Remote Sensing of Plants and Topography in R

Kyla M. Dahlin, Michigan State University
Author Profile


This module introduces students who are already familiar with remote sensing and R to doing quantitative analyses with large spatial data sets. Students will explore different possible abiotic drivers of plant growth, defined as greenness and height. In the final step, students will analyze data from around the United States and consider macroscale patterns of vegetation controls.

Strengths of Module

This module asks a fundamental question to landscape ecologists - why do plants grow differently in different places? However, even for non-ecologists, this module allows students to use raster data to ask quantitative questions at multiple scales (local and continental). Code is provided in R that can either be taught line by line in class, or can be run simply by changing the directory name at the beginning of the script, depending on time available and students' skill level.

What does success look like

Successful students will produce a series of figures in R representing different vegetation and landscape variables, then compare these via scatterplots and regression. They will do this first with a single data set, then on a different data set of their choosing. Students will understand how to quantify the influence of topography on vegetation across multiple biomes. Students will practice statistical methods (regression, graphing) and develop higher-order thinking skills including hypothesis generation and synthesis. The final goal is for students to interpret the large scale spatial patterns of correlations, attributing their variation to geographical drivers like latitude, biome, or geologic history.

Overall Learning Goals are for students to:

  • Test whether plant growth (greenness and height) is driven more by elevation, slope, or aspect.
  • Investigate an ecological question at both local and continental scales.
  • Analyze spatial raster data in R, moving between making maps and doing non-spatial statistical tests.
  • Consider macroscale (continental scale) patterns of relationships between topography and vegetation.

Context for Use

This module has been developed for an upper-level remote sensing class, where students have a lot of familiarity with gridded (raster) data sets and concepts like NDVI and digital elevation models. They have had some introduction to R. The class that the module was tested in is a 15-person lab, with computers pre-loaded with R, RStudio, and the data. The labs are 2 hours each, and the expectation is that this module could be completed over the course of two lab sessions, with some of the answer writing done as homework, if necessary.

It could easily be adapted to another small class, if students were able to bring laptops; or to a landscape ecology course as a relatively small quantitative component. For example, this could serve as a quantitative activity in connection to the "Ecosystem Processes in Heterogeneous Landscapes" chapter in the textbookLandscape Ecology in Theory and Practice(Turner & Gardner 2015), which references the Riera et al. (1998) paper that is listed in the references below.

Description and Teaching Materials

Why this Matters:

At the global scale, we know that there is more above-ground biomass near the equator than there is towards the poles. At the local scale, however, many different environmental gradients (e.g. topography, geology, hydrology) can influence plant growth. Within each of these environmental factors, there are many possible ways the gradients can be quantified. Understanding these environmental drivers is important because they are likely to change in the face of global change pressures - for example, as temperatures increase, plants may become greener at higher elevations, indicating either an increase in growth of the existing plants or a change in the plant species. Similarly, plants of the same species tend to grow larger under optimal growth conditions. In this module, students will learn how to assess whether plant greenness and height are correlated with topographic variables (elevation, slope, aspect) at one location and then across many locations in the US.

Quick outline/overview of the activities in this module

  • Pre-module work: Discussion of papers read for class and PowerPoint presentation. Students should have familiarity with DEMs, slope, aspect, and NDVI (greenness). Students should have some comfort with R and have access to R and RStudio. See the References and Resources section for suggested tools.
  • Activity A: Calculate slope and aspect from a DEM, map elevation, slope, aspect, NDVI, and vegetation height. Calculate northness (cosine of aspect). Based on visualization, generate hypotheses about the most likely driver(s) of vegetation growth.
  • Activity B: Calculate correlations between topographic variables and vegetation variables. At the end of this activity, students decide on one metric and gradient combination to explore across a set of NEON sites to determine to what extent the local pattern applies at large scales.
  • Activity C: Each student chooses a NEON location, accesses data, and calculates correlations as had been done in Activity B. Then, the correlations chosen at the end of Activity B are mapped together (by hand or digitally) on a map of the U.S. that is accessible to all students, and macroscale patterns are considered qualitatively (as a group or as a written assignment).

Activity A

Students will load the NDVI, DEM, and canopy height model (CHM) data into R and plot it and calculate slope and aspect. They will then build a multi-panel figure of DEM, slope, aspect, NDVI, and CHM. Discuss challenges of using aspect (0 and 360 are adjacent), and have students convert aspect to cosine of aspect ('northness'). Based on visual analyses of these maps students should generate hypotheses about which environmental gradients are the likely strongest drivers.

