Back to search

South Dakota State University

Evaluation of an Empirical Piecewise Regression Model in Simulating GPP in the Northern Great Plains

Abstract

dc:description.abstract

An empirical piecewise regression (PWR) model was developed to estimate gross primary production (GPP) and improve the understanding of carbon fluxes for the grassland ecosystems in the Northern Great Plains. The PWR model spatially scales up the localized flux tower measurements across the grassland ecoregion at 1-km resolution. In this study, cross-validation was used to evaluate the robustness of the PWR model in the Northern Great Plains. The results showed that the PWR modeling approach was robust with a good agreement (index of agreement d = 0.71-0.97) between the PWR GPP and tower-measured GPP by withholding site, and a good agreement (d = 0.86-0.91) by withholding year. The PWR model was developed from five Northern Great Plain flux towers to predict GPP in the ecoregions. The PWR GPP and MODIS GPP were then compared with the tower-measured GPP. The results showed a good agreement of GPP among PWR, MODIS, and tower measurements at the Fort Peck, Mandan, and Cheyenne site. The MODIS GPP, however, did not agree well with the tower measurements at the Miles City and Lethbridge sites (d = 0.62-0.79). Differences between MODIS GPP and tower measurements at the Miles City and Lethbridge implied that the MODIS GPP failed to capture the seasonal dynamics of the growing season and locally overestimated or underestimated the tower-measured GPP at the two sites. Those discrepancies may be attributed to three potential problems: 1) pixel misregistration, 2) flux tower measurements, and 3) model estimates. The GPP spatial maps from the PWR and MODIS models were also compared for grasslands for the entire study area. The PWR GPP was lower than, or similar to, the MODIS GPP in the east and higher in the west and south. Environmental factors that may contribute to the spatial patterns of the GPP differences between the two models were then evaluated using a decision tree technique. The results of the decision tree analysis suggested that percentage of C4 grasses, soil water holding capacity, percentage of clay, and percentage of cropland mixed in the grassland contributed to the GPP difference patterns of the PWR and MO DIS models. Keywords: Carbon cycle; Decision tree; Gross primary production (GPP); MODIS GPP; Model evaluation; Northern Great Plains; Remote sensing

Degree

thesis:*
Name thesis:degree_name
Master of Science (MS)
Level thesis:degree_level
Thesis - University Access Only
Discipline thesis:degree_discipline
Geography

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Zhang, Li

Identifiers

dc:identifier.*
Repository record dc:identifier
https://openprairie.sdstate.edu/etd2/1222

Chain of custody

source
Harvested from
South Dakota State University
Base URL
openprairie.sdstate.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Zhang, Li. Evaluation of an Empirical Piecewise Regression Model in Simulating GPP in the Northern Great Plains. Thesis - University Access Only thesis, https://openprairie.sdstate.edu/etd2/1222