**4. Discussion**

The background information provided by our 51 experts demonstrates that they qualified as such given their backgrounds including work experience. The experts' backgrounds also shows that our sourced experts cover a wide range of positions and institutions so that bias towards interests of a particular group of researchers or decision makers is unlikely. Our top 10 list of key knowledge gaps addresses knowledge gaps from all three spheres of sustainability consisting of three ecological, three socio–economic knowledge gaps, four political, as well as two general knowledge gaps. None of the identified top 10 knowledge gaps was directly related to SDG 2 (Reduce Hunger), only four to SDG 10 (Reduce inequality) and three to SDG 15 (Life on Land). In the following paragraphs, we will first discuss the top 10 knowledge gaps by linking them to all SDGs, ABTs and the 2 ◦C target of the Paris Climate Agreement and then show recent local or larger scale attempts starting to address these knowledge gaps. Acknowledging the implications of our study and its limitations, we will provide suggestions for further work.

#### *4.1. Top 10 Knowledge Gaps*

The first two overall top priority knowledge gaps address the need to find "strategies to improve justice, fairness, enforcement of laws/rules" and to "gain more knowledge about the role of corruption as well as how to reduce it". Both knowledge gaps are strongly linked to another knowledge gap in the top 10 ranking: "Strategies to improve security from violence/theft/corruption" (rank 5). All three are addressed in SDG 16—one of the nine SDGs Madagascar failed to achieve according to the Africa SDG Index and Dashboard Report 2018 [45]. The issue of justice and fairness is also related to SDG 10 (Reduced inequalities), ye<sup>t</sup> for African countries, no metrics exist for the achievement of this SDG [45]. So far, the issue of corruption has been completely lacking from the ABTs [15]. Lack of law enforcement, Zebu cattle theft and burglaries remain important issues for the rural population and the attempts to reduce corruption via an Anti-corruption Commission in 2002, Anti-Corruption Agency in 2004, or a governmental decree to ban illegal logging in 2005 have not helped to significantly reduce illegal activities [39,90]. Achieving better law/rule enforcement and fighting corruption is di fficult as politicians, higher o fficials and local elites are often themselves involved in bribery or put no e fforts in law enforcement for monetary, social or other benefits [91–94]. Corruption considerably undermines the protection of nature [95] as it reduces law enforcement and investments by international conservation agencies [91,95]. Since poor people often rely on the extraction of resources from protected areas to secure protein supply or to cover other subsistence needs, conservation laws and rules are often perceived as unfair and have been internationally criticized as "green-grabbing" [23,96,97]. This issue

is addressed in the socio–economic knowledge gap "Economic benefits for local small-holder farmers from biodiversity" ranking 4th overall (see further below).

Internationally, the most well-known type of corruption in Madagascar is probably associated with the illegal trade of precious wood species (particularly *Dalbergia* spp.) due to the involvement of politicians and international media attention. Madagascar's new president Andry Rajoelina who was elected in December 2018 promised to make the fight against corruption a priority [98]. However, given his past involvement in illegal rosewood trade, his agreemen<sup>t</sup> to reinstate the ban of rosewood trade that lacked any reinforcement [41], and governmen<sup>t</sup> corruption levels which spiked under his de facto presidency between 2009 and 2013, many, particularly the international community, doubt his promise to curb corruption [99].

Since the illegal rosewood trade occurs predominantly in eastern Madagascar [100] and also because it is only one of many corruption issues, we deliberately included "corruption beyond logging" in this knowledge gap. Corruption affects all sorts of sectors, for example funds destined for education [101], the church [102], undermining of local land rights, and access associated with agribusinesses and mining operations [39]. Political instability and multiple political crises have resulted in difficulties to establish or monitor new policies [54]. This has been accompanied by an increase in deforestation for agriculture [63] as well as in other types of illegal resource exploitation [33,42,69,100]. Those trying to fight corruption (e.g., local forest guards, environmental groups, members of watchdog organizations, researchers) live dangerous lives as they have to fear being evicted as "rebellions" or being confronted with death threats to them and their family members [96,103].

