Introduction

Sustainability and multi-functionality principles have become increasingly emphasized in forest management and forest policies around the globe (Hujala et al. 2009; McDonald and Lane 2004; Nichiforel 2010; Nichiforel et al. 2018). The implementation of these principles results in more complex management with higher requirements of precision, knowledge and documentation of the forest system for the manager as well as the policy-makers (Segura et al. 2014). Forest management plans (FMP) are pivotal in the knowledge and information system that provides managers with the information to support strategic, tactical and/or operational forest management planning. Due to its perceived importance for forest management, FMP is a key forest policy instrumentFootnote 1 towards small-scale private forestry in several countries (Brukas and Sallnäs 2012; Hokajärvi et al. 2009; Serbruyns and Luyssaert 2006). In addition, having FMP is also a certification criterion in many countries (PEFC 2015). However, FMP differ greatly in purpose, scope and level of details (Bettinger et al. 2017; Solli 2013). In this paper, we focus on the FMP as decision-support tools for non-industrial private forest owners (NIPF) typically consisting of inventory results, maps and management proposals (Eid 2006; Nuutinen 2006; Størdal et al. 2006).

As FMP are closely linked to forest policy and national certification requirements, the development, users and content of FMP are country-specific (Ficko and Boncina 2015). For instance, Nuutinen (2006) and Brukas and Sallnäs (2012) considered FMP mainly as a policy instrument for proper forest management among Finnish and Swedish private forest owners, respectively. Hujala et al. (2009) in Finland regarded FMP as an important instrument for implementing forest policies; however, Eyvindson et al. (2010) stated that FMP compiled in Finland by professional planners are without proper consultation with forest owners. Bouriaud et al. (2013) stated that FMP is a compulsory forest planning tool in Eastern and Central European countries.

In many European countries, enhancing forest management activity among NIPF owners is a policy goal (Forest Europe 2015). This policy is further emphasized to achieve objectives of climate change mitigation, and rural development (Weiss et al. 2019; UNECE/FAO 2020). In the US, Joshi and Arano (2009) observed that forest owners are more likely to engage in forest management if they have FMP. Findings from Slovenia and Spain suggest that forest owners consider FMP as important decision-aid tools (Bruña-García and Marey-Pérez 2017; Ficko and Boncina 2015). However, studies from France and the US show that only 3–6% of forest owners have FMP (Agreste 2013; Butler and Leatherberry 2004).

The success of FMP as a policy instrument depends not only on the forest owners acquiring the plan, but also that they know its content and are implementing its proposals (Ficko and Boncina 2015). However, to the best of our knowledge, there is a lack of empirical studies addressing the influence of owner characteristics, objectives and motivations of managing forests, socio-economic situation and property characteristics on the acquisition of the FMP, awareness of its content and finally its implementation. A better understanding of these relationships ought to be of interest for improved policy design and better tailoring of FMP and extension services. A deeper insights of the success of the current FMP among forest owner subgroups is valuable. This insight can aid in developing FMP products and services that can reach larger audiences and aid mitigating the longstanding problem of low participation of NIPF owners in forest management and planning. We fill part of this void by carrying out quantitative analyses to answer three research questions that together form the FMP pathway:

  1. 1.

    Which socio-demographic, financial, informational and forest-management factors characterize forest owners who have FMP?

  2. 2.

    Which socio-demographic, financial, informational and forest-management factors characterize forest owners who know the content of their FMP?

  3. 3.

    Which socio-demographic, financial, informational and forest-management factors characterize forest owners who implement the proposals of their FMP?

We base our analyses on a large survey dataset representing the population of Norwegian NIPF owners. Even if we are undertaking the study in Norway where about 80% of the productive forest land is owned by NIPF owners, its relevance extends to other countries with NIPF owners.

Due to the lack of literature of forest owners’ knowledge and implementation of FMP, we chose an open, explorative data mining approach to find subgroups using decision-tree (DT) analyses. Based on the DT outcomes, we built regression models to compare the DT outcomes and obtain the importance of individual predictors for gaining more insight into the role of FMP in NIPF owners’ decision-making.

We continue by presenting the study area, the organization of FMP projects in Norway and methods.

Methodology

Study area and survey design

Norway has approximately 8.6 million hectares of productive forest land, representing 26% of the total land area (Fig. 1). The predominant and most economically important tree species are Norway Spruce (Picea abies) and Scots Pine (Pinus sylvestris) which comprise about 75% of the total forest area (Statistics Norway 2021a). NIPF ownerships extend to approximately 80% of the productive forestland, divided into more than 118 000 properties (Rognstad et al. 2016; Statistics Norway 2021b). The total net annual growth is about 24 million cubic meters (2015–2019) while removals averaged 11.1 million cubic meters in the years 1996–2019, and NIPF lands constitute approximately 70% of the net annual growth and 86% of the removals (Statistics Norway 2021a).

Fig. 1
figure 1

Map of counties with forest resources in Norway (kilden.nibio.no)

Full size image

This study analyzed data collected in 2014, by a nationwide mail survey sent to 3150 randomly selected Norwegian NIPF owners having more than 2.49 hectares of productive forest land. The survey was developed and administered collaboratively by Statistic Norway and the Norwegian University of Life Sciences. To increase the survey efficiency the methodology followed the Tailored Design Method (Dillman 1978). The questionnaire survey included three mailings: first the questionnaire was sent to all respondents. In the second and third mailing, a reminder card enclosed with the questionnaire was sent to respondents one and two months after the first mailing. A total of 1637 respondents returned the survey, giving a response rate of 52%. We gathered information from forest landowners across all counties in Norway, excluding Finnmark, due to its insignificant private ownership. The survey included questions of the use and management of forest, attitudes, reasons for owning forest and FMP. Tax record information about forest owner’s taxable income and wealth as well as productive forest area was appended to the survey data by Statistics Norway.

Forest management plans in Norway

In Norway, FMP are not mandatory by law, but it is an important forest policy instrument regulated by law (Norwegian Ministry of Agriculture and Food 2004). The main objectives aimed at by the FMP policy are increased harvest, improved silviculture, better control of the management and documentation of key habitats for ensuring sustainable forestry (Norwegian Agricultural Authority 2013). FMP is typically renewed every 15–20 years, and due to the property structure with many small forest ownerships, the vast majority of new FMP are organized as projects for larger areas, i.e. one or more municipalities. The initiative to make new FMP is taken by the county-level public management in collaboration with local public management and forest owner organizations (Norwegian Agricultural Authority 2010, 2020). Once it is decided that an area will have new FMP, a steering board for the process consisting of representatives from public management, forest owners and timber buyers is appointed by the local public management. While the authorities set regulations about the basic prerequisites of the content of the plan, the parties may choose the inventory method (Norwegian Ministry of Agriculture and Food 2004). The authorities provide direct subsidies and tax deductions to owners who buy the plans given that the technical requirements are met and that owners provide the authorities with the plan. Environmental mapping as demanded by the Program for the Endorsement of Forest Certification (PEFC) certification is carried out as an integrated part of the FMP inventory. Through the group PEFC certification, timber buyers as the certificate holders ensure that the mapping of the environmental values is carried out in accordance with the certification requirements (PEFC 2015).

Participation in these FMP projects is voluntary, but subsidies are used to enhance participation. Also, for supplying timber, the current certification requires that if necessary, remapping of key habitats should be undertaken at a maximum of fifteen years (PEFC 2018). This remapping will often coincide with the cycles of the local FMP projects and acts as an incentive for forest owners to participate. Depending on the design of the local FMP projects, owners may buy only environmental mapping with no FMP. Still, the FMP participation rates vary from about half of the area in the coastal region with limited commercial forestry traditions to more than 90% of the productive forest area in the traditional forestry areas in the Eastern part of Norway (Korsvold 2020).

The final FMP products to the owners typically consist of a forest map with delineated stands, timber inventories of the stands and a description of treatments (Eid 2006). The description of treatments is usually a standardized projection of harvest and silviculture investments, based on an underlying growth and yield simulator that provide harvest prognoses. The environmental mapping part consists of a map with set-aside hotspot (key habitat) areas and a description of the required treatment for maintaining their biological values.

Quantitative modelling framework

The three research questions representing the adoption of FMP, consisted of a two-stage procedure: In the first step, the respondents’ answers on the question of having a FMP were analyzed (AcquireFMP). In the second stage, where the sample consisted of only forest owners indicated having FMP in the first stage, two outcome components were analyzed: the forest owners’ self-reported knowledge of the content of their FMP (ContentFMP) and the owners’ self-reported implementation of the FMP’s proposals (ImplementFMP). Thus, the two stages provide increasingly more details into the actual success of FMP as the forest owners’ decision-support tool.

To answer the three research questions, we applied the model of adoption framework (Rogers 1995; Wejnert 2002) adjusted to our needs. In our study, the adoption model stipulates that socio-demographic, financial, information and forest management factors are important predictors for determining the rate of adoption of FMP, its awareness and implementation. We examined the factors that influence a forest owner’s decision to adopt FMP. We consider adoption is a complex process pertaining to forest owner decision to acquire and implement FMP. Understanding characteristics of forest owners that adopt FMP may assist in identifying appropriate extension services and FMP products to forest owners in order to improve adoption rates which in turn may improve forest management.

The methodology in this study includes, firstly utilizing the DT for assessment of a broad set of variables related to the property, attitudes, cultural factors, future plans for the property, demographic and socio-economic factors to understand which groups of owners acquire FMP, know the content and implement the proposals. Secondly, the significant variables from DT were brought forward to build the regression models.

The dependent variable in the first stage is discrete; a value of 1 was assigned if the respondent possessed a FMP and 0 otherwise. The two dependent binary variables for the second stage, the awareness of the content of FMP and its implementation, were both responded on a four-point Likert scale turned into dichotomous variables (Table 1).

Table 1 Definitions of dependent and independent variables used in the decision and regression models: Acquisition of forest management Plan (AcquireFMP), knowing the content of the forest management plan well (ContentFMP) and implement the proposals from the forest management plan (ImplementFMP)
Full size table

Principal component analysis (PCA)

The respondents stated the importance of totally twelve financial, cultural, recreational, and nature-preservation ownership objectives on four-point Likert scales. Several of the objectives were highly correlated. We therefore employed PCA dimensionality reduction procedure to merge them into a set of principal components that consist of interpretable and uncorrelated combined variables to avoid multicollinearity in further analyses (Kuuluvainen et al. 1996). The three components with eigenvalues greater than one were selected (the so-called Kaiser’s rule) with a lower cut-off of PC loadings of 0.30 (Favada et al. 2009; Kuuluvainen et al. 1996). The new composed ownership objective variables were included in the DT and regression analyses.

DT and regression analysis

DT analysis is a well-established non-parametric supervised learning and data mining technique used for classification in complex and big datasets (Durán-Román et al. 2021; Han et al. 2012). DT’s appeal in comparison to general regression analysis lies in its straightforward interpretation of associations between the response and a set of predictive variables and between predictive variables (Durán-Román et al. 2021; Hothorn et al. 2006; Loh 2014).

In this study, we utilized conditional inference classification DTs to explore multivariate relationships between the outcome variables and a set of candidate predictor variables, using the ctree function of the “party” package in R (Hothorn et al. 2006). The conditional inference DTs build models using algorithms that recursively partition the data into a number of binary splits called nodes (Loh 2014). The algorithm determines the variables to be split at each node and the nodes are connected to each other by branches. The null hypothesis that there is no relationship between the predictor and response variables is tested. The partitioning maximizes the homogeneity within the branches and the process will come to an end when the null hypothesis cannot be rejected (Hothorn et al. 2006). The stop criteria are based on adjusted p-values following the Bonferroni test type (Bland and Altman 1995). We chose 0.1 as the maximum p-value. This procedure made certain that the appropriate-sized tree is grown. This also implies that pruning or cross-validation to avoid overfitting is not required (Pinet et al. 2015; Hothorn et al. 2006).

We thereafter built logistic regression models with variables that were significant in the respective classification DTs. The two variables Distance and Productive area were log-transformed in the regressions to satisfy the normal distribution assumption, and each Probit model was tested for multicollinearity using the Variance Inflation Factor (vif) test in R. Sample weights were added to the observations for the statistical analyses, but not for the DT that do not provide population estimates, but merely classify the respondents.

Results

PCA dimensional reduction

PCA was conducted on 1637 complete records of the 12 ownership objective statements described in Table 2, where respondents rated statements of owning forest land on a four-point scale from “not at all important” to “very important”. The highest means were observed in the nature, leisure and intrinsic value statements while the lowest means were seen in the financial objectives.

Table 2 Descriptive statistics of the ownership objectives. The results are used in the PCA. The quantification was done using ordinal 4-point scale: Not important at all (1); slightly important (2); of relatively great importance (3); very important (4)
Full size table

The Principle Component (PC) loadings for the first three PCs with eigenvalues > 1 are displayed in Table 3. The three PC accounted for 65% of the variance. In PC1, the loadings of all twelve statements were in the range of 0.4 to 0.8, suggesting a strong multi-functionality as the primary objective of the NIPF owners. Component 1 was thus labeled multiobjective (MO). Component 2 consisted of the three statements describing economic objectives (Economic security, Income, and Investment) with loadings of 0.6–0.7, named ECON. The two nature diversity and preservation statements in component 3 were recorded with loadings of 0.3–0.4 and labelled Environmentalist (ENV).

Table 3 PCA summary statistics for 12 statements of reasons for owning forests
Full size table

Descriptive statistics

Table 4 presents summary statistics for the full sample of forest owners and the two subsamples of owners having and not having FMP, and significance levels of z-tests of differences between owners with and without FMP. Out of the total sample of 1637 respondents, 37% responded having a FMP. Among the owners with FMP, 66% stated having knowledge about the content in the FMP and 40% that they implement the proposals of the FMP.

Comparing the subsample of owners with a FMP with those without a FMP, we find that all variables except for taxable net wealth are significantly different. Owners with FMP are on average younger, have more education, live closer to the property, are more urban, have higher income, larger properties, more future plans for the property and acquire considerably more information from a set of sources. In addition, more of them are male and do farming and they have more pronounced ownership objectives. The mean age was 58.3 years for all forest owners, and owners with FMP were on average three years younger than others. The average property size was for all owners of 49.9 hectares, varying from a mean of 25 hectares for owners without FMP to 91.9 hectares among owners with FMP. On average, owners lived 56.2 km from the property with owners having FMP living on average 28.2 km away and owners not having FMP living on average 72.8 km from the property. 45% of the forest owners had harvested timber over the fifteen years, varying from 27% of owners without FMP to 74% of owners with FMP. Owners with FMP used various information sources considerably more than owners without FMP.

Table 4 Descriptive statistics (mean and standard deviation) of the variables (see Table 1 for definitions). Significance levels (10% = *; 5% = **; 1% = ***) refer to z-tests of differences between forest owners with FMP and forest owners without FMP. Weighted observations
Full size table

Acquisition of forest management plan (AcquireFMP)

The AcquireFMP conditioned classification tree produced a total of 15 terminal nodes in the final model (Fig. 2) classifying owners with and without FMP. The barplots at the terminal nodes represent the proportion of respondents having FMP in the subsample. The root node, hence the single most important predictor and given on the top of the tree, in AcquireFMP tree is harvest activity, which presents the maximum significant difference between NIPF owners with or without FMP. Other important variables in the model were contact with the municipality for advice on forestry, knowledge about public support schemes, property size, county of forest holding, forestry or agricultural education, economic reasons for owning forests, distance to property from residence and plans of transferring the land.

In the AcquireFMP tree the number of owners with FMP decreases when moving from the right to the left. In the seven nodes (18, 22, 25–29) on the right side of the tree, 50 to 100% of owners have FMP. The owners with FMP in these nodes harvest timber and do either receive advice from the municipality or have large properties. More than 80% of the owners who harvest timber, receive advice on forestry from the municipality and have forest in a traditional forestry region, have FMP. Out of those, among economically oriented owners who have plans of transferring the property and live close to the forest, 99% have FMP. In addition to previous harvests and advice from the municipality, the size of the holding is a decisive factor for having FMP. Out of the owners with property exceeding 26.14 hectares, 67% of owners have FMP (node 18), compared to 23% of the owners with smaller properties (node 17).

On the left side of the tree, low shares of forest owners have FMP. These are inactive owners without knowledge of public schemes. Among owners who do not harvest and do not have knowledge of public support schemes, the effect of forestry municipality advice is conditional upon the forest area as among the owners with properties smaller than 122 hectares who receive advice, the share having FMP is as low as among owners who do not receive advice from the municipality (nodes 4, 6, 7).

Fig. 2
figure 2

Conditional classification tree showing the variables’ influence on NIPF owners’ acquisition of Forest management plan (AcquireFMP). Each oval in the tree contains a particular variable. The n values at the leaves display the total number of observations that fall in the terminal nodes. The bars indicate the share of forest owners in each subgroup having forest management plan. For visibility, the counties were denoted with codes in the tree: Agder (A), Innlandet (I), Møre og Romsdal (M), Rogaland (R), Troms (TM), Trøndelag (TN), Vestfold og Telemark (VT)

Full size image

The estimates and marginal effects for each of the three probit models are provided in Table 5 using the significant variables from each DT. Knowledge of public support schemes, contact with the municipality and previous harvest activity increased the probability of having a FMP by 13%, 13%, and 15% respectively. This means that owners that have harvested at least once during the 15-year period with knowledge of public support schemes who receive advice from the municipality have about 41% higher probability of having a FMP compared to other forest owners. However, forest owners possessing property in county class 1 and 2 (defined in Table 5) have greatly lower chances of having FMP.

Table 5 Coefficients and marginal effects of the empirical probit model of acquisition of FMP. Only significant variables from the DT (Fig. 2) were brought into the regression model. The grouping of counties into three categories is based on DT outcome (Base: A, I, VK), (Countyclass1: VL, R, N), (Countyclass2: TM, TN, M, VT). The counties were denoted with codes Agder (A), Innlandet (I), Viken (VK), Møre og Romsdal (M), Rogaland (R), Troms (TM), Trøndelag (TN), Vestfold og Telemark (VT)
Full size table

Knowledge of content of forest management plan (ContentFMP)

In the ContentFMP classification tree (Fig. 3), the right side of the tree includes the largest share of NIPF owners that know the content of plan well, with 65–100% of owners in each subgroup. These owners think that the FMP is well adapted to their objectives, and either do not want to outsource the management (nodes 15–19) or are willing to outsource the management but receive advice from the municipality. Among owners who consider the plan as relevant, do not want to outsource the management and use media and journals as information sources, the geographical factor plays only a minor role. The subgroup on the left side of the tree (node 3, 6, 8) have lower shares of owners that know the content of the plan well. A part of these owners do not think that the FMP is well adapted to their objectives and do not have knowledge about public support schemes. Among the owners who do not think the plan is well adapted to their objectives, but do have information about public support schemes, owner age and county steer the extent to which forest owners know the content of the FMP, with older owners having less knowledge about the FMP.

Fig. 3
figure 3

Conditional classification tree showing the variables’ influence on NIPF owners having good knowledge of the content of the FMP (ContentFMP). Each oval in the tree contains a particular variable. The n values at the leaves display the total number of observations that fall in the terminal nodes. The bars indicate the share of forest owners in each subgroup having knowledge of the content of the FMP. For visibility, the counties were denoted with codes in the tree: Agder (A), Innlandet (I), Møre og Romsdal (M), Rogaland (R), Troms (TM), Trøndelag (TN), Vestfold og Telemark (VT)

Full size image

In the ContentFMP probit model (Table 6) all included variables were significant. If the FMP is adapted to the objectives of the owner, the probability of knowing the content of plan well increases by 26%. Likewise, a forest owner who has direct contact with the forestry section of the municipality and receives information about forestry from media and journals is more likely to have good knowledge about the content of the FMP. Owners with the intention to outsource the management of the forest have lower probability of knowing the content of plan. County is also an important predictor.

Table 6 Coefficients and marginal effects of the empirical probit model knowledge of the content of the FMP. Only significant variables from the DT (Fig. 3) were brought into the regression models. The grouping of counties into three categories is based on DT outcome (Base: M, R), (Countyclass1: I, VL, VK, N), (Countyclass2: A, TM, TN, VT). The counties were denoted with codes Agder (A), Innlandet (I), Viken (VK), Møre og Romsdal (M), Rogaland (R), Troms (TM), Trøndelag (TN), Vestfold og Telemark (VT)
Full size table

Implementation of forest management plan (ImplementFMP)

The ImplementFMP tree used fewer variables than the other two trees (Fig. 4). The variable stating that FMP is well adapted to the owner objectives takes the root node position in this tree similar to the ContentFMP tree, explaining the most of the difference between owners who implement the plan and not. On the right side of the tree, forest area steers decisively the extent to which owners that do not think the plan is well adapted to their objectives, implement the proposals. 94% of economically oriented owners with higher education and who think the FMP is well adapted to their objectives and with property in counties other than Agder or Vestland implement the FMP. On the left side of the tree, the subgroup of owners who think the FMP is well adapted to their objectives and have less of economic objectives, contact with the municipality makes a difference. Out of the owners who are not in contact with the municipality, 37% implement the FMP, but this number grows to 62% among the owners who are in contact with the municipality.

Fig. 4
figure 4

Conditional classification tree showing the variables’ influence on NIPF owners implementing the proposals in the FMP (ImplementFMP). Each oval in the tree contains a particular variable. The n values at the leaves display the total number of observations that fall in the terminal nodes. The bars indicate the share of forest owners in each subgroup who do implement the proposals in the FMP. For visibility, the counties were denoted with codes in the tree: Agder (A), Innlandet (I), Møre og Romsdal (M), Rogaland (R), Troms (TM), Trøndelag (TN), Vestfold og Telemark (VT)

Full size image

Finally, in the ImplementFMP model (Table 7), owners with a positive value of the most influential variable FMP adaptedforobjectives have 38% higher probability to implement the proposals of the FMP than others. In addition, owners who receive advice from the forestry section of the municipality are 13% more willing to implement the proposals of FMP on their property compared to other forest owners. County also makes a significant impact on the probability of implementing the FMP proposals.

Table 7 Coefficients and marginal effects of the empirical probit model of implementation of the proposals in the FMP. Only significant variables from the decision tree (Fig. 4) were brought into the regression models. The grouping of counties into three categories is based on DT outcome (Base: N, VL), (Countyclass1: A, VT), (Countyclass2: I, M, R, TM, TN, VK). The counties were denoted with codes Agder (A), Innlandet (I), Viken (VK), Møre og Romsdal (M), Rogaland (R), Troms (TM), Trøndelag (TN), Vestfold og Telemark (VT)
Full size table

Discussion

Our study sheds new light on the complexity of FMP as a decision-support tool for forest owners. The study indicates that less than half of the forest owners who buy the plan (40%) actually implement its proposals. This implies that high acquisition rates is no assurance of the success of the FMP as a policy instrument or decision-aid management tool.

The two variables that were significant across all trees and all regression models were direct contact with the municipality for advices of forestry and county group while the perceived relevance of the FMP (FMP adapted to owner objectives) was significant in the two trees and in the models where it was included. In addition, knowledge of public support schemes was important for the acquisition and knowledge of the FMP content, but not its implementation. Looking into the significant variables across the FMP adoption pathway in light of the theoretical adoption framework, we see that the information and forest management factors are dominating. Financial factors are the least important, with gross income and wealth not appearing in the DT; several socio-demographic variables like gender, distance between residency and property, centrality and farming are also not significant.

We found that previous harvest is a strong predictor and classifier for having FMP, but not for knowing the content of or implementing the FMP. Joshi and Arano (2009) suggest that owners with a FMP are more likely to harvest timber using self-reported activities. There are arguments that the causal effects go in both directions. Information about the timber resources may be a trigger for harvest, but owners may be informed about the resources and make management plans in other ways than by written FMP (Kittredge 2004; Kittredge et al. 2008). However, analyzing our data including a panel of harvested timber over a 15-year period (Bashir et al. 2020), we found no effect of the acquisition of FMP on subsequent harvest (data not shown). Purchasing FMP in Norway is a large investment for the forest owner that typically is covered by the owner’s mandatory forestry investment fund (Skogfond). The fund is credited with timber incomes and previous harvests provide the necessary means for buying the plan. Even if timber is cut and sold without a written FMP, owners may experience that having a FMP eases the operational planning considerably and may thus be motivated to acquire a FMP. More targeted information about the dependency between harvest, Skogfond and FMP towards forest owners, for instance through the municipalities’ forestry sections could help in disseminating FMP more broadly. Despite the association between previous harvests and having a FMP, owners who harvest may as well base the decisions on other information sources, suggesting the broader cognitive basis for management (Davis and Fly 2010). To summarize, more time-series research is warranted about the causal relationships between FMP and harvest, whose effects may be country-specific due to variations in law and certification requirements.

In line with several previous studies like Butler et al. (2007), Ficko (2019) and Majumdar et al. (2008), our findings suggest that forest area is a significant predictor for forest management planning. The lower FMP enrollment rate on smaller properties can be due to a complexity of reasons connected to forest types and area (Bruña-García and Marey-Pérez 2017; Hirschnitz-Garbers and Stoll-Kleemann 2010). Smaller properties may face more practical management constraints (Best 2004; Best and Wayburn 2001). Pan et al. (2007) found that forest owners with larger properties put more time and effort into management and with the higher potential incomes from it, are more willing to acquire plans. However, the effect of FMP to reduce the barrier to harvest among small owner could be important.

Our study results show that a forest owner’s perceived relevance of the FMP is an important predictor for the familiarity and use of the FMP, as 80% of owners who consider that the FMP is relevant, know the content of the plan well and 61% implement the proposals of the plan. Earlier studies (Davis and Fly 2010; Ficko and Boncina 2015) suggest that forest owners’ participation is lower at the implementation stage of forest planning because the FMP are not adapted to their objectives and understanding of forest management which can result in forest owners staying away from outreach and extension programs (Davis and Fly 2010). Hujala et al. (2007) found that the perceived FMP meaningfulness is higher if the plan is adapted to the forest owner’s objectives. In Norway, most FMP are based on standardized “best practice” timber-production forest management even if non-timber objectives are important among owners (Sjølie et al. 2019) which possibly creates a discrepancy between forest owners’ objectives and the FMP content and proposals, reducing its considered relevance and acting as a barrier for enhanced use of the FMP.

In addition to the plan adaptation variable that only was part of the second stage, counties and contact with the municipality forestry section for advice were significant across all trees and regressions. The marginal effect of the latter variable in the ImplementFMP model is as high as 38%, indicating that improving outreach through the public management may contribute to realizing not only the extensiveness of the FMP, but also considerably its potentials as a decision-aid tool for forest management planning. County effects are strong, and even if they vary between the models, the general picture is that forest owners in the traditional forest regions in Eastern Norway have 10–20% higher likelihood to acquire, knowing the content and implement proposals of FMP (Tables 5, 6 and 7; Figs. 2, 3 and 4) than in other regions. Large regional variation in NIPF owners’ possessions of FMPs was likewise reported by studies in the US (Butler and Leatherberry 2004; Measells et al. 2005; Rasamoelina et al. 2016).

Looking ahead, some studies suggest that in the near future, economic constraints may drive alternatives for FMP. As a consequence, web-based training can be more utilized as a tool to increase activity on NIPF owners’ landbase (Butler et al. 2014; Kueper et al. 2014; Sagor et al. 2014). However, these kinds of initiatives possess certain challenges like reaching out to owners with lower digital skills (VanBrakle 2015). Therefore, including these factors in future studies is important for proper understanding of forest-owners' decision-making towards FMP.

Our study used a combined approach of conditional classification tree and logistic regression models to visualize and estimate factors explaining forest owners’ FMP behavior. There are relevant studies that have utilized other approaches like structural equation modelling (Ficko and Boncina 2015), factor analysis (Koskela and Karppinen, 2021) and cluster analysis (Majumdar et al. 2008; Butler et al. 2022) to unveil the factors influencing forest owner behavior. However, we believe that our combined method provided new information as these two techniques complement each other. While the DT point to the forest owner subgroups and thresholds for a positive response and reveal unconsidered structures in the data, the probit regressions provide the estimated impact of each independent variable and sort all variables according to their importance.

The questions of our study is relevant to countries with a private forest landbase. However, interpreting the results of our study, there are some uncertainties associated with the data and methods that should be kept in mind. The dataset contained missing values due to incomplete responses from the respondents varying from 5 to 20% across all the variables. We utilized multiple imputation to mitigate its impacts. The number of respondents responding to a survey question may be different from those not responding. For example, in our study active or engaged owners might have responded more often than inactive owners, thus creating biases. Similar results were reported by Butler et al. (2016) and Butler et al. (2022). In addition, the survey data was collected in 2014. Therefore, there is need to continue studies along the same lines to validate the results and observe any changes that might have occurred due to the digital transition in FMP during the last decade. This transistion may have reduced the adoption barrier for some groups of forest owners. However, given the demographic distribution of forest owners, it is an open question whether this transition has imposed new barriers to some groups of owners. With data mining, we unveiled important variables that otherwise might have been overlooked. We believe the insight from our study can act as a foundation for further research. However, some variables might need to be updated due to the digital transition in FMP. It should be kept in mind that the DT have a tendency to overfit the training data leading to poor predictions (Shawar et al. 2021). However, combining the DT with the regression models provide more robust evaluation of the important predictors.

Conclusions

FMP have a paramount role to play in the sustainable management of forests. The results of this study indicate that the forestry sections in the municipalities can play an important role by reaching out and aiding more forest owners to use the FMP as a decision-support tool. Furthermore, for the FMP to be actually used, it has to be deemed relevant by the forest owners, thus adapted to the forest owner’s objectives. We show that owners with and without plan differ across a broad set of variables, indicating that actions to stimulating the interest of owners who do not acquire FMP could be targeted towards specific owner groups. Information and forest management factors were dominating in explaining the adoption of FMP; this insight may aid in triggering the diffusion. High-quality information encourages and educates forest owners to execute efficient forest management decisions and maximize the benefits from their forests. For reaching that goal, policymakers should be well informed about forest management decisions pathway of FMP, from acquisition to implementation of FMP among NIPF owners.

Our study can be utilized by FMP suppliers for adapting products and services, forestry consultants, bureaucracy and policy-makers to better reach out to various forest owner groups to increase the success of the FMP as a decision-aid tool and forest policy instrument. FMP in Norway are highly subsidized; enhancing its actual use will increase the cost-effectiveness and efficiency of this policy instrument.