A Proposal for Climate Change Resilience Management through Fuzzy

Controllers

J. Rub

en G. C

ardenas

Angela Nebot

and Francisco Mugica

IUSS UME School, Via Ferrata 45, Pavia, Italy

Soft Computing Group, Universitat Polit

ecnica de Catalunya, Jordi Girona Salgado 1-3, Barcelona, Spain

Keywords:

Fuzzy Sets, Risk Management, Natural Hazards, Social Vulnerability, Resilience Management, Fuzzy

Controllers, Inference System, Decision Support System, Artiﬁcial Intelligence.

Abstract:

We aim towards the implementation of a set of fuzzy controllers capable to perform automated estimation of

the period of time necessary to recover a resilience level through the non-linear inﬂuence of a set of interrelated

climate change resilience indicators constrained by social-based variables. This fuzzy controller set, working

together with a fuzzy inference system type Mamdani, will be capable to estimate the proper adjustments to be

done onto system’s elements in order to achieve a certain resilience level, while a general estimation of required

costs is appraised. The ﬁnal tool can then be used to provide guidelines for strategic vulnerability planning

and monitoring through a clear understanding between investments and results, while an open evaluation and

scrutiny of applied policies is made. In this paper the main strategy to achieve the mentioned objectives is

presented and discussed.

1 INTRODUCTION

Resilience is one of the risk components that might

be viewed as real interface between analysis and de-

cisions. Moreover, resilience can be considered as

a linkage between mother’s Nature uncertainty and

social-based structures. Such as risk, in order to

enhance societies resilience capacities, a transparent

and consistent decision framework must be designed

while including an assertion of its capacity to be im-

plemented over those ofﬁcial bodies responsible of

binding either social behaviors and/or interactions.

Up to some point, it is when these social institutions

fails that susceptibility to external or internal stressors

increases, and the whole management structure seems

to be compromised. Resilience however, is a concept

not completely well understood, even if its ﬁnal aim

is in somehow clear.

It is by correlating the concept of recovery to real

factors, that a true improvement on resilience strate-

gies assessments and implementations on a real sce-

nario can be achieved, however the shift from qualita-

tive resilience models towards quantitative resilience

assessment models represents a very recent branch in

the ﬁeld of disaster risk management. Consequently,

most of techniques dealing with disaster resilience

modeling and assessment rely on either: probabilistic

approaches (Cimellaro et al., 2010; Miles and Chang,

2006; Bruneau et al., 2003), indexing (Carre

no et

al., 2013; Cardona, 2001) and/or qualitative ratings

(Astles et al.,2009; Hobday 2011). Even practical,

these approaches do leave information aside, either

by not considering system’s elements interrelation-

ships or dynamics, or because they fail to handle con-

cepts intimately related with impreciseness and per-

ceptions. In the same way, most of the resilience as-

sessments models leave the temporal dimension of re-

silience unattended. Our purpose in this research is

to achieve reliable recovery time estimations while

a proper supervision and control of resilience indi-

cators’ progress is performed. Because of its sim-

ple conﬁguration, the set of fuzzy controllers might

be very helpful in provide guidelines suitable for

medium and long term strategic adaptation planning

since the ﬁnal ’target’ resilience level can be selected

by the ﬁnal user. From there, the resilience con-

troller is capable to estimate the proper adjustments

that needs to be done to each indicator’s performance

rate in order to achieve such level, while a general es-

timation of required costs is apprised.

376

Cárdenas, J., Nebot, À. and Mugica, F.

A Proposal for Climate Change Resilience Management through Fuzzy Controllers.

DOI: 10.5220/0006031703760382

In Proceedings of the 6th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2016), pages 376-382

ISBN: 978-989-758-199-1

2 MEASURING CLIMATE

IMPACTS THROUGH

INDICATORS

According to Ellis (2014), there is no absolute con-

sensus whether it is better to focus on vulnerability or

resilience when using indicators to describe and ex-

plain climate impacts to communities. Still, Malone

(2009) stated that although most of climate change

research followed in its origins a more vulnerabil-

ity focused approach, there is an ongoing slow shift-

ing towards an adaptive resilience-based approach be-

cause the concept of resilience, often considered as

an ability can be embraced as an integral part of gen-

eral and particular development goals, while vulnera-

bility alone is often assumed as a degree. Certainly,

both concepts are related and someone may think that

one establishes the other and vice versa, nevertheless

theirs is not a causal relationship since a resilient com-

munity is still vulnerable and in any case, a state of

minimum vulnerability would be a complex outcome

that goes beyond of the resilience goal only.

