Vegetation Dynamics of Semiarid Rangelands in Eastern New Mexico

Authors: Sarah J. Turner

The shortgrass prairie ecosystem of North America has faced extensive land conversion since colonization. Vegetation community composition changes over time due to natural and anthropogenic influences and it is often difficult to determine the direction of change within these highly utilized rangelands. Many land management agencies use state-and-transition models to characterize compositional shifts due to weather and management activities, though long-term data are lacking to quantify management influences and validate these models.

This work analyzes long-term vegetation composition data collected using the Land Condition Trend Analysis method on Melrose Air Force Range, a 28,218-hectare military training installation in eastern New Mexico. All resources excluding survey data utilized in this work were from open resources to derive methods for the common land manager. I used traditional ground-based sampling methods and remote sensing technology to detect the magnitude and direction of vegetation composition change over a 13-year time span. I utilized multivariate ordination techniques to determine the gradients of species associations and to identify spatiotemporal compositional shifts. Second, I used multivariate ordination techniques to derive a link between spectral reflectance and species composition using Landsat 7 ETM and Landsat 8 OLI imagery and subsequently conducted an unsupervised classification to detect community-level changes across a 6-year time frame. Finally, I evaluated the effectiveness of the Land Condition Trend Analysis method to meet the monitoring needs on Melrose Air Force Range by calculating the effect size of spatiotemporal comparisons followed by a suite of power analyses using native, invasive, and exotic species.

I found that soil and weather gradients constrained 47% of the variance in transect vegetation composition. This study indicated that though vegetation composition at the transect-level varied annually and spatially across the 13-year period, there was no evidence of a unidirectional composition shift to infer irreversible vegetation change. I found that at a 30-meter resolution, unconstrained ordination techniques were better suited than constrained techniques to differentiate spectral reflectance values. An unsupervised classification across 2015 and 2021 indicated minimal shrubland conversion across Melrose Air Force Range but no change in the dominant graminoid species, supporting an overall stable vegetation state. Finally, I found that the current implementation of the Land Condition Trend Analysis on Melrose Air Force Range meets the requirement of installation-level vegetation monitoring but lacks statistical power to compare composition changes across smaller, management-scale units. Likewise, my results indicate that this method may not be the best suited for monitoring rare or aggregated species.

Overall, this effort provides evidence to support the state-and-transition model theory of vegetation dynamics within the shortgrass prairie of eastern New Mexico, shows the difficulty in linking transect-level and spectral data, and highlights the need for sampling designs based on well-defined management objectives. The methods outlined can support management actions on other semiarid rangelands and can help further inform vegetation succession theories.