Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → États-Unis → 2014 → Bridging structure and function in semi-arid ecosystems by integrating remote sensing and ground based measurements

University of New Mexico (2014)

Bridging structure and function in semi-arid ecosystems by integrating remote sensing and ground based measurements

Krofcheck, Dan

Titre : Bridging structure and function in semi-arid ecosystems by integrating remote sensing and ground based measurements

Auteur : Krofcheck, Dan

Université de soutenance : University of New Mexico

Grade : Doctor of Philosophy (PhD) 2014

The Southwestern US is projected to continue the current significant warming trend, with increased variability in the timing and magnitude of rainfall events. The effects of these changes in climate are already evident in the form of multi-year droughts which have resulted in the widespread mortality of woody vegetation across the region. Therefore, the need to monitor and model forest mortality and carbon dy- namics at the landscape and regional scale is an essential component of regional and global climate mitigation strategies, and critical if we are to understand how the imminent state transitions taking place in forests globally will affect climate forcing and feedbacks. Remote sensing offers the only solution to multitemporal regional observation, yet many challenges exist with employing modern remote sensing so- lutions in highly stressed vegetation characteristic of semi-arid biomes, making one of the most expansive biomes on the globe also one of the most difficult to accu- rately monitor and model. The goal of this research was to investigate how changes in the structure of semi-arid woodlands following forest mortality impacts ecosys- tem function, and to determine how this question can be addressed using remotely sensed data sets. I focused primarily on Pinus edulis and Juniperous monosperma (pi ̃ non-juniper) woodlands, and took advantage of an existing manipulation experi- ment where mortality was imposed on all of the large pi ̃ non (¡ 7 cm dbh) in a 4 ha PJ woodland in 2009 and the ecosystem functional responses have been quantified using eddy covariance. A nearby intact PJ woodland, also instrumented with eddy covariance, was used as a control for this experiment. I tested the ability of high resolution remote sensing data to mechanistically describe the patterns in overstory mortality and understory green-up in this manipulated woodland by comparing it to the intact woodland, and observed the heterogeneous response of the understory as a function of cover type. I also investigated the relationship between changes in soil water content and the greenness of the canopy, noting that in the disturbed woodland, I observed a decoupling between how the canopy was measured remotely (e.g., via vegetation indices, VI) and photosynthesis. This is significant in that it potentially represents a significant source of error in using existing light use efficiency models of carbon uptake in these disturbed woodlands. This research also suggested that leveraging remote sensing data which measures in the red-edge portion of re- flected light can provide increased sensitivity to the low leaf area, ephemeral pulses of greenup that were identified in the disturbed woodland, post-canopy mortality. Given these findings, I developed a hierarchy of simple linear models to test how well vegetation indices acquired through different spatial resolution sensors (Land- sat and RapidEye) were able to predict carbon uptake in both intact and disturbed pi ̃ non-juniper woodlands. The vegetation indices used were a moisture sensitive VI, and a red-edge leveraging VI from these sensors, and I compared estimates of carbon uptake derived from these models to the Gross Primary Productivity estimated from tower-based eddy covariance at both the manipulated and intact pi ̃ non-juniper sites. I determined that the red-edge VI and the moisture sensitive VI both constrained uncertainty associated with carbon uptake, but that the variability in satellite view angle from scene to scene can impose a significant amount of noise in sparse canopy ecosystems. Finally, given the extent and prevalence of J. monosperma across the region, and its complex growth morphology, I tested the ability of aerial lidar to quantify the biomass of juniper. In this simplified case study, I developed a method- ology to relate the volume of canopy measured via lidar to the equivalent stem area at the root crown. By working in a single species ecosystem, I circumvented many challenges associated with driving allometries remotely, but also present a work-flow that I intend to adapt to more complex systems, namely pi ̃ non-juniper woodlands. Together, this work describes and addresses existing challenges with respect to us- ing remote sensing to understand both the structure and function of pi ̃ non-juniper woodlands, and how it changes in response to widespread pi ̃ non mortality. It provides several new techniques to mitigate the difficulties associated with monitoring mor- tality / recovery dynamics, predicting canopy function, and determining ecosystem state parameters in these complex, sensitive biomes

Mots clés : Remote Sensing GPP Eddy-Covariance Vegetation Structure


Version intégrale (4,36 Mb)

Page publiée le 4 mars 2017, mise à jour le 9 avril 2019