Bayesian Network Approach [Brazil]

Abordagem com Redes Bayesianas

approaches_1975 - Brazil

Completeness: 92%

1. General information

1.2 Contact details of resource persons and institutions involved in the assessment and documentation of the Approach

Key resource person(s)

SLM specialist:
SLM specialist:
SLM specialist:

Steinmetz Liron

Berlin Institute of Technology (Technische Universität Berlin), Environmental Assessment and Planning Research Group

Secr. EB 5, Straße des 17. Juni 145, 10623 Berlin, Germany

Name of project which facilitated the documentation/ evaluation of the Approach (if relevant)
Book project: Making sense of research for sustainable land management (GLUES)

1.3 Conditions regarding the use of data documented through WOCAT

When were the data compiled (in the field)?


The compiler and key resource person(s) accept the conditions regarding the use of data documented through WOCAT:


2. Description of the SLM Approach

2.1 Short description of the Approach

Assessment of the probability and effectiveness of management options or innovations to describe cause-effect-relationships and to make recommendations for action on sustainable land management in the Itaparica region in Northeast Brazil.

2.2 Detailed description of the Approach

Detailed description of the Approach:

The study focus was on the determinants behind the adoption of innovations developed under a scientific project. The specific innovation analysed was intended to benefit both the environment and local smallholder farmers: namely cultivating a multi-purpose, low-growing, to the prevailing harsh semiarid environment well-adapted tree species (Spondias tuberosa L. – so called umbuzeiro). The assessment method was selected as it allows the combination of qualitative and quantitative data, and can be applied even in data-scarce situations. Moreover, it allows downscaling from a broad overview to small-scale management.

Knowledge is collected from different disciplines to support decision-making through the inter- and transdisciplinary approaches of constellation analysis and Bayesian networks. A Bayesian Network (BN) is a probabilistic graphical model that represents a set of variables (elements, nodes) and their conditional dependencies. There are three input components to a Bayesian Network: (a) a set of elements representing factors relevant to a particular environmental system or problem, (b) the links between these elements, and (c) the conditional probability tables (CPTs) behind each node (element) used to calculate the state of the node. Collected data and ratings are arranged in a hierarchical Bayesian Network model in Netica software (Netica 5.12 - freeware up to 15 nodes).

The creation of a Bayesian Network model is as follows: the objectives and necessary interventions for the innovation process aimed at sustainable management are characterized, with scientists arranging a conceptual diagram, including the mapping of elements. States of the nodes are determined through study of the literature and expert consultation (by scientists, stakeholders and experts on related topics). In a final step, a sensitivity analysis is performed on the Bayesian Network to highlight crucial nodes with the highest influence on objectives in order to derive actions to be recommended.

Stakeholder participation is the core process of designing Bayesian Networks. In pre-consultations stakeholders help identifying major influencing factors and relationships. Assessments are compiled in interview sessions enabling the states of the nodes to be quantified later. In this case study, the stakeholders were farmers, farmer-supporting institutions, and expert in soils, vegetation and crops.

2.3 Photos of the Approach

2.5 Country/ region/ locations where the Approach has been applied



Region/ State/ Province:

Pernambuco, Brazil

Further specification of location:

Itaparica Reservoir, Petrolândia

2.6 Dates of initiation and termination of the Approach

Indicate year of initiation:


Year of termination (if Approach is no longer applied):


2.8 Main aims/ objectives of the Approach

The principle of Bayesian Network modelling is the integration of multiple issues and system components, where information from different sources can be integrated, while also handling missing data and uncertainty. The outcome may be recommendations that support local management decision-making. As the method is strong in transdisciplinary knowledge integration, it has the potential to become one of the core methods in environmental management.

2.9 Conditions enabling or hindering implementation of the Technology/ Technologies applied under the Approach

social/ cultural/ religious norms and values
  • enabling
availability/ access to financial resources and services
  • enabling

Potential for financial support could be through national small-scale farmer programs; suitable government-sponsored credit programs, public and governmental institutions such as bulk purchasers of agricultural commodities (for instance SEBRAE in Brazil).

legal framework (land tenure, land and water use rights)
  • enabling
knowledge about SLM, access to technical support
  • enabling

Use available free programmes. Use of visual aids such as smileys for evaluation to make questionnaire more comprehensible.

3. Participation and roles of stakeholders involved

3.1 Stakeholders involved in the Approach and their roles

  • local land users/ local communities

Farmers of a resettlement community on dryland; Representatives of the indigenous tribe of Pankararu

  • SLM specialists/ agricultural advisers

Experts in soil and crop sciences; Expert in vegetation and biodiversity science of the Caatinga

  • national government (planners, decision-makers)

Institute of Agriculture in Pernambuco (IPA); A private company as the hired institution by the National Institute for Colonization and Agrarian Reform - INCRA

  • Company of plant breeding, seed science
3.2 Involvement of local land users/ local communities in the different phases of the Approach
Involvement of local land users/ local communities Specify who was involved and describe activities
initiation/ motivation passive Interviews
planning none
implementation interactive Interviews
monitoring/ evaluation interactive Interviews

3.3 Flow chart (if available)


Simplified work flow of Bayesian Network (BN) showing different steps:
­ Defining: apply or use already applied constellation analysis (see A_BRA003en) for information and visualization of node setting for the BN model and for stakeholder identification.
­ Identifying: clarify objectives, implementation factors, interventions, intermediates and controlling factors. Give every node a state, e.g. date, temperature range, amount of precipitation, or a classification: high / low…
­ Building: Collect data to fill the conditional probability tables (CPTs) behind every node. Prepare questionnaires, ask experts and conduct a literature search. Avoid too much states and no more than four nodes indicating the next node. Finish the model by entering all data in a programme (e.g. Netica).
­ Evaluating: Compare different scenarios by changing the state of inputs (e.g. from low to high). Show a baseline (without changes), a most improved and least improved scenario to justify recommendations. Finally, hand over recommended actions to stakeholders.


Liron Steinmetz, Verona Rodorff

3.4 Decision-making on the selection of SLM Technology/ Technologies

Were decisions on the selection of the Technology(ies) made:
  • The approach was initiated by scientists.

The approach can be applied to SLM technologies, but also can be employed for other purposes. In our data-scarce case, the method was very helpful as able to deal with different data sources and types.

Specify on what basis decisions were made:
  • research findings

4. Technical support, capacity building, and knowledge management

4.1 Capacity building/ training

Was training provided to land users/ other stakeholders?


Specify who was trained:
  • land users
  • field staff/ advisers
Form of training:
  • farmer-to-farmer
  • demonstration areas
Form of training:
  • workshops
Subjects covered:

Detecting decisive factors for an ideal scenario of implementation being adopted by land users. For the participants it was interesting to participate in preparing a joint view of their action space - this is generally known in its parts though not with its major interconnections and complexity. Participants especially acknowledged this value added for them.

4.2 Advisory service

Do land users have access to an advisory service?


4.3 Institution strengthening (organizational development)

Have institutions been established or strengthened through the Approach?
  • yes, greatly
Specify the level(s) at which institutions have been strengthened or established:
  • local
Describe institution, roles and responsibilities, members, etc.

Decisive factors for the adoption of innovations were identified, including favoring cultivation techniques for Umbuzeiro agriculture (e.g. soil additives).

Specify type of support:
  • capacity building/ training

4.4 Monitoring and evaluation

Is monitoring and evaluation part of the Approach?


If yes, is this documentation intended to be used for monitoring and evaluation?


4.5 Research

Was research part of the Approach?


Specify topics:
  • sociology
  • economics / marketing
  • ecology
  • technology
Give further details and indicate who did the research:

Research on the situation of local action and governance was a major driver for the workshops. University project members prepared and held the workshops, while also did extended interpretation and integration of results across a number of different workshops.

