Tree-species mixture effects (e.g., complementarity and facilitation) have been found to increase individual-tree productivity, lessen mortality, and improve recruitment in forests worldwide. By promoting more efficient and complete resource use, mixture effects may also lessen individual-tree-level water stress, thus improving drought-resistance. We investigated the influence of mixture effects on tree productivity, mortality, and recruitment across broad compositional and moisture gradients in high-elevation Interior West US mixed-conifer communities, where Populus tremuloides (trembling aspen) is the major contributor to functional diversity. Our goal was to provide a more complete scientific foundation for managing these drought-prone, fire-excluded systems under an uncertain climate.

We used landscape-scale national forest inventory data to examine mixture effects on P. tremuloides and the major associated conifer species, Pseudotsuga menziesii, Pinus contorta, Abies lasiocarpa, and Picea engelmannii. Using generalized linear mixed modeling, we isolated the influences of P. tremuloides relative density and climate on tree-level (stems ≥ 12.7 cm DBH) growth, mortality, and stand-level recruitment (presence/absence of new trees). Cold-season precipitation (PPT) and warm-season vapor pressure deficit (VPD) served to represent soil moisture supply and demand, respectively.

Populus tremuloides growth declined as interspecific density increased. In contrast, Pinus contorta and A. lasiocarpa growth increased with P. tremuloides density. For all species except A. lasiocarpa and P. menziesii, growth increased under higher PPT and VPD. Populus tremuloides mortality increased under high VPD but not with interspecific relative density. We found limited evidence that A. lasiocarpa mortality decreased as P. tremuloides density increased. Populus tremuloides recruitment declined steeply above 25% interspecific relative density. We found a decline in conifer recruitment odds as P. tremuloides density increased, ranging from strong in P. contorta to insubstantial in P. engelmannii.

Synthesis. Our findings have implications for sustaining mixed-conifer communities impacted by climate change and historical fire exclusion. Mixtures of P. tremuloides and conifers may improve conifer growth while adversely impacting P. tremuloides growth relative to pure stands. Higher conifer productivity combined with lower P. tremuloides recruitment odds at conifer relative density may accelerate succession.

The primary dataset for this study were collected through the United States Department of Agriculture, Forest Serivce, Forest Inventory and Analysis Program. These data were collected under a common sampling and measurement protocol, as fully detailed in this agency document valid through the date of database queries: https://www.fs.fed.us/rm/ogden/data-collection/pdf/P2%20Manual_70_Feb2sm.pdf. These data are part of the Public Domain and freely available. Because the FIA database is a dataset periodically updated with new plot measurements, we based on study on data current as of September 2019.

We have provided a workflow for querying FIA plots, applicable both to individual U.S. states as well as the aggregated regional dataset. An SQL script was designed and tested for use with postgreSQL, but originally developed for single-state queries via MS Access. Many researchers may opt instead to query and summarize the FIA Database with the rFIA R package (Stanke and Finley, 2020; https://cran.r-project.org/web/packages/rFIA/index.html), which does not require SQL querying or other database skills. Please see the link to this R package in "Related Works."

We have also provided the R script for building the analysis files based on our flattened data query. This script details our data filtering steps, such as dropping distributed or multi-stand plots, not performed in the SQL query. This annotated file details the process metadata used to calculate derived data fields such as competition index and species composition matrices. We have also provided script for integrating climatic data layers.

Climate data (vapor pressure deficit and precipitation) were built using raster functions in QGIS v. 3.2. The original PRISM data may be freely accessed via https://prism.oregonstate.edu/recent/. Climatic water deficit data used in the supplementary material may be freely accessed via Terraclim. Both PRISM and Terraclim are cited in "Related Works".   

Lastly, performing the same query on more up-to-date versions of the FIA database will yield subtly different results. The FIA dataset is continuously updated, correcting minor errors as well as adding additional plot measurements. Based on our criteria, disturbance, harvesting, or conversion to non-forest land would be grounds for dropping plots from inclusion in analysis. Because our climatic data are also dynamic, the same analysis repeated in subsequent years may also yield different results for a given region (for example, southwestern Colorado) depending on recent drought history.

In some cases queries may return missing data values for FIA tree records, such as tree height or diameter. In postgreSQL database format, the Interior West FIA Database has blank fields for trees dead as of the most current plot measurement. While dead trees were used for mortality analyses, currently living trees were used for growth and recruitment analyses. Occasional missing values may occur for site class, initial diameter, or stand age. These trees were dropped from modeling.

The SQL script is inefficient (John Shaw, personal communication), as an up-front "WHERE" statement would save processing time and greatly reduce file size. Difficulties coding SQL 'WHERE" statements movitated us to filter data in R using subsetting and table joins.

Forest Inventory and Analysis data are collected and maintained in U.S. units. For the purposes of this study, conversion to metric was only strictly necessary for Methods/Results descriptions as well as graphical summaries. For those cases, conversions of responses such as volume increment were performed in scripts used to generate graphics. Climatic data are in metric units.