1. Introduction
Forest management objectives have undoubtedly evolved during the last decades and new criteria associated with goods and services with no market price (e.g., biodiversity conservation) have been incorporated. However, in some classic studies on forest management, biodiversity conservation was not included as a forest planning objective. Thus, it is not surprising to find that only one objective associated with biodiversity aspects, i.e., hunting (game preserve) [
1], whereas this objective is not even explicitly mentioned in others [
2]. However, as shown in some reference books, this new good is currently considered in the analysis [
3,
4,
5].
The gradual incorporation of these new objectives (biodiversity, sustainability, watershed protection, etc.) has increased the complexity of forest management since, according to some authors, the spatial component should be included in the forest plans [
6]. In synthesis, a situation in which the time component was the leading decision of forest management from a strategic point of view (when to cut) has changed to another in which it is not so important when or how much should be cut, but where and how the final cuts are done.
In this paper, we have used the classic concepts of strategic, tactical, and operational planning [
7,
8]. In short, leaving aside the links and conflicts between these levels [
9,
10], and the differences between countries, strategic planning involves long-term goals, is highly aggregated [
11], and often includes other forecasts of different economic, ecological, and social outcomes [
12]. Tactical planning focuses on medium-term or medium-scale goals [
7] and on the areas or trees to be harvested [
13]. Operational planning states when and how the operations are performed [
14]. The use of Operations Research (OR) techniques in these planning levels has frequently been described for several decades [
12,
15,
16].
Thus, although the inclusion of an objective related to biodiversity conservation may justify an increase in the length of the rotation [
17], once biodiversity conservation aims are included, tactical decisions gain more relevance. However, this change involving the incorporation of spatial restrictions has triggered a search for more sophisticated analysis tools than those initially used in forest management (linear programming), as we will show in the next sections.
Furthermore, the importance of biodiversity conservation has led to certain critical decisions in forest management, which has been modified in some aspects. For instance, similar to the changes in the rotation length [
18], some silvicultural models have been proposed that avoid clearcuttings. These models suggest leaving some trees in order to mitigate the impact of cuttings on biodiversity [
19,
20] or even the use of systems like Continuous Cover Forestry (CCF), which favors the creation of mixed stands and is widely accepted for biodiversity management [
21].
In keeping with the above comments, it should be added that numerous techniques based on OR have been applied in an attempt to develop acceptable forest plans from a biodiversity perspective. These techniques have been used at different spatial scales, under diverse silvicultural treatments, and with multiple objectives. However, to our knowledge, no previous review has been made to date, although it is true that there has been a proliferation of different surveys related to the integration of biodiversity and some aspects of forest management in the past few years, especially in recognizing the interactions between the spatial relationships and scales. One of these approaches was focused on the application of Geographical Information Systems (GIS) in forest management when biodiversity is incorporated as an objective [
22]. Furthermore, there is a review of studies using optimization tools to integrate spatial variables into forest management [
23]. Unlike this current study, the bibliographic references are grouped into four categories according to the type of problem, but without taking into account the optimization technique used. Another study which should be mentioned is a review where the publications using OR techniques in North America on forest management issues up to 2001 are given [
24]. On the other hand, some aspects related to forest planning at a spatial scale appear, with an emphasis on various objectives, among which biodiversity is found [
13,
25]. Also, different techniques (planning tools) to solve biodiversity conservation problems can be founded in the scientific literature [
26]. A more recent work analyzes the application of different OR techniques to aspects related to the biodiversity in five particular cases, one of them being forest exploitation [
27]. Finally, some reviews on the use of Multiple Criteria Decision Making (MCDM) techniques in forest management include references related to their application in problems in which biodiversity is an objective [
28,
29].
The main aim of this study is to critically compile and assess papers which tackle issues related to the integration of biodiversity into forest management when using techniques based on OR, as well as analyzing the technique employed in each case. We have differentiated these papers in two classes: case studies and methodological papers with illustrative examples. As secondary objectives, we examined aspects related to planning and the spatial scale used, or attributes associated with silviculture, like the stand structure or the biodiversity indices employed. Also, for some of these attributes, it was aimed to analyze their evolution over time throughout the set of articles selected.
It is not an objective of this study to justify which OR method would be the best one for a particular forest management topic including objectives related to biodiversity conservation. Selecting the appropriate analysis tool for a particular issue is a difficult decision, as many authors have confirmed, both when talking about multi-criteria decision theories [
30] and metaheuristic techniques [
31]. To sum up, as it was stated: “There is no method that is universally best or even applicable for all situations.” [
5]
2. Material
To compile the papers included in this review, first, it should be noted that no articles present in non-Journal Citation Reports (JCR) journals and books or book chapters have been considered in this survey. These papers have been located mainly in a series of searches made in ISI Web of Science and SCOPUS databases including the following fields: biodiversity + forest management + optimization; biodiversity + conservation + optimization + forest; biodiversity + conservation + multi-criteria + forest; biodiversity + forest + multicriteria; AHP (Analytic Hierarchical Process) + conservation + biodiversity + forest heuristics + conservation + biodiversity + forest; simulated annealing + conservation + biodiversity + forest; multiobjective + conservation + biodiversity + forest; linear programming + conservation + biodiversity + forest; integer programming + conservation + biodiversity + forest.
In a second phase, a special search was made by only including the names of the principal OR techniques, which are described in a later section, together with the expression “forest management”, and hence selecting possible studies not included previously. Discarding some surveys, an initial collection of 512 papers in all was taken, 179 of which were finally considered (up to 15 of which were from September 2015).
Once the papers were selected, the following step was to define a set of fields covered by all the articles analyzed. Firstly, the OR techniques employed, the objective functions, and the main constraints considered were analyzed. Similarly, for each case, the temporal (strategic or long-term, tactical or medium-term) and/or spatial (ecosystem, forest, or stand) planning level used was computed. The typology of the objective function (net present value, biodiversity, timber volume, and others) and some constraints (related to timber production or biodiversity) were studied. In addition, the number of different animal species present in the analysis, as well as the biodiversity indices included in these articles, has been recorded. According to the silviculture employed, diverse aspects related to the typology of the cuttings (final or intermediate ones), the use of systems like Green Tree Retention (GTR,) or the reserve selection destined for protection has been considered. The stand age structure (even or uneven) in each case study has also been analyzed. As a complement, in accordance with this growing concern about biodiversity conservation, the presence of protection figures in the various case studies has been recorded. Finally, we have analyzed whether the papers included in our database were mainly case studies, directed towards solving real forest planning problems, or papers with a methodological orientation, but which included a case study as an illustrative example.
Biodiversity as a concept can be very broad in scope and difficult to pin down [
32]. Thus, it was necessary to define what attributes were presented in the articles considered in this study. In principle, the review was restricted to papers considering conservation objectives mainly related to wildlife and natural vegetation, excluding those articles related to hunting or fishing activities. However, we have not considered studies related to the following issues: biodiversity related to the water courses, wildfires, prescribed burning, pesticide or fertilizer treatments, and pest management.
4. Results
Table 1 shows, quantitatively, the number of papers which have used the different techniques previously introduced. These techniques have been divided into four large groups: classic optimization, heuristics, MCDM, and other techniques. Given that in one paper various methodologies may have been used (this occurs in 36.9% of the cases), the sum of all the papers employing each methodology obviously exceeds the number of papers considered (179). The whole list of papers appears in
Supplementary Materials.
The temporal evolution of these publications is shown in
Figure 1. It can be seen how the prolific years correspond to the decade 2000–2010 (123 papers), in which both heuristics and multi-criteria techniques have been very often applied. The applications using classic optimization techniques keep up a relatively stable trend throughout the time, with the most abundant ones occurring before and after the aforementioned decade.
Table 2 shows the characteristics of those papers when addressing the composition of their objective function and their constraints, indicating which are the most common constraints in forest management problems included in our database. The average number of objectives included in the case study is 2.3%, and 57.5% of the papers include constraints associated with timber production, whereas 84% include those related to biodiversity conservation. On average, the case studies included 2 to 3 (2.3) different objectives with 57.5% and 84% of the papers including constraints associated with timber production and biodiversity conservation, respectively. Among the constraints associated with biodiversity conservation, the most frequent ones are those related to the proportion percentage of mature patches that ensure a minimum viable population of focal species, and adjacency constraints. These constraints prevent harvest in adjacent areas for one or several periods, preventing large harvesting areas [
25]. Finally, we have included in this table the main OR techniques used when considering conservation as the objective function in the papers analyzed. In analyzing the papers where this objective function appeared, MCDM methodologies were the most numerous OR group.
Table 3 displays some aspects related to the forest management applied in these papers. It can be seen how a strategic type of planning has been more commonly carried out than a tactical one, although in a reduced number of studies both scales have been approached—in most cases due to the addition of spatial considerations.
As shown in
Figure 2, concerning the temporal evolution in approaching this type of forest management problem, a strategic temporal scale rather than a tactical one has been most frequently employed.
On the one hand, at a spatial level, ecosystem analyses are in a majority compared to other more disaggregated forms. With regard to the silviculture practiced in these cases, a notable number of papers apply silvicultural practices associated with the presence of areas with no final cuttings, although clearcutting is performed in almost one third of the studies. On the other hand, a limited number of papers indicate the structure of the forest stand in each case study, and only a few correspond to uneven-aged structures.
Table 4 shows the use of different groups of OR techniques in specific case studies, encompassing a total of 54.7% of the papers reviewed, and 45.3% of the papers describing methodological approaches. Furthermore, the percentage of papers using hybrid methods has been included. This percentage is defined as the use of two or more OR techniques included in the different groups shown in
Table 1.
Going on to analyze how biodiversity has been introduced into these types of papers,
Table 5 shows how more than half of them have tackled problems related to wildlife. In some cases, several indices have been used to define it, although this has not been a generalized trend in all the studies.
5. Discussion
In this study, we have shown that the integration of biodiversity conservation objectives into forest management with the help of tools proceeding from OR has constituted a fruitful work area during the past few years. However, although we have previously said that the justification of choosing an OR method to address biodiversity was not an objective of this work, before discussing the results, it is necessary to emphasize the fact that none of the studies analyzed explain the benefits and drawbacks of all OR techniques shown here. The reasons provided for using a given OR method are always inadequate and partial, depending in some cases on the quality of the information available [
5], not allowing one to extract reliable conclusions regarding the utilization of these methodologies.
Following the division into large blocks made in this study, which does not coincide with that reported in other studies [
25], it was observed how papers employing multi-criteria decision and MH techniques predominated over classic optimization ones. However, following
Table 1, the most common methodology is a metaheuristic (SA) one. These results are quite different from another one, which analyzed optimization in forestry using 85 papers published since 2010 [
12]. Their results show how heuristics are the techniques most used at the stand level, with some optimization techniques (MIP and LP) also used at the forest and landscape level. As for the temporal evolution of these papers, it was seen that there has been an apparently increasing trend in the whole period considered up to 2008, followed by an abrupt drop from which it has recovered in recent years. That trend up to 2008 coincides with what has been noted in other similar reviews [
29].
When analyzing our results, it is seen that no OR technique was used more than others for solving these type of problems. As observed in
Table 1, none of the techniques were applied in more than 20% of the studies analyzed. This dispersion in the use of different methodologies may indicate that this is an open problem in which various approaches are entirely compatible. This conclusion was observed in a recent study [
9], in which a set of open forest management problems where OR techniques had been applied were included, and some of them were precisely directly related to biodiversity conservation.
Regarding the temporal scale of forest planning, and following the results shown in
Table 3 and
Figure 2, strategic planning clearly predominated, although the tactical and strategic levels were merged in some papers. It is hard to obtain conclusions from these results for several reasons: how biodiversity is considered in each case study, how strategic and tactical planning has been defined, the different spatial component of each problem, and the integration of strategic and tactical planning, that is not necessarily univocal [
7,
43]. It should be pointed out that operational planning has not been included as it was only approached directly or indirectly in 5 of the 179 studies. Similarly, at a spatial level, the ecosystem component apparently prevails, and the most frequent silvicultural practice is the one that includes aspects close to GTR. Finally, the forest structure predominating in the papers analyzed was even-aged systems. This circumstance is surprising because, traditionally, it has been considered that more complex structures (multiaged stands) presented higher biodiversity levels [
44]. However, it is worth noting that less than 40% of the studies analyzed have detailed the structure type of the stand ages in their respective case studies. Actually, these more complex structures are more difficult to model, and they have therefore been investigated to a lesser extent [
17].
The consideration of objectives not presenting a market price, and which need a greater knowledge applied of disciplines like ecology, offers, from an economic perspective, a more significative challenge. However, nowadays, forest management has assumed the idea that it is necessary to integrate timber production and biodiversity conservation into the areas where cuttings are produced [
45]. All this has led to a duality within forest management between production and conservation objectives, with different tools being applied to integrate both objectives, and with different relationships (positive or negative) between each other [
46]. Thus, although some authors suggest that the cost of applying certain measures in order to favor biodiversity can be estimated [
3], other papers introduce economic concepts like the Production Possibility Frontier (PPF) to tackle problems in forest management which include, among others, those associated with timber production and biodiversity conservation [
47,
48]. It is widely known that the combination of a production objective and a conservation purpose that are located in this curve are, at least, efficient (i.e., that conservation objectives are fulfilled at a lower cost) [
49]. However, and independently from the objectives of each study analyzed, what does come out of the previous results (
Table 2) is that the production-conservation duality has been taken into account in numerous studies mentioned here. Indeed, there were a notable number of papers in which objectives measured private profitability from the production of different forest outputs, or that included elements of an economic nature in the constraints of the models used.
One aspect to be considered is the gradual incorporation of spatial results which accompanied the solutions provided by the different operations research methods collected in this study. Approximately half of the studies analyzed already explicitly incorporated GIS data and/or used software fully integrated into GIS. Obviously, GIS produces displays of geographic information for analysis, but does not ensure a spatial approach to biodiversity. However, in our study, all the papers dealing with connectivity and fragmentation, and forest mature patches have used a GIS. The integration of spatial aspects with several OR methods using GIS could be feasible [
25], and it has been already mentioned as being a beneficial research line for years [
11]. Finally, some authors suggest an interest in integration in the future, mainly referring to the merging of MCDM techniques with GIS [
39].
Since the biodiversity concept is very broad, and throughout this paper it has been included under different objective functions and constraints, it might be useful to see how these aspects have been considered in various geographical regions and countries. Thus, we have selected some results corresponding to the seven most represented countries in our database: United States, Finland, Sweden, Canada, Portugal, Australia, and Germany (see
Appendix B). These countries constitute more than 78% of the papers. This Table shows some differences between the countries relative to the use of strategic and tactical planning, the use of mono-criteria or MCDM techniques, the objectives and constraints, and also the silvicultural practices specified in the models.
It is interesting to note that some techniques (SA, AHP, MAUT) are clearly associated with a few of the countries, while in others their application is much lower. Also, comparing these figures with others obtained at the forest and stand level, their results are quite different in the geographical distribution and the OR technique used in each of the 85 papers analyzed [
12]. However, it should be noted (
Table 2 and
Table 5, and
Appendix B) that some key elements in the management of biodiversity issues using spatial requirements like connectivity or adjacency are not any of the topics most addressed with OR techniques in this context. This can also be said about the use of biodiversity indicators in these optimization models, even when they provide interesting results [
50,
51]. These issues could be lines of further research.
If we raise the issue as to what direction biodiversity modeling could go within forest management using these OR tools, one clear trend is that of employing more sophisticated models which hybridize diverse techniques. For some authors, this increase in model complexity is due to the introduction of biology principles in biodiversity conservation problems [
26]. This circumstance has been verified in the papers analyzed here since around 35% of the articles already include more than one different OR technique (
Table 4). Thus, the hybridization of methods could improve decision-making in comparison with the use of a single technique [
39].
In this sense, some aspects are gradually being introduced into forest management incorporating this set of techniques. One of them is the inclusion of stakeholders’ preferences in relation to problems associated with biodiversity conservation through group decision methods [
52]. In fact, most of the few studies that have used these techniques when incorporating biodiversity conservation into forest management have been published in the last few years. An example of another field in vogue is in the solving of these problems under non-deterministic environments, as mentioned in some papers [
53] and having shown an increased tendency to appear in the last decades [
54]. Other methodologies that, a priori, could be addressed to integrate biodiversity would be the Decision Support Systems (DSS). However, in the papers analyzed this methodology is still of a minor importance (it only appears in 7 of the 179 papers). This fact had already been pointed out in reviews dealing with the implementation of DSS in forest management [
55].
Although, as discussed above, there is not a clearly predominant OR technique for addressing biodiversity conservation problems in forest management, when partitioning our database some clear trends appear (besides the importance of the hybrid methods). These methods have been developed in spatial forest planning issues [
25]. On the other hand, what should be highlighted is the wider use of MCDM techniques when the problem includes an objective function related to biodiversity conservation (41.9% of the papers, see
Table 2), especially in countries like Finland (
Appendix B). This latter trend appears again in
Table 4, in which the use of MCDM techniques in studies with the aim of solving a given real-life planning problem reaches 46.9%. This finding suggests that, as some authors indicate, many biodiversity management problems have more than one dimension [
56], so MCDM techniques fit very well. Furthermore,
Table 4 clearly shows how the use of non-deterministic techniques is practically concentrated in those papers associated with case studies.
6. Conclusions
The analysis of the integration of biodiversity into forest management using OR tools has demonstrated which techniques are preferred for solving a specific problem. This has been characterized by the techniques and groups of techniques employed, the components (objective functions and constraints) of those OR models, of the different scales (temporal, spatial), and of the silvicultural and management aspects. Likewise, the biodiversity issues most frequently dealt with in these types of problems have been detailed.
Regarding the methodologies used to integrate biodiversity objectives into forest management plans using OR techniques, our results show how MCDM techniques and MH predominated over classic optimization ones. Additionally, the most frequent MCDM methodologies are the techniques applied to discrete problems (AHP, MAUT, etc.), while SA is the metaheuristic one most often used. However, it has been verified that the use of certain techniques has been concentrated in some countries in which scientific production on this topic has been more prolific, probably due to the fact that the selection of a certain technique is directly related to the abilities and experiences of the authors. Finally, from the point of view of forest management, these optimization models have been designed for different silvicultural scenarios, although GTR and clearcutting predominate, especially in even-aged stands.
As for the cases analyzed, the aspects related to biodiversity using OR techniques have been approached mainly at more aggregated spatial scales (ecosystem level) in different problems associated with wildlife, and in forest systems which do not have to be subjected to protection figures. However, it should be realized that this problem type is one of the cases analyzed, and that, unlike other objectives, biodiversity consideration as a criterion in forest management can be defined or introduced into the model in different ways.
The same as for other ecosystem services which do not correspond to productive aspects, there is a great diversity in the optimization techniques employed for the integration of aspects related to biodiversity in forest management, and no universal and valid combination for all the cases analyzed has been found. However, forest managers, when confronting these problems, and once the temporal and spatial scales of the model to be developed have been fixed, should first ask themselves which OR technique fits their approach best. For this purpose, they should define its nature (i.e., mono- or multicriteria), as well as whether they need to integrate the opinions of the different stakeholders and use a DSS to carry this out. By answering these questions the next step is obtained towards designing a model which permits them to solve the problem being tackled with only one item lacking: the consideration of a deterministic or stochastic environment. With the latter, there will be occasions when conceptually simpler models can be justified, and on others in which it will be necessary to hybridize different methodologies in order to solve certain types of problems in an adequate manner. Throughout this work numerous examples have been cited which can serve as an aid in this respect.