Branch: master
Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
.ipynb_checkpoints
2018-07-12-BisQue_upload_images.ipynb
2018-07-12-Leaflet_Folium_example.ipynb
2018-07-12-NEON_api_python_lidar.ipynb
2018-07-12-NEON_api_python_orthophoto.ipynb
2018-07-12-NEON_api_python_reflectance.ipynb
README.md

README.md

QUBES NEON teaching exercises

Focus on NEON AOP lidar and hyperspectral imagery, conducting analyses on cloud and HPC.

Author: Tyson Lee Swetnam

Intent: I am not teaching this as a course (yet). I’ve got some flexibility in developing the lesson plan and pacing.

Teaching Philosophy: I want to take a reproducible science approach, using pre-written Jupyter notebooks with empty code sections that can be filled in by the student programmers.

Learning Objectives: NEON’s data science skills provide notebooks and lesson plans for analyzing hyperspectral imagery. I plan to develop complementary notebooks for downloading AOP lidar data; segmenting individual organisms (herbaceous plants, woody shrubs, and trees) using open-source code, vertically sampling (all bands and time series) hyperspectral signatures for each organism.

The lessons are targeted toward advanced graduate students and researchers. Lessons are expected to take at least one week, e.g. 2-3 lectures, one computer lab session, with online nightly homework.

Students will be allowed to work in small teams (paired programming) to develop their individual notebooks.

Prerequisites

Lesson 1: Data Science Workbench

Goals: Provision virtual machines on CyVerse Atmosphere or XSEDE Jetstream, and Discovery Environment, for data analyses.

Value: Learn to launch remote metal running on fast Internet2.

Lesson 2: Get NEON AOP data

Goals: Learn to download NEON data using the NEON API

Value: Get large quantities of data using batch commands.

  • NEON Data API
    • Python Jupyter Notebooks for downloading lidar & imagery data
    • Rmarkdown notebooks for downloading lidar & imagery data
  • CyVerse iRODS client
    • getting data to and from the CyVerse DataStore.
  • 3rd Party software
    • CyberDuck

Lesson 3: Lidar data QAQC and Visualization

Goals: Learn basic lidar processing and visualization techniques.

Value: Improve classification (bare earth, vegetation, infrastructure) of lidar and visualize data anywhere.

  • IntroducePDAL (w/ Docker) for lidar classification.

    • CLI executions.
    • JSON scripting.
    • Colorize lidar data with aerial orthophotography
  • Introduce Potree, Entwine & Greyhound for point cloud visualization.

  • Introduce CloudCompare

Lesson 4: Lidar stem segmentation in R (lidR)

Goals: Learn how to segment individual trees/shrubs/herbaceous plants from lidar data in R.

Value: Generate ecologically relevant inventory (census) of physical characteristics.

  • Run lidR stem segmentation

  • Develop an individual object (plant) inventory for various scales.

  • Leaflet mapping in R.

Lesson 5: NEON Hyperspectral QAQC & simple calibrations

Goals: Learn to use NEON hyperspectral imagery.

Value: Utilize hyperspectral imagery which will allow species level ID and phenology (health).

Lesson 6: Point/polygon sampling of individuals w/ corresponding hyperspectral signatures

Goals: Learn to point and polygon sample individual plants in R

Value: Ecologically relevant inventory with phenology and species.

  • Attribute individual organisms with their corresponding hyperspectral reflectances

Lesson 7: Statistical Analyses in RStudio

Goals: Conduct statistical analyses on your new datasets.

Value: Hypothesis testing using data.

  • Graphical representations of the data

    • ggplot2
    • shiny
  • Statistical analyses

    • linear modelling, maximum likelihood, generalized additive models (gam), random forests

Lesson 8: Statistical Analyses in Jupyter

Goals: Conduct statistical analsyes on your new datasets

Value: Hypothesis testing using data.

  • Graphical representations of the data
    • matplotlib
    • pandas

Lesson 9: Scaling analyses to multiple NEON sites

Goal: Develop workflows which utilize data from multiple NEON sites and time series analyses

  • Introduce Makeflow and Singularity
  • Writing job requests for HPC