Environmental impact assessment for large carnivores: a methodological review of the wolf (Canis lupus) monitoring in Portugal

The continuous growth of the global human population results in increased use and change of landscapes, with infrastructures like transportation or energy facilities, being a particular risk to large carnivores. Environmental Impact Assessments were established to identify the probable environmental consequences of any new proposed project, find ways to reduce impacts, and provide evidence to inform decision making and mitigation. Portugal has a wolf population of around 300 individuals, designated as an endangered species with full legal protection. They occupy the northern mountainous areas of the country which has also been the focus of new human infrastructures over the last 20 years. Consequently, dozens of wolf monitoring programs have been established to evaluate wolf population status, to identify impacts, and to inform appropriate mitigation or compensation measures. We reviewed Portuguese wolf monitoring programs to answer four key questions: do wolf programs examine adequate biological parameters to meet monitoring objectives? is the study design suitable for measuring impacts? are data collection methods and effort sufficient for the stated inference objectives? and do statistical analyses of the data lead to robust conclusions? Overall, we found a mismatch between the stated aims of wolf monitoring and the results reported, and often neither aligns with the existing national wolf monitoring guidelines. Despite the vast effort expended and the diversity of methods used, data analysis makes almost exclusive use of relative indices or summary statistics, with little consideration of the potential biases that arise through the (imperfect) observational process. This makes comparisons of impacts across space and time difficult and is therefore unlikely to contribute to a general understanding of wolf responses to infrastructure-related disturbance. We recommend the development of standardized monitoring protocols and advocate for the use of statistical methods that account for imperfect detection to guarantee accuracy, reproducibility, and efficacy of the programs.

Methods
We reviewed all major wolf monitoring programs developed for environmental impact assessments in Portugal since 2002 (Table S1, Supplementary material). Given that the focus here is on the adequacy of targeted wolf monitoring for delivering conclusions about the effects of infrastructure development, we reviewed only monitoring programs that were specifically designed for wolves and not those concerned with general mammalian assessment.

The starting point was a compilation from the 2019-2021 National Wolf Census (Pimenta et al., 2023), where every wolf monitoring program that occurred between 2014 and 2019 in Portugal was identified. The list was completed with projects that started before 2014 or after 2019 based on personal knowledge, inquires to principal scientific teams, governmental agencies, and EIA consultants. Depending on duration, wolf monitoring programs can produce several, usually annual, reports that are not peer-reviewed and do not appear on standard search engines (e.g., Web of Science or Google Schoolar) but are publicly available from the Portuguese Environmental Agency (APA – www.apambiente.pt). We conducted an online search on APA´s search engine (https://siaia.apambiente.pt/) and identified a total of 30 projects. For each of these projects, we were interested in the first and the last report to identify any methodological changes. If the last report was not present, we reviewed the most recent one. If no report was present, we requested it from the team responsible.

Our investigation centred on characterizing and quantifying four components of wolf monitoring programs that are interlinked and that should be ideally determined by the initial objectives: (1) biological parameters, i.e., what wolf parameters were studied to assess impacts; (2) study design, i.e., what sampling schemes were followed to collect and analyse data; (3) data collection, i.e., which sampling methodology and how much effort was used to collect data; and (4) data analysis, i.e., how data were analysed to estimate relevant parameters and assess impact.

Biological parameters were identified and classified under two categories: occurrence and demography, which broadly correspond to the necessary inputs to assess impacts like exclusion effect and changes in reproductive patterns. Occurrence-related parameters refer to variables used to measure the presence or absence of wolves, whereas demographic parameters refer to variables that intend to measure population-level effects such as abundance, density, survival, or reproduction. We also recorded whether any effort was made to quantify prey population distribution or abundance as recommended in the guidelines.

For study design, we reviewed the sampling design of the project, with specific focus on the spatial and temporal aspect of the study such as total area surveyed, the definition of a sampling site within this region (i.e., resolution), the duration of the study and the number of sampling seasons. The goal here was to determine whether the sampling scheme used was appropriate for assessing infrastructure impacts on wolf distribution or demography, depending on what the focus was.

For data collection, we identified the main data collection methodologies used and the corresponding sampling effort. By far the most frequent method used is sign surveys, and specifically scat surveys, and for these studies we recorded whether genetic identification of species or individuals based on faecal DNA was attempted. We compare how sampling effort varies by the various inference objectives and, as above, assess which, if any, project or data collection approach is most likely to produce evidence of impact.

We divided the Analysis component into two groups: single-year and multi-year analyses. For single-year analysis we identified how monitoring projects used data to make inferences about the state biological parameters of interest and discuss the associated strengths and weaknesses. For multi-year analyses, we recorded how differences or trends were quantified and associated with infrastructure impacts, commenting on the statistical robustness of the analyses used across the projects.