Activity B

Students will test the hypotheses generated in Activity A using scatterplots and regression. They will generate a multi-panel figure with regressions between vegetation height and the three metrics (elevation, slope, northness) and between greenness and these metrics. This activity concludes with a discussion of the different plots and students will make a decision about which to test in Activity C.

Activity C

Each student will choose a NEON site location, and do all the regression tests they had done in Activity B, and then report the one selected by the group for further continental-scale analysis. Students will then all contribute their R2 values to a map, discuss possible patterns and drivers qualitatively, then write a response to the prompt

"Is [[chosen driver]] equally important to [[NDVI or height]] across the US, or does it appear to vary with a large-scale geographical pattern? Which geographical pattern looks like it is the most important? Why do you think this is? What follow up analyses would you propose to further explore this finding?"

Teaching Materials:

  • R code for students ( 6kB Oct29 19)
  • Student Handout (Microsoft Word 2007 (.docx) 104kB Oct28 20)
  • Dataset 1 (Zip Archive 7.6MB Oct29 19)- NEON AOP DTM, DSM, and NDVI for A & B use this 2018 California site.
  • Dataset 2 (Zip Archive 519.4MB Jul3 20) - NEON AOP DTM, DSM, and NDVI for all sites where data was collected in 2018 & 2019 (note this zip file is >500 MB!)
  • Spreadsheet to record student data (Comma Separated Values 9kB May28 20) (for Activity C)
  • R code to make final map (R script 3kB May28 20)

Teaching Notes and Tips

If time is limited, or if students are learning online, it might be better to provide them with the R code in advance. However, students will have a more 'authentic' experience if they type out the code themselves - this can be done either by providing a paper copy that they have to transcribe or, (recommended), the students follow along in real time as the instructor types and projects the code on a screen. So much of writing code is catching small typos or inconsistencies; it is good for students to see this happen in real time.

Workflow of this module:

  1. Assign pre-class readings; students should have already learned about NDVI and topographic variables (either via lecture for this lab or elsewhere in class). Students should also do some introductory work in R ( is great).
  2. Give students their handout (Microsoft Word 2007 (.docx) 104kB Oct28 20) and printed code ( 6kB Oct29 19) when they arrive to class.
  3. Instructor leads students through Activities A and B, pausing for discussions and to troubleshoot as needed.
  4. Students work through Activity C on their own, reporting correlation values to the instructor.
  5. Final re-group for students to look at the map of all values (or post online so students can interpret large scale patterns on their own).

Notes on the student handout:

  • The student handout that is included in this module packet was generally designed for an advanced remote sensing or plant ecology course and to be used in combination with the PowerPoint presentation. We recommend that the instructor revise the handout and presentation as appropriate for their own classroom. For example, integrating information from the presentation into the text (such as the overview of trendlines and slope) may be necessary if the instructor does not include this in their lecture. Alternatively, for a more advanced class some material may be removed from the module.
  • All of the data needed for this module is provided, though for part C, there is a large zipped file that needs to be unzipped, then within that are many individual site x year zip files to unzip.

Potential pre-class readings:

  1. R you ready for EDDIE (Microsoft Word 2007 (.docx) 23kB Nov19 18) from the Macrosystems EDDIE Teleconnections Module.
  2. Dahlin, K.M., Asner, G.P. and C.B. Field. 2012. Environmental filtering and land-use history drive patterns in biomass accumulation in a mediterranean-type landscape. Ecological Applications. 22(1): 104–118.
  3. Armesto, J. J., and J. A. Martinez. 1978. Relations between vegetation structure and slope aspect in the Mediterranean region of Chile. Journal of Ecology 66:881–889.
  4. Riera, J., Magnuson, J., Vande Castle, J. et al. 1998. Analysis of Large-Scale Spatial Heterogeneity in Vegetation Indices among North American Landscapes.Ecosystem s 1: 268–282.

Measures of Student Success

Student success can be measured by considering the following:

  • Were students able to produce the maps and plots with the California data set when led by the instructor?
  • Were students able to compare the regression values and identify stronger and weaker relationships?
  • Were students able to generate testable hypotheses about the relationships between the topographic and vegetation variables?
  • Did students compare the vegetation patterns in their site to the California site in a meaningful way, considering larger environmental patterns?
  • Were students able to connect the continental vegetation patterns to other continental scale gradients (climate, biome, geologic history)?

References and Resources

National Ecological Observatory Network (NEON)

NEON Airborne Observation Platform (AOP)

NEON Data Portal

Teaching Tips from Software Carpentry