Optimizing sampling across transect-based methods improves the power of agroecological monitoring data

Authors: McCord, Sarah; Webb, Nicholas; Van Zee, Justin; Courtright, Ericha; DUNIWAY, MICHAEL - Us Geological Survey; EDWARDS, BRANDON - New Mexico State University; KACHERGIS, EMILY - Bureau Of Land Management; Moriasi, Daniel; MORRA, BRIAN - University Of Nevada School Of Medicine; NAFUS, ALETA - Bureau Of Land Management; Newingham, Beth; Scott, Drew; Toledo, David

Submitted to: Journal of Environmental Quality
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/19/2024
Publication Date: 3/17/2025
Citation: McCord, S.E., Webb, N.P., Van Zee, J.W., Courtright, E.M., Duniway, M.C., Edwards, B., Kachergis, E., Moriasi, D.N., Morra, B., Nafus, A., Newingham, B.A., Scott, D.A., Toledo, D.N. 2025. Optimizing sampling across transect-based methods improves the power of agroecological monitoring data. Journal of Environmental Quality. 1-4. https://doi.org/10.1002/jeq2.20678.
DOI: https://doi.org/10.1002/jeq2.20678

Interpretive Summary: Ecosystem monitoring is most effective when it is designed with research questions and monitoring objectives in mind (Fischman & Ruhl, 2016). We demonstrated that plot-scale inference is strongly affected by sample design, and consequently our understanding of ecological dynamics and the status, condition and trend of ecosystems, may be influenced by plot sampling decisions. Decisions regarding monitoring effort and plot design should be driven by 1) ecological process(es) of interest, 2) amount of monitoring resources available, 3) ecosystem spatial patterns and expected changes over time, and 4) tolerance for sampling error. Monitoring for research studies may require more intensive sampling to reach a greater confidence level (e.g., 95%) and lower sampling errors, as researchers seek to describe ecosystem processes and patterns with high precision. If greater uncertainty is acceptable and monitoring resources are limited, less intensive sampling may be appropriate. This study examines multiple monitoring methods to provide improved information to those designing monitoring studies and analyzing monitoring data to identify the appropriateness of sample design to meet monitoring objectives. Appropriate sample design will ensure that we are neither over- nor under-allocating the extensive resources needed to monitor ecosystem change.

Technical Abstract: Transect-based monitoring has long been a valuable tool in ecosystem monitoring to measure multiple ecosystem attributes. The line-point intercept (LPI), vegetation height, and canopy gap intercept methods comprise a set of core methods, which provide indicators of ecosystem condition. However, users often struggle to design a sampling strategy that optimizes the ability to detect ecological change using transect-based methods. We assessed the sensitivity of each of these core methods to transect length, number, and sampling interval in 1-ha plots to determine: (1) minimum sampling required to describe ecosystem characteristics and detect change; and (2) optimal transect length and number to make recommendations for future analyses and monitoring efforts. We used data from 13 National Wind Erosion Research Network locations, including five LTAR sites, spanning the western United States, which included 151 plot sampling events over time across five biomes. We found that longer and increased replicates of transects were more important for reducing sampling error than increased sample intensity along fewer transects per plot. For all methods and indicators across biomes plots, three 100-m transects reduced sampling error such that indicator estimates fell within a 95% confidence interval of ±5% for canopy gap intercept and LPI-total foliar cover, ±5 cm for height, and ±2 species for LPI-species counts. For the same criteria at 80% confidence intervals, two 100-m transects are needed. Site-scale inference was strongly affected by sample design, consequently our understanding of ecological dynamics may be influenced by sampling decisions.