Corruption indicators used to evaluate the SDG 16 (e.g., number of victims of intentional homicide, conflict-related deaths per 100,000 population, proportion of children who experienced any physical punishment and/or psychological aggression) lack many of the above described components of corruption issues in Madagascar [104]. The International Corruption Perception Index relies on experts' opinions regarding transparency, accountability and corruption in the public sector [44], even though Transparency International experts are mainly business people who seem ill-suited to represent locally affected residents. Involving the local communities via questionnaires or interviews as done by Gore et al. (2013) for a different conservation area in Madagascar would give a better understanding of the "hotspots" of corruption activities and thus easier ways to tackle these issues right where they occur. More general strategies to fight corruption would be fair wages, stringent accounting procedures and managemen<sup>t</sup> partnerships [95]. Media coverage of corruption activities and the work of brave activists to fight them may also be helpful [101], as is pressure from the international community [91].

The highest ranking ecological knowledge gap (third highest overall rank) directly addresses the need to know more about appropriate restoration beneficial to the protection and sustainable use of biodiversity. Restoration is key to providing essential ecosystem services (ABT 14) and has been given increased attention by the Global Partnership for Forest Restoration and the Bonn Challenge (e.g., Ockenden et al. 2018). Ecosystem restoration and resilience is explicitly mentioned in ABT 15, but not in the targets of SDG 15. It is linked to SDG 12 and ABTs 4 and 7 due its consequences for sustainable production in terms of most sustainable reforestation methods. Technological advances (Perring et al. 2015), green finance options (FAO and Global Mechanism of the UNCD, 2015) and compensation measures, particularly in regards to telecoupling effects, can all support restoration activities (e.g., Ockenden et al. 2018). Restoration projects require long-term efforts of monitoring as well as associated adjustments to effectively protect biodiversity and enable sustainable use of resources within restored areas. Given that some species will be unable to cope with even slight anthropogenic disturbance, some restoration areas, particularly corridors enabling migration between remaining habitat patches, will have to meet the needs of these sensitive species. Other restoration areas allowing for human resource utilization may still provide habitats for a number of less sensitive species. "Strategies to secure long-term funding" (rank 6) will be necessary to ensure that restoration will be sustainable.

If wisely done, reforestation can be extremely beneficial for biodiversity by providing suitable habitat connected by corridors or stepping stones [105,106]. Madagascar committed to the Bonn Challenge to reforest four million hectares of forest, but little progress has been made and the benefits only address the economy and climate [40,107]. Madagascar's National Biodiversity Strategy and Action Plan (NBSAP) lists restoration of at least 15% degraded habitats as one of its five strategic goals (Objective 15 in [108], p. 89). As a solution it mentions agro–ecological techniques as "e ffective tools for the degraded vegetation" but does not provide any details on what these tools to restore degraded vegetation are. Similarly, "appropriate strategies are set up to safeguard these ecosystems [ ... ] for human well-being especially local communities through restoration activities" are mentioned, but actual strategies on how this can be achieved are only vaguely mentioned: "Tools or Handbook of conservation and/or Ecological restoration of various existing tropical forest types are developed" [ibid.] (p. 106) and "Number of recovery programs for protected areas of degraded ecosystems is developed and implemented" [ibid.] (p. 114). The knowledge gap concerning appropriate restoration strategies is also linked to SDG 13 and the 2 ◦C target of the Paris Climate Agreement as restoration will help mitigate climate change due to increased carbon storage, a decreased albedo and associated effects caused by revegetation measures [105,109,110].

The socio–economic knowledge gap considered most important (overall rank 4), "Economic benefits for local small-holder farmers from biodiversity", is directly related to SDG 8 (Decent work) in combination with the protection of biodiversity (SDG 15, ABT 11) and ecosystem services (ABT 14) and explicitly addressed as important in Madagascar's NBSAP [Objective 2 in 108] (p. 70f). Herein, it has been acknowledged that a fair distribution of benefits of ecosystem services requires more research, and where paymen<sup>t</sup> schemes for ecosystem services have been successful these approaches should be used as a guideline and adjusted to local contexts [65,111–113]. To date, most of these payments never covered the opportunity costs locals had to endure for the conservation of biodiversity, patrols were reluctant to convict fellow community members of illegal actions, and positive trends were reported although forest degradation actually increased [111,114,115].

