Regression tree modeling of spatial pattern and process interactions

Trisalyn Nelson, Wiebe Nijland, Mathieu L. Bourbonnais, Michael A. Wulder

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

In forestry, many fundamental spatial processes cannot be measured directly and data on spatial patterns are used as a surrogate for studying processes. To characterize the outcomes of a dynamic process in terms of a spatial pattern, we often consider the probability of certain outcomes over a large area rather than on the scale of the particular process. In this chapter we demonstrate data mining approaches that leverage the growing availability of forestry-related spatial data sets for understanding spatial processes. We present classification and regression trees (CART) and associated methods, including boosted regression trees (BRT) and random forests (RT). We demonstrate how data mining or machine learning approaches are useful for relating spatial patterns and processes. Methods are applied to a wildfire data and covariate data are used to contextualize the quantified patterns. Results indicate that fire patterns are mostly related to processes influenced by people. Given the growing number of multi-temporal and large area datasets on forests and ecology machine learning and data mining approaches should be leveraged to quantify dynamic space-time relationships.

Original languageEnglish (US)
Title of host publicationMapping Forest Landscape Patterns
PublisherSpringer New York
Pages187-212
Number of pages26
ISBN (Electronic)9781493973316
ISBN (Print)9781493973293
DOIs
StatePublished - Sep 7 2017

Fingerprint

Data Mining
data mining
Data mining
Forestry
artificial intelligence
Learning systems
forestry
modeling
spatial data
Ecology
wildfires
wildfire
Fires
Availability
ecology
methodology
machine learning
method
Datasets
Machine Learning

ASJC Scopus subject areas

  • Agricultural and Biological Sciences(all)
  • Environmental Science(all)
  • Engineering(all)

Cite this

Nelson, T., Nijland, W., Bourbonnais, M. L., & Wulder, M. A. (2017). Regression tree modeling of spatial pattern and process interactions. In Mapping Forest Landscape Patterns (pp. 187-212). Springer New York. https://doi.org/10.1007/978-1-4939-7331-6_5

Regression tree modeling of spatial pattern and process interactions. / Nelson, Trisalyn; Nijland, Wiebe; Bourbonnais, Mathieu L.; Wulder, Michael A.

Mapping Forest Landscape Patterns. Springer New York, 2017. p. 187-212.

Research output: Chapter in Book/Report/Conference proceedingChapter

Nelson, T, Nijland, W, Bourbonnais, ML & Wulder, MA 2017, Regression tree modeling of spatial pattern and process interactions. in Mapping Forest Landscape Patterns. Springer New York, pp. 187-212. https://doi.org/10.1007/978-1-4939-7331-6_5
Nelson T, Nijland W, Bourbonnais ML, Wulder MA. Regression tree modeling of spatial pattern and process interactions. In Mapping Forest Landscape Patterns. Springer New York. 2017. p. 187-212 https://doi.org/10.1007/978-1-4939-7331-6_5
Nelson, Trisalyn ; Nijland, Wiebe ; Bourbonnais, Mathieu L. ; Wulder, Michael A. / Regression tree modeling of spatial pattern and process interactions. Mapping Forest Landscape Patterns. Springer New York, 2017. pp. 187-212
@inbook{750d292e47cf4518959d301c3192f1fe,
title = "Regression tree modeling of spatial pattern and process interactions",
abstract = "In forestry, many fundamental spatial processes cannot be measured directly and data on spatial patterns are used as a surrogate for studying processes. To characterize the outcomes of a dynamic process in terms of a spatial pattern, we often consider the probability of certain outcomes over a large area rather than on the scale of the particular process. In this chapter we demonstrate data mining approaches that leverage the growing availability of forestry-related spatial data sets for understanding spatial processes. We present classification and regression trees (CART) and associated methods, including boosted regression trees (BRT) and random forests (RT). We demonstrate how data mining or machine learning approaches are useful for relating spatial patterns and processes. Methods are applied to a wildfire data and covariate data are used to contextualize the quantified patterns. Results indicate that fire patterns are mostly related to processes influenced by people. Given the growing number of multi-temporal and large area datasets on forests and ecology machine learning and data mining approaches should be leveraged to quantify dynamic space-time relationships.",
author = "Trisalyn Nelson and Wiebe Nijland and Bourbonnais, {Mathieu L.} and Wulder, {Michael A.}",
year = "2017",
month = "9",
day = "7",
doi = "10.1007/978-1-4939-7331-6_5",
language = "English (US)",
isbn = "9781493973293",
pages = "187--212",
booktitle = "Mapping Forest Landscape Patterns",
publisher = "Springer New York",
address = "United States",

}

TY - CHAP

T1 - Regression tree modeling of spatial pattern and process interactions

AU - Nelson, Trisalyn

AU - Nijland, Wiebe

AU - Bourbonnais, Mathieu L.

AU - Wulder, Michael A.

PY - 2017/9/7

Y1 - 2017/9/7

N2 - In forestry, many fundamental spatial processes cannot be measured directly and data on spatial patterns are used as a surrogate for studying processes. To characterize the outcomes of a dynamic process in terms of a spatial pattern, we often consider the probability of certain outcomes over a large area rather than on the scale of the particular process. In this chapter we demonstrate data mining approaches that leverage the growing availability of forestry-related spatial data sets for understanding spatial processes. We present classification and regression trees (CART) and associated methods, including boosted regression trees (BRT) and random forests (RT). We demonstrate how data mining or machine learning approaches are useful for relating spatial patterns and processes. Methods are applied to a wildfire data and covariate data are used to contextualize the quantified patterns. Results indicate that fire patterns are mostly related to processes influenced by people. Given the growing number of multi-temporal and large area datasets on forests and ecology machine learning and data mining approaches should be leveraged to quantify dynamic space-time relationships.

AB - In forestry, many fundamental spatial processes cannot be measured directly and data on spatial patterns are used as a surrogate for studying processes. To characterize the outcomes of a dynamic process in terms of a spatial pattern, we often consider the probability of certain outcomes over a large area rather than on the scale of the particular process. In this chapter we demonstrate data mining approaches that leverage the growing availability of forestry-related spatial data sets for understanding spatial processes. We present classification and regression trees (CART) and associated methods, including boosted regression trees (BRT) and random forests (RT). We demonstrate how data mining or machine learning approaches are useful for relating spatial patterns and processes. Methods are applied to a wildfire data and covariate data are used to contextualize the quantified patterns. Results indicate that fire patterns are mostly related to processes influenced by people. Given the growing number of multi-temporal and large area datasets on forests and ecology machine learning and data mining approaches should be leveraged to quantify dynamic space-time relationships.

UR - http://www.scopus.com/inward/record.url?scp=85037061957&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85037061957&partnerID=8YFLogxK

U2 - 10.1007/978-1-4939-7331-6_5

DO - 10.1007/978-1-4939-7331-6_5

M3 - Chapter

AN - SCOPUS:85037061957

SN - 9781493973293

SP - 187

EP - 212

BT - Mapping Forest Landscape Patterns

PB - Springer New York

ER -