Literature searches
The initial literature search will be conducted in English in the Web of Science with the aim of answering the primary research question. The terms will be searched in titles, abstracts and keywords and will be based on the Disturbance and the Intervention. The results will not be date-restricted but will be limited to the fields of Environmental Sciences and Ecology/Forestry/Biodiversity Conservation/Zoology/Plant Sciences/Meteorology and Atmospheric Sciences/Entomology/Water Resources.
To find other publications, including grey literature, complementary searches will be performed with simplified Disturbance and Intervention terms. These searches will be made in the Directory of Open Access Journals (https://doaj.org/) and the CABI database of forest science (http://www.cabi.org/forestscience/).
Specific searches will be conducted using the websites of the Canadian Forest Service (http://cfs.nrcan.gc.ca/publications) and US Forest Service (http://www.treesearch.fs.fed.us/).
Searches using Google Scholar, also including Disturbance and Intervention terms, will be made in English, Spanish, and German, and the first 100 hits will be screened.
As complementary bibliographic searches, the reference lists of relevant articles will be screened, and articles citing them will be searched in the Web of Science to complement the list. If necessary, authors of relevant articles will also be contacted to clarify study designs or provide additional data. A list of the publications will be sent to all the authors of the review here proposed and to other experts in the field to assess its completeness.
Initial scoping exercise
A preliminary search was performed on 9 March 2015 in the Web of Science. The initial terms considered were:
-
(a)
Disturbance terms = (disturb* OR wildfire OR fire OR windthrow OR storm OR pest ((insect* OR beetle*) AND (outbreak OR attack)))
-
(b)
Intervention terms = (salvag* OR logg* OR harvest*)
-
(c)
Outcome terms = (ecosystem OR environment* OR pollin* OR dispers* OR carbon OR sequestr* OR sink OR pest OR invasive* OR water OR purif* OR climate OR weather OR waste OR flood* OR erosion* OR air)
Searching the Disturbance terms alone yielded 416,471 hits. Searching Disturbance + Intervention terms (strings joined with the Boolean operator AND) resulted in 25,664 hits. A further search including Disturbance + Intervention + Response terms still yielded an excessive 17,580 hits. Thus, after this preliminary search we refined the intervention terms to include only logging that happens after disturbance, i.e. “salvage logging”, “salvaging”, or post-(fire/disturbance/…) logging/harvesting:
Intervention terms = ((salvag* OR post*) AND (logging OR harvest*))
This search provided 1,423 hits. As the list of ecosystem services needs to be refined according to what is found in the literature, it would be desirable to obtain a comprehensive list of the studies on salvage logging to avoid missing any study. Thus, after eliminating the restrictions resulting from the inclusion of Outcome terms, we got 1,845 hits. After screening some titles and obtaining suggestions during peer review we decided to include the terms “snag remov*”, “cutting” and “felling” in the Intervention terms. In this way, on 5 June 2015 we obtained a total of 5,039 hits.
Search string
After the scoping exercise described above, the final search string used is as follows:
(disturb* OR fire OR wildfire OR windthrow OR storm OR pest ((insect* OR beetle*) AND (outbreak OR attack)))
AND
(“snag remov*” OR ((salvag* OR post*) AND (logging OR harvest* OR cutting OR felling)))
Study inclusion criteria
To be considered for the review, studies must be empirical and fulfil all of the following inclusion criteria:
-
(a)
Relevant population: forest after one of: fire, insect outbreak, or storm disturbance (including both natural and anthropogenic causes).
-
(b)
Relevant intervention: salvage logging.
-
(c)
Relevant comparator: disturbed forest without (or prior to) salvage logging.
-
(d)
Relevant outcome: response variable that can be regarded as and indicator or used as a proxy of a regulating or supporting ecosystem service.
Modelling and review articles will be recorded as additional materials.
Article screening
The relevance of the articles resulting from the literature searches will be assessed through a stepwise elimination procedure. The articles will be screened in the following steps:
-
1.
All the titles will be read in the first step, and articles with irrelevant titles will be discarded. This step will be carried out in a conservative way to avoid discarding any potentially relevant publications, so that in doubt of the relevance of an article it will be kept (e.g. articles on post-fire ecosystem dynamics will be saved, as they might include salvage logging as a factor without it being reflected in the title). Before screening all the titles, two members of the review team will screen a subset of titles and the difference in outcomes will be assessed through a kappa test. If necessary, the inclusion criteria will be discussed again prior to screening all the titles. Kappa tests will also be conducted in steps 2 and 3 below.
-
2.
The abstracts of articles with relevant titles will be read in the second step, and articles with irrelevant abstracts will be discarded. To be classified as relevant in this step, the abstracts must fulfil all of the inclusion criteria (a), (b), and (c). If in doubt about the relevance of a publication, it will be kept for the next step.