全球人口的持续增长导致了土地利用方式的改变与景观格局的重塑，交通、能源设施等人类基础设施的建设，对大型食肉动物（large carnivores）构成了尤为突出的威胁。环境影响评价（Environmental Impact Assessments）体系的建立，旨在识别新建拟建项目可能带来的环境影响，制定减缓影响的解决方案，并为决策制定与生态影响减缓提供科学依据。葡萄牙境内现存约300只狼，该物种已被列为濒危物种并受到全面法律保护。狼栖息于葡萄牙北部山区，而该区域也是过去20年间人类新建基础设施的集中区域。因此，当地已建立数十个狼类监测项目，用于评估狼种群现状、识别人类活动的影响，并为制定合理的减缓或补偿措施提供参考。本研究对葡萄牙境内的狼类监测项目开展综述，旨在回答四个核心问题：1. 狼类监测项目是否选取了足够的生物学参数以达成监测目标？2. 研究设计是否适用于影响评估？3. 数据采集方法与工作量是否足以支撑既定的推断目标？4. 数据的统计分析能否得出稳健可靠的结论？

总体而言，本研究发现狼类监测的既定目标与报告结果之间存在脱节，且二者往往均未契合现行的国家狼类监测指南。尽管现有监测投入了大量工作量并采用了多样的方法，但数据分析几乎仅使用相对指数（relative indices）或描述性统计量（summary statistics），极少考虑由（不完全）观测过程所引发的潜在偏差。这使得跨空间与跨时间的影响对比难以开展，因此无法为深入理解狼类对基础设施相关干扰的响应提供通用的认知基础。本研究建议制定标准化监测方案（standardized monitoring protocols），并倡导采用能够校正不完全检测（imperfect detection）偏差的统计方法，以保障监测项目的准确性、可重复性与实施效能。

方法
本研究综述了2002年以来葡萄牙境内所有针对环境影响评价开展的主要狼类监测项目（补充材料表S1）。鉴于本研究聚焦于针对性狼类监测在评估基础设施建设影响方面的适用性，本综述仅纳入专门针对狼类设计的监测项目，而非针对通用哺乳类的监测项目。

本研究的初始数据集源自2019-2021年全国狼类普查（National Wolf Census，Pimenta等，2023），该普查已识别出2014至2019年间葡萄牙境内开展的所有狼类监测项目。此外，本研究通过个人调研、联系主要科研团队、政府机构及环境影响评价顾问，补充了2014年前或2019年后启动的监测项目，最终形成完整的项目列表。

根据监测周期的不同，狼类监测项目可产出多份（通常为年度）报告，此类报告未经同行评议（peer-reviewed），且未在Web of Science、谷歌学术（Google Scholar）等标准搜索引擎中收录，但可从葡萄牙环境署（APA，www.apambiente.pt）公开获取。本研究通过葡萄牙环境署的搜索引擎（https://siaia.apambiente.pt/）开展线上检索，共识别出30个监测项目。针对每个项目，本研究均获取其首份与末份报告，以识别方法学上的变更；若未获取到末份报告，则调取最新版报告；若未检索到任何报告，则向负责该项目的团队申请获取。

本研究的分析围绕狼类监测项目的四个相互关联的核心维度展开，且这些维度的设定应契合项目初始目标：（1）生物学参数：即用于评估影响的狼类相关参数；（2）研究设计：即用于采集与分析数据的抽样方案（sampling scheme）；（3）数据采集：即数据采集所采用的方法与投入的采样工作量（sampling effort）；（4）数据分析：即用于估算相关参数并评估影响的数据分析方法。

研究将识别出的生物学参数分为两类：分布与种群动态。二者分别对应评估狼类栖息地排斥效应、繁殖模式改变等影响所需的核心输入数据。分布相关参数用于衡量狼类的存在与否，而种群动态参数则用于估算种群层面的效应，如种群数量、种群密度、存活率或繁殖率。此外，本研究还记录了项目是否按照指南要求，开展了猎物种群分布与数量的量化工作。

针对研究设计，本研究审查了项目的抽样方案，重点关注研究的时空维度，包括调查总面积、区域内样地的划定标准（即采样分辨率（resolution））、研究周期及采样季数。本环节的目标是判断所采用的抽样方案是否适用于评估基础设施对狼类分布或种群动态的影响，具体取决于项目的研究重点。

针对数据采集环节，本研究识别了项目采用的主要数据采集方法及对应的采样工作量。目前最常用的方法为痕迹调查，尤其是粪便调查（scat surveys）；针对此类研究，本研究记录了是否尝试通过粪便DNA（faecal DNA）开展物种或个体的遗传鉴定。本研究对比了不同推断目标下采样工作量的差异，并如前文所述，评估了哪些（若存在）项目或数据采集方法最有可能获取影响评估所需的证据。

本研究将数据分析环节分为两类：单年度分析与多年度分析。针对单年度分析，本研究梳理了监测项目如何利用数据对目标生物学参数的状态开展推断，并讨论了相应方法的优缺点。针对多年度分析，本研究记录了项目如何量化差异或趋势并将其与基础设施影响关联，同时评价了各项目所采用分析方法的统计稳健性（statistical robustness）。