improve our documentation, and launch a new search interface to accompany our public release of data in November 2019 ... Stay tuned!

Here is a summary of the data available in our initial release: