Search for: All records

Summary The Jornada Basin Long‐Term Ecological Research Site (JRN‐LTER, or JRN) is a semiarid grassland–shrubland in southern New Mexico, USA. The role of intraspecific competition in constraining shrub growth and establishment at the JRN and in arid systems, in general, is an important question in dryland studies.Using information on shrub distributions and growth habits at the JRN, we present a novel landscape‐scale (c. 1 ha) metric (the ‘competition index’, CI), which quantifies the potential intensity of competitive interactions. We map and compare the intensity of honey mesquite (Prosopis glandulosa, Torr.) competition spatially and temporally across the JRN‐LTER, investigating associations of CI with shrub distribution, density, and soil types.The CI metric shows strong correlation with values of percent cover. Mapping CI across the Jornada Basin shows that high‐intensity intraspecific competition is not prevalent, with few locations where intense competition is likely to be limiting further honey mesquite expansion.Comparison of CI among physiographic provinces shows differences in average CI values associated with geomorphology, topography, and soil type, suggesting that edaphic conditions may impose important constraints on honey mesquite and growth. However, declining and negative growth rates with increasing CI suggest that intraspecific competition constrains growth rates when CI increases abovec. 0.5. 

1. The ecological and environmental science communities have embraced machine learning (ML) for empirical modelling and prediction. However, going beyond prediction to draw insights into underlying functional relationships between response variables and environmental ‘drivers’ is less straightforward. Deriving ecological insights from fitted ML models requires techniques to extract the ‘learning’ hidden in the ML models. 2. We revisit the theoretical background and effectiveness of four approaches for deriving insights from ML: ranking independent variable importance (Gini importance, GI; permutation importance, PI; split importance, SI; and conditional permutation importance, CPI), and two approaches for inference of bivariate functional relationships (partial dependence plots, PDP; and accumulated local effect plots, ALE). We also explore the use of a surrogate model for visualization and interpretation of complex multi-variate relationships between response variables and environmental drivers. We examine the challenges and opportunities for extracting ecological insights with these interpretation approaches. Specifically, we aim to improve interpretation of ML models by investigating how effectiveness relates to (a) interpretation algorithm, (b) sample size and (c) the presence of spurious explanatory variables. 3. We base the analysis on simulations with known underlying functional relationships between response and predictor variables, with added white noise and the presence of correlated but non-influential variables. The results indicate that deriving ecological insight is strongly affected by interpretation algorithm and spurious variables, and moderately impacted by sample size. Removing spurious variables improves interpretation of ML models. Meanwhile, increasing sample size has limited value in the presence of spurious variables, but increasing sample size does improves performance once spurious variables are omitted. Among the four ranking methods, SI is slightly more effective than the other methods in the presence of spurious variables, while GI and SI yield higher accuracy when spurious variables are removed. PDP is more effective in retrieving underlying functional relationships than ALE, but its reliability declines sharply in the presence of spurious variables. Visualization and interpretation of the interactive effects of predictors and the response variable can be enhanced using surrogate models, including three-dimensional visualizations and use of loess planes to represent independent variable effects and interactions. 4. Machine learning analysts should be aware that including correlated independent variables in ML models with no clear causal relationship to response variables can interfere with ecological inference. When ecological inference is important, ML models should be constructed with independent variables that have clear causal effects on response variables. While interpreting ML models for ecological inference remains challenging, we show that careful choice of interpretation methods, exclusion of spurious variables and adequate sample size can provide more and better opportunities to ‘learn from machine learning’. 

Abstract Dryland vegetation is influenced by biotic and abiotic land surface template (LST) conditions and precipitation (PPT), such that enhanced vegetation responses to periods of high PPT may be shaped by multiple factors. High PPT stochasticity in the Chihuahuan Desert suggests that enhanced responses across broad geographic areas are improbable. Yet, multiyear wet periods may homogenize PPT patterns, interact with favorable LST conditions, and in this way produce enhanced responses. In contrast, periods containing multiple extreme high PPT pulse events could overwhelm LST influences, suggesting a divergence in how climate change could influence vegetation by altering PPT periods. Using a suite of stacked remote sensing and LST datasets from the 1980s to the present, we evaluated PPT‐LST‐Vegetation relationships across this region and tested the hypothesis that enhanced vegetation responses would be initiated by high PPT, but that LST favorability would underlie response magnitude, producing geographic differences between wet periods. We focused on two multiyear wet periods; one of above average, regionally distributed PPT (1990–1993) and a second with locally distributed PPT that contained two extreme wet pulses (2006–2008). 1990–1993 had regional vegetation responses that were correlated with soil properties. 2006–2008 had higher vegetation responses over a smaller area that were correlated primarily with PPT and secondarily to soil properties. Within the overlapping PPT area of both periods, enhanced vegetation responses occurred in similar locations. Thus, LST favorability underlied the geographic pattern of vegetation responses, whereas PPT initiated the response and controlled response area and maximum response magnitude. Multiyear periods provide foresight on the differing impacts that directional changes in mean climate and changes in extreme PPT pulses could have in drylands. Our study shows that future vegetation responses during wet periods will be tied to LST favorability, yet will be shaped by the pattern and magnitude of multiyear PPT events.