Sampling Intensities and Sampling Errors Associated with Pre-and Post-treatment Forest Restoration Monitoring: The Ute Valley Inventory

Although silvicultural prescriptions may vary depending on the ecosystems treated and the objectives of the landowner, forest restoration in the southwestern United States is often designed to reduce the potential deleterious effects of grazing and other land-use practices across ranges, stands, forests and watersheds. The Collaborative Forest Restoration Program (CFRP), administered by the USDA Forest Service (USFS) Southwestern Region in Albuquerque, is a New Mexico-wide initiative to reduce the potential of catastrophic wildfire on public and tribal lands while building collaborations and partnerships among diverse stakeholders and interest groups (USFS 2001).

Among the questions associated with forest restoration monitoring is whether the inventory performed provides information consistent with both landowner objectives and the anticipated use of inventory results. On the subject of sampling intensity, current CFRP inventory protocols: suggest using land area to determine sampling intensity (USFS 2003, Savage et al. 2006, New Mexico Forest and Watershed Restoration Institute, pers. comm.); advise establishing enough plots to “make monitoring reliable,” without further discussion or any indication of what is meant by reliable (Moote et al. 2010); or ignore sampling error completely, while at the same time stressing the importance of “good” baseline data to compare with future project monitoring data (Derr et al. 2005). Curiously, a CFRP-funded monitoring document asserted that “one quarter of the (CFRP) projects had ecological monitoring methods that were assessed as having low reliability,” yet did not describe what was meant by “reliability” or how the assessments were made (Derr et al. 2008). Further, since the terms were not mentioned, sampling intensity and sampling error did not appear to be factors in determining monitoring or data reliability.

Table 1.

Allowable errors for various forest inventory objectives.

One common CFRP project inventory protocol calls for establishing a 0.04-ha plot for every four acres of land (New Mexico Forest and Watershed Restoration Institute, pers. comm.). While this approach may work for some inventories, it is not directly sensitive to the inherent variability associated with the site and stand attributes being measured and therefore it will not be appropriate for all site conditions or inventory situations. As a result, sampling errors may be either too high or too low compared to pre-inventory, targeted allowable errors. While there is flexibility associated with the most appropriate allowable error targeted for a given inventory project, in general the decision will be based on inventory objectives, the resources available to conduct the inventory, and, in the case of timber and/or land sales, the value of the timber. For example, one set of guidelines for allowable errors for various forest inventory objectives (J. Barrett, University of New Hampshire, retired, pers. comm.) is suggested in Table 1.

In 2007, following generally accepted inventory protocols suggested by the CFRP (Derr et al. 2008), a restoration monitoring crew established 21 0.02-ha sample plots in the New Mexican Ute Valley prior to a hazardous fuels reduction treatment. In 2009, after CFRP-funded restoration treatments were applied, a crew returned to the same site, re-measuring 16 of the original 21 sample plots (five of the 2007 pre-treatment plots had not been exposed to restoration treatment, and, therefore, were not re-measured in 2009). Results indicated that the restoration treatment reduced the trees/ha from 1,808 to 148 trees/ha and the basal area from 28.7 m²/ha to 10.3 m²/ha (Table 2). Much of the treatment could be classified as a low thinning, that is, trees in smaller diameter classes were removed. Therefore, the mean stand diameter increased from 14.2 cm in the pre-treatment stand to 29.7 cm in the post-treatment stand, while average tree height increased from 12.7 m to 19.2 m (Table 2).

Importantly, the sampling error around the estimated mean basal area per hectare almost doubled in the post-treatment stand vs. pre-treatment, a reflection of the increased variability in the distribution in basal area/ha after the treatment (Table 2). This...