Journal cover Journal topic
Biogeosciences An interactive open-access journal of the European Geosciences Union
Journal topic
Discussion papers
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Research article 20 Jun 2019

Research article | 20 Jun 2019

Review status
This discussion paper is a preprint. It is a manuscript under review for the journal Biogeosciences (BG).

Comparing Stability in Random Forest Models to Map Northern Great Plains Plant Communities Using 2015 and 2016 Pleiades Imagery

Jameson Brennan1, Patricia Johnson1, and Niall Hanan2 Jameson Brennan et al.
  • 1South Dakota State University West River Agricultural Center 1905 N Plaza Dr. Rapid City, SD 57702, USA
  • 2Jornada Basin LTER, New Mexico State University Plant and Environmental Sciences Las Cruces, NM 88003, USA

Abstract. The use of high resolution imagery in remote sensing has the potential to improve understanding of patch level variability in plant structure and community composition that may be lost at coarser scales. Random forest (RF) is a machine learning technique that has gained considerable traction in remote sensing applications due to its ability to produce accurate classifications with highly dimensional data and relatively efficient computing times. The aim of this study was to test the ability of RF to classify five plant communities located both on and off prairie dog towns in mixed grass prairie landscapes of north central South Dakota, and assess the stability of RF models among different years. During 2015 and 2016, Pleiades satellites were tasked to image the study site for a total of five monthly collections each summer (June–October). Training polygons were mapped in 2016 for the five plant communities and used to train separate 2015 and 2016 RF models. The RF models for 2015 and 2016 were highly effective at predicting different vegetation types associated with, and remote from, prairie dog towns (misclassification rates < 5 % for each plant community). However, comparisons between the predicted plant community map using the 2015 imagery and one created with the 2016 imagery indicate 6.7 % of pixels on-town and 24.3 % of pixels off-town changed class membership depending on the year used. Given the low model misclassification error rates, one would assume that low changes in class belonging between years. The results show that while RF models may predict with a high degree of accuracy, overlap of plant communities and inter-annual differences in rainfall may cause instability in fitted models. Researchers should be aware of similarities between target plant communities as well as issues that may arise with using single season or single year images to produce vegetation classification maps.

Jameson Brennan et al.
Interactive discussion
Status: open (until 15 Aug 2019)
Status: open (until 15 Aug 2019)
AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment
[Subscribe to comment alert] Printer-friendly Version - Printer-friendly version Supplement - Supplement
Jameson Brennan et al.
Jameson Brennan et al.
Total article views: 164 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
114 47 3 164 2 1
  • HTML: 114
  • PDF: 47
  • XML: 3
  • Total: 164
  • BibTeX: 2
  • EndNote: 1
Views and downloads (calculated since 20 Jun 2019)
Cumulative views and downloads (calculated since 20 Jun 2019)
Viewed (geographical distribution)  
Total article views: 143 (including HTML, PDF, and XML) Thereof 143 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
No saved metrics found.
No discussed metrics found.
Latest update: 22 Jul 2019
Publications Copernicus
Short summary
Satellite imagery can be used to map plant communities to better understand plant animal interactions at landscape scales. We used two years of Pleiades imagery to compare the stability of random forest models to map plant communities on and off prairie dog towns in South Dakota. Overall error rates was comparable between years, but a large number of pixels changed classification between separate yearly models. Plant community overlap may play a role in the accuracy of model predictions.
Satellite imagery can be used to map plant communities to better understand plant animal...