5. Financing and external material support

5.1 Annual budget for the SLM component of the Approach

If precise annual budget is not known, indicate range:
  • < 2,000
Comments (e.g. main sources of funding/ major donors):

German Federal Ministry of Education and Research (BMBF) 100%

5.2 Financial/ material support provided to land users

Did land users receive financial/ material support for implementing the Technology/ Technologies?


5.3 Subsidies for specific inputs (including labour)

  • none
If labour by land users was a substantial input, was it:
  • voluntary

5.4 Credit

Was credit provided under the Approach for SLM activities?


5.5 Other incentives or instruments

Were other incentives or instruments used to promote implementation of SLM Technologies?


6. Impact analysis and concluding statements

6.1 Impacts of the Approach

Did the Approach help land users to implement and maintain SLM Technologies?
  • No
  • Yes, little
  • Yes, moderately
  • Yes, greatly

The different scenarios of the BN tested highlight the good probability of adoption, which then can support sustainable land management.

Did the Approach empower socially and economically disadvantaged groups?
  • No
  • Yes, little
  • Yes, moderately
  • Yes, greatly

The approach was conducted especially for smallscale farmers without sophisticated irrigation tecniques and as well for the indigenous tribe Pankararu.

Did the Approach improve issues of land tenure/ user rights that hindered implementation of SLM Technologies?
  • No
  • Yes, little
  • Yes, moderately
  • Yes, greatly

The BN model offers alternative sources for soil additives in case land use rights are hindering availability.

Did the Approach lead to improved food security/ improved nutrition?
  • No
  • Yes, little
  • Yes, moderately
  • Yes, greatly

Not inmediately but a long-term influence is possible.

Did the Approach lead to improved livelihoods / human well-being?
  • No
  • Yes, little
  • Yes, moderately
  • Yes, greatly

The different scenarios of the BN tested highlight the good probability of adoption, which can then benefit the livelihoods of adopters.

6.2 Main motivation of land users to implement SLM

  • increased production
  • increased profit(ability), improved cost-benefit-ratio
  • payments/ subsidies
  • environmental consciousness
  • well-being and livelihoods improvement

6.3 Sustainability of Approach activities

Can the land users sustain what has been implemented through the Approach (without external support)?
  • yes
If yes, describe how:

Lessons learnt (especially on most favorable soil additive mixture) improve effectiveness of potential Umbuzeiro cultivation. Stakeholder pool of BN-creation comprises business networking opportunities for land users.

6.4 Strengths/ advantages of the Approach

Strengths/ advantages/ opportunities in the compiler’s or other key resource person’s view
The combination of input variables from any given background is possible.
Via Bayesian networks changes to the modelled system can be tested prospectively. The space and potential effects of management options can be shown to decision makers.
Combining Bayesian networks with Constellation Analysis allows easy determination of major nodes of the model and supports the process of decision-making for sustainable land management activities; methods proved to be very transdisciplinary.

6.5 Weaknesses/ disadvantages of the Approach and ways of overcoming them

Weaknesses/ disadvantages/ risks in the compiler’s or other key resource person’s view How can they be overcome?
The statistical component of the Bayesian network approach can be hard to grasp for less educated or near-illiterate stakeholder groups. Percentages of probability estimations can be translated to a graphical equivalent (e.g. gradual scale of emoticons).

7. References and links

7.1 Methods/ sources of information

  • field visits, field surveys
  • interviews with land users

7.2 References to available publications

Title, author, year, ISBN:

Rodorff V., Steinmetz L., Siegmund-Schultze M., Köppel J. (2015) Using Bayesian networks to depict favouring frame conditions for sustainable land management: Umbuzeiro-tree planting by smallholders in Brazil

Available from where? Costs?

Session: Methods, tools and impact applications. Tropentag ‘Management of land use systems for enhanced food security - conflicts, controversies and resolutions’, September 16 – 18, 2015, Humboldt-Universität zu Berlin, Berlin, Germany

7.3 Links to relevant information which is available online

Title/ description:

Innovate project information


Links and modules

Expand all Collapse all