Payments for ecosystem services like those within the REDD program (Reduced Emissions from Deforestation and Forest Degradation) can only be successful under certain conditions [115]. MacKinnon et al. (2017) reported that the amount of money reaching a community di ffered strongly between methods and that projects directly compensating for the loss of forest resources (e.g., by fish farming or bee keeping) were the most successful. Anticipated and promoted benefits for local people through tourism [69] rarely materialized [48,116], although there have been some cases of livelihood benefits from eco-tourism in a few tourism hotspots [114]. However, usually these benefited only a few people and not whole communities [102,117].

More knowledge regarding "strategies to improve security from violence/theft/corruption" has been the fifth highest ranking knowledge gap. This comes as no surprise, as Malagasy people have also ranked crime and insecurity as top priority issues for the governmen<sup>t</sup> [118]. This knowledge gap concerns aspects of inequality (SDG 10) as poor people are most a ffected by these issues, but also to SDG 16 as strong institutions are necessary to tackle these issues. To avoid cattle theft, which is considered as "extremely worrisome" [119] cattle are kept hidden in forests, which promotes forest degradation and hampers forest regeneration [39]. People who have no choice but to continue to use forest resources for their survival also fear violence and fines from forest police [119]. Since law enforcement is largely lacking, laws are no deterrent for committing crimes in Madagascar, but an increase in law enforcement personnel locally and at least temporally decreases cattle theft (see [120] and references therein). Residents from villages with missions, police or military presence feel safer [119]. However, in remote areas, governmental control is particularly weak (or even completely lacking) [103]. While cattle theft is usually practiced by well-organized groups which cooperate with local authorities, burglaries and crop theft are rather associated with poverty and hunger of the rural poor as a coping strategy under high survival risk [120].

"Identification of strategies to improve long-term funding" has been identified as the 6th highest ranking knowledge gap and relates to all four spheres of sustainability [82]. Limited and short-term funding have been impeding long-term monitoring and successful participatory strategies [119]. Securing long-term funding for research, environmental as well as human aid projects is a global problem which is why this issue is picked up in ABT 20 (Mobilizing resources from all sources). It is strongly tied to the issue of short election cycles, awareness of the benefits of long-term funding instead of funding invested in many short-term projects and an issue of changes in staff when running long-term projects.

We combined the aspects of appropriate livestock managemen<sup>t</sup> (here zebu cattle farming) and fire regimes in one knowledge gap (see Table 1, rank 7) as both have experienced little research and both are closely interrelated: grasslands, woodlands and forests are burnt to provide fresh pastures for zebu cattle and even where fire is not deliberately set, these pasture fires often escape unintentionally and burn adjacent habitats [39,121–123]. This knowledge gap is linked to the protection of biodiversity (SDG 15, ABT 7), sustainable managemen<sup>t</sup> (SDG 12, ABT 4 and 7) and also to climate change (SDG 13, 2 ◦C target) as fire and cattle farming increase atmospheric CO2-levels. Research needs for "appropriate managemen<sup>t</sup> of livestock and fire regimes" are only partly addressed in Madagascar's National Biodiversity Strategy and Action Plan [108]: "programs aimed at strengthening the control of bush fires" and those that minimize fire impacts in areas with significant biodiversity by creating effective buffer zones around protected areas and more training of fire extinguishing personnel are explicitly mentioned [ibid.] (p. 107). Appropriate and more sustainable livestock managemen<sup>t</sup> are not addressed though, indicating that this issue has not even gained due attention.

Zebu farming has a negligible role as a protein source, but cattle have important cultural value (slaughtered for special cultural occasions) and herd size indicates social status [39,122,124]. Zebu cattle are also used for agricultural labor and transport [125] and as an insurance asset to buffer income loss during times of hardship like droughts or low crop market prices [119,122,126]. The cultural importance of zebu cattle is particularly high for temporary Antandroy migrants from the south who use slash-and-burn agriculture for cash crops (particularly corn) to purchase cattle and thus increase their social status when returning home [119,125]. Climate change has already induced higher frequency and intensity of droughts in southern Madagascar which has led to increased migration of Antandroy from southern Madagascar into central western Madagascar severely increasing the pressure on the dry forests [39,122]. Traders involved in illegal corn and peanut plantation farms actively lure these migrants in for illegally clearing the forests for them by promising them quick and easy paymen<sup>t</sup> for each hectare of forest cleared as well as covering their travel and accommodation costs [127]. This climate change-induced migration issue has been the only additional knowledge gap suggested by several experts in our study. We agree that more knowledge about this issue and strategies to mitigate it via prosecution of and high fines for traffickers have a very high priority and that it should be on the priority list of current decision makers [40].

The third highest ranking ecological knowledge gap (overall rank 8) concerned ecosystem services at risk from slash-and-burn farming. Like the previous ecological knowledge gap concerning livestock and fire management, it is linked to biodiversity (SDG 15, ABT 7) and ecosystem service loss (ABT 14), as well as to consequences of fire on CO2-levels and hence climate change (SDG 13, 2 ◦C target). We principally know that dry forests provide essential ecosystem services and that these are at risk from slash-and-burn farming but ranking this knowledge gap so high indicates that our understanding of which ecosystem services are affected with what potential cascading consequences is not ye<sup>t</sup> perceived as being sufficient. Madagascar's NBSAP is strongly committed to ensure future provisioning of ecosystem services through the protection of 10% of terrestrial ecosystems, sustainable managemen<sup>t</sup> (including certification), sustainable tourism, compensation activities like restoration, as well as additional studies to develop and implement appropriate strategies [108].

This knowledge gap has direct consequences for the sociological knowledge gap concerning the effectiveness of education and awareness-raising on biodiversity conservation linked to SDG 4 (Quality education) and ABT 1 (Awareness of biodiversity has increased). People who understand the implications of their actions are less likely to undertake them if they anticipate negative consequences for themselves or their families. Creating this understanding is particularly di fficult when the e ffects of actions are complex and only visible in the more distant future. Better education may increase chances of income diversification [128] and improve community forest managemen<sup>t</sup> [113]. Both, education and awareness raising are essential for the protection of biodiversity as they can lead to more understanding of ecosystem services nature provides [129]. For example, Reibelt, et al. [130] have shown that in south–western Madagascar, awareness of local people rises through direct contact with endangered species on which they rely for subsistence.

One of the two knowledge gaps at rank 9 concerns the improvement of sustainable methods by "more frequent and regular scenario-updates based on long-term monitoring". This general knowledge gap is of course directly related to the protection of biodiversity (SDG 15, ABT 7). Building the necessary trust for participatory approaches, negotiating rules and calculating long-term costs and benefits of biodiversity loss are all time-consuming activities. The e fforts required for collecting solid data at realistic temporal scales need to be considered in biodiversity strategies and action plans (ABT 17). We considered this knowledge gap to be a general one as long-term monitoring for more frequent and regular scenario updates is also important in socio–economic and political planning, like long-term costs and successes of education programs (SDG 4) or the e ffectiveness of PES on political decisions.

Demonstrating how interrelated our identified key knowledge gaps are, the knowledge gap concerning appropriate strategies for restoration (overall rank 3) also relies on long-term monitoring—which again depends on long-term funding (rank 6), and consequently they all are strongly linked to ABT 20 ("Mobilizing resources from all sources").

The other knowledge gap at rank 9 "Traditional knowledge about sustainable natural resource use" belongs to the socio–economic sphere of sustainability and is directly linked to ABT 18 (Traditional knowledge): "In 2015, the initiatives put in place to protect traditional knowledge, innovations and practices of local communities relevant to biodiversity. The traditional sustainable use of biodiversity and their contribution to conservation are respected, preserved and maintained" which has been given funding priority by Madagascar's NBSAP when compared to restoration (Objective 15) and the evaluation of ecosystem services for PES schemes (Objective 2) (both in [108], (p. 154f)).

We need to become aware of traditional knowledge before it will be lost and to find out how traditional conservation values overlap with those of scientists [131]. Local people's knowledge has been shown to provide important insights for sustainable land managemen<sup>t</sup> and use of resources in south-western Madagascar [132]. Several authors have shown that to achieve e ffective conservation in Madagascar, every ethnic group should be integrated from the start [133,134]. More use should be made of existing forms of traditional, local agreements and institutions, because such local rules are generally more adhered to [133,135].

The lowest ranking knowledge gap within our top 10 list "Interdisciplinary work to generate most comprehensive data sets" is strongly linked with the higher-ranking ones. It is in some ways addressed in SDG 17 (Partnerships of the goals) and ABT 19 (Sharing information and knowledge) as well as ABT 20 (Mobilizing resources from all sources). In 2015, the journal Nature dedicated a special issue on the topic of interdisciplinarity demonstrating why "scientists must work together to save the world" [136]. However, interdisciplinarity should go beyond that of scientists and–in a transdisciplinary way–include knowledge from other sources like that of local/indigenous people or organizations with long-term on the ground experiences. Interdisciplinary work has been considered particularly important for studying the ecological impacts of climate change [137]. Finding ways to deal with environmental challenges associated with such impacts and their consequences for human well-being requires incorporating social and human-centered approaches by using participatory approaches [ibid.].

Most knowledge gaps which focused on generating more data within their own discipline received the lowest mean scores (Table S3), potentially indicating that our experts consider it more important to make use of existing data by using it in an interdisciplinary way than investing money into research of deficient, but highly specific research fields. A focus on improving decision-making and a call for more action rather than further investment in data collection has also been proposed by others [138]. However, we agree with Stuart et al. [139] that investing into more data generation, particularly those concerning threatened and data deficient species, remains important for conducting appropriate actions and adjusting managemen<sup>t</sup> strategies as needed.

#### *4.2. E*ff*ects of Di*ff*erent Expert Backgrounds on Ranking Scores*

Our study showed that our top 10 list of key knowledge gaps would not have di ffered, would we have conducted this study without experts from other countries, only with women or only with experts of a certain age group, except for the knowledge gap concerning frequent and regular scenario updates which was considered of higher importance by Malagasy than other nationalities (Table S4, Figure S1b). Given the influence of place attachment, gender and age on the judgement of conservation related issues which has been pointed out by others [86], we are pleased about predominantly Malagasy experts ranking our knowledge gaps, about the relatively even gender ratio and that the bulk of our experts were medium and older aged and had long working experience in the area.

#### *4.3. Limitations, Implications and Recommendations for Future Work*

We are aware of the limitations of this study. To keep the questionnaire short, some knowledge gaps were already combined. Ideally, knowledge gaps should be kept separate and maybe in addition also presented in combination to give participants the option to score whether they found the combination more important than its parts. We tried to accommodate for this by giving the opportunity to express this in the comments sections or by listing additional highest priority knowledge gaps. Network sampling of experts as well as expert consultations in a relatively short time frame, which are typical for many projects, always have limitations as they are rarely comprehensive [84]. Our questionnaires were in English; future studies should include versions in French and potentially even Malagasy to be more inclusive. For other case studies, particularly those with more time and resources as well as those with a narrower, i.e., local focus, we recommend involving experts as well as local people right from the beginning, i.e., when the questions to be ranked are gathered [21]. To avoid translation problems when involving local people, highly skilled translators will be necessary. These need to be trained not only in translating western conservation ideas into local language and values, but also and most likely more importantly, the locals' values, beliefs and proverbs into concepts that can be integrated into conservation policy [116]. Even participatory approaches involving community members cannot claim to be the gold standard [140]. Since not every community member can be involved, participation is usually only possible for high ranking representatives of communities. These are usually elderly men so that a gender and age bias in selecting key knowledge gaps may become an issue. However, we agree that, given more time and resources, a Delphi-like process (i.e., at least two rounds of questionnaires, each accompanied by accumulation of responses and anonymous feedback to participants) should be used to prioritize knowledge gaps [21]. Ideally, representatives of local communities should be involved, particularly when attempting knowledge gap prioritization at a local level. We tried to accommodate for the lack of expert engagemen<sup>t</sup> from the start by providing an option to list additional knowledge gaps. We also acknowledge that the number of participants was relatively low and dominated by academics due to the networks of the authors. Despite those limitations, we hope that our work substantially contributes to pinpointing the most urgen<sup>t</sup> knowledge gaps which need to be addressed to secure local livelihoods and better protect biodiversity in the dry forests of western Madagascar–and hence better achieve the SDGs, ABTs and the 2 ◦C target of the Paris Climate Agreement. Our priority list of knowledge gaps should be considered a first attempt of identifying most urgen<sup>t</sup> research needs and help funding bodies to streamline investments at the regional level (i.e., the dry forests of western Madagascar). Like the global research questions identified for example by Oldekopp et al. (2016), the knowledge gaps identified for the slash-and-burn problematic in western Madagascar can be used at a regional level as a starting point for research project designs, collaborations and debates between academics, practitioners, politicians and stakeholders.