-
3.
The articles with relevant abstracts will be read in full. At this stage, articles failing to fulfil one of the study inclusion criteria will be discarded. To find studies that fulfil inclusion criterion (d), the main objectives of the studies will be assessed as well as the study-site descriptions (including tables and figures). Relevant articles will be categorised according to the study quality assessment criteria defined below. A list of articles excluded at this stage, together with the reason for exclusion, will be provided as additional material.
Study quality assessment
The retrieved studies that pass the above steps of article screening will be appraised (1) to decide on their inclusion in the narrative part of the synthesis and/or the meta-analysis on the basis of their susceptibility to bias, and (2) to consider possible lines of action among included studies to minimise the effect of potential bias in the original studies on the conclusions of the review. For this, the studies will be included in one of the following three broad categories (which may be revised according to the standards found in the retrieved literature):
-
1.
Empirical studies with treatments applied at appropriate spatial scales and with replication (at least three replicate units) will be used for the meta-analysis and the narrative synthesis. We will perform sensitivity analyses to test for the potential effects of study quality. For example, we expect many studies to make use of areas where the intervention was applied for silvicultural rather than experimental motivations, thus lacking randomisation, and sensitivity analyses will aim to show whether including or excluding such studies yields similar results. Other characteristics to be noted for each study and considered for sensitivity analyses will be whether the methods are clearly stated and repeatable (including a clear description of the disturbance and the salvage intervention) and confounding factors are properly dealt with.
-
2.
Empirical studies with experimental flaws—e.g. due to the lack of replication—may be included for the narrative synthesis only.
-
3.
Empirical studies whose conclusions are compromised due to strong experimental flaws will not be considered for the review.
Potential effect modifiers and reasons for heterogeneity
We expect heterogeneous responses to our primary research question. Variation in site characteristics, disturbance properties, salvage logging practice, and response variables/measurements across studies should lead to considerable differences in the direction and magnitude of effects. The main potential effect modifiers (or covariates) have been identified according to the reviewer team’s knowledge of the subject and are listed below under “Data extraction strategy”. However, this list is not necessarily exhaustive, and new modifiers may be added during the process of acquisition and classification of publications. The existence and magnitude of the effect of these modifiers will be quantitatively assessed through meta-analytical statistics.
Data extraction strategy
The following information from the relevant literature will be extracted and stored in Excel spread sheets. The title, journal, author, year, and methodological approach of all the relevant articles will be stored in one file. Another file will include the data for quantitative analysis. Each row will constitute one data point, and several data points might arise from a single study (e.g. due to 2 years of sampling). The columns will contain the numerical data for the response variable in intervened and control areas (mean, standard deviation, and number of replicates) as well as the meta-data (which will be assessed as effect modifiers). For the numbers obtained for the response variables we will always take care of potential pseudoreplication, i.e. subsampling that does not constitute true replication. The meta-data will include:
-
Study site (which will be used as a random effect in mixed effects models)
-
Country
-
Forest biome (according to [20])
-
Forest type (conifer, broadleaf, mixed)
-
Disturbance type (fire, windthrow, insect outbreak)
-
Disturbance intensity
-
Years after disturbance
-
Years after logging
-
The method of logging/wood extraction (helicopter, forwarder, winching, mules, etc.)
-
Intensity of salvage logging
-
Time of the year of logging
-
Additional post-disturbance actions (e.g. planting)
-
Mean sampling-plot area
-
Study quality: clarity in the methods (specified as a percentage), proper handling of confounding effects (yes/no), and randomisation (yes/no).
We will also register the category of ecosystem service to perform complementary analyses separated by ecosystem service type. We will make an effort to obtain all these meta-data for each data point; authors will be contacted to clarify any missing information if necessary.
Risk of bias
The potential for bias between studies will be assessed in complementary ways, as follows. First, geographical and disturbance-related bias (e.g. most studies coming from Europe or North America, which is what we expect) will be assessed by quantifying the studies falling into the different categories of disturbance type, forest type, time since disturbance, and geographical region, and the results will be discussed. Second, bias in quantitative analysis due to more than one data point coming from the same study area will be handled through random effects in mixed effects models. And third, publication bias will be tackled both prior to analysis, through contacting researchers to provide unpublished data, and after analysis, with the trim and fill method [21].
Data synthesis and presentation
The data obtained during the review process will be presented in both narrative and quantitative ways. The narrative part will make use of conceptual diagrams and tables to describe the studies that were found (e.g. the number of studies of salvage logging after each kind of disturbance affecting different ecosystem services).
The quantitative data extracted from empirical studies will be assessed with statistical methods through quantitative meta-analysis. The meta-analysis will be formulated so as to answer the primary and all the secondary questions. The results will be presented in the form of figures and tables and thoroughly explained and discussed in the text.