Research on resilience indicators linked to climate

change is limited, and different approaches may be

used for its development. One of the most recurrent is

the one that consider resilience as the opposite to vul-

nerability, which is deﬁned in the context of climate

change as a combination of exposure, sensitivity and

adaptive capacity (IPCC, 2005). In this sense, vulner-

ability will be reduced through improving resilience

capacities.

On the other hand, resilience operability can be

considered as a matter of purely adaptive capacities.

Malone (2009) deﬁnes adaptive capacity as: ”the

ability to design and implement effective adaptation

strategies, or to react to evolving hazards and stresses

so as to reduce the likelihood of the occurrence and/or

the magnitude of harmful outcomes resulting from

climate-related hazards.”

By adopting this view it is possible to use the

wide range of research developed to measure adap-

tive capacity to climate change, to develop resilience

indicators, hence turning resilience into a more mea-

surable entity. However, resilience and adaptive ca-

pacity are not simply interchangeable and their re-

lationship must be established with caution. Car-

penter et al., (2001) highlighted that relationship be-

tween resilience and adaptive capacity can be estab-

lished in terms of the capacity of the social system

to endure and overcome changes coming from ex-

ternal or internal forces. For Carpenter, the capac-

ity of self-organization and learning capacity build-

ing determines the rate of change that a systems can

undergo and therefore, a way to estimate resilience.

Since we intend to perform a monitoring scheme over

the performance of a set of socio-economic indica-

tors while considering the well functioning of a com-

munity ofﬁcial bodies to estimate the climate im-

pacts that may occur over certain ecosystem services

and socio-economic conditions, in this study we are

choosing a combination of the both aforementioned

resilience frameworks, that is: an adaptive resilience-

based approach to estimate the inﬂuences of coordi-

nate ofﬁcial bodies, and a vulnerability approach that

considers all components assumed to integrate such a

concept in space and time.

3 RESILIENCE INDICATORS

The use of indicators to quantify an intangible con-

cept such as resilience has always caused some con-

troversy. In fact, the common thought now is that re-

silience cannot be measured directly and instead, a

set of proxy variables to perform comparative moni-

toring and assessment over time is needed (Silver et

al., 2008). In doing so, a very extended methodol-

ogy to describe a complex reality in terms of indi-

cators is the so called composite indices, which are

intended to integrate trough aggregation all compo-

nents described by individual indicators into a single

dimensionless number or class. Despite the difference

in approaches there are some common features to all

of the resilience indicators considered in this study.

3.1 Vulnerability Approach Indicators

We intent to use resilience indicators presented by

Moss et al. (2001): Ibarraran et al. (2010) in the

so called Vulnerability- Resilience Indicators Model

(VRIM), which focuses on sensitivity and adaptive

capacity, while exposure is regarded as implicit. The

VRIM model is divided in 8 different sectorial in-

dicators obtained trough the aggregation of different

proxy variables attempted to describe each sector (see

Table 1). In the VRIM model, all proxy variables de-

scribed by indicators are aggregated using geometric

averages, after a process of normalization.

3.2 Adaptive Capacity Approach

Indicators

In order to develop indicators under this category, a

set of components or determinants of resilience spe-

ciﬁc to a particular area or sector is deﬁned ﬁrst. Fo-

cusing on urban climate resilience, Swanson et al.,

(2009) proposed a framework for climate resilience

indicators development considering six components:

A Proposal for Climate Change Resilience Management through Fuzzy Controllers

377

Table 1: Vulnerability-Resilience Indicators Model (VRIM) (Moss et al. 2001, Ibarraran et al. (2010).

Sectorial Indicators Proxy Variables

Food Security Cereals production/ crop land area

Protein consumption/ capita

Water resource sensitivity Renewable supply and inﬂow of water

Settlement / infrastructure sensitivity Population at ﬂood risk from sea level rise

Population without access to clean water

Population without access to sanitation

Human health sensitivity Completed fertility

Life expectancy

Ecosystem sensitivity % Land managed

% Fertilizer use / cropland area

Human and civic resources Dependency ratio

Literacy

Economic capacity GDP(market) / capita

An income equity measure

Environmental capacity % Land unmanaged

/area

Population density

• Economic resources:

• Technology

• Information, skills and management

• Infrastructure

• Institutions and networks

• Equity

We are now in the process of selecting the proper

set of sectors to cover all these dimensions.

4 APPROACH

The proposed strategy is divided in two main stages,

each on one involving different activities. The ﬁrst

part of the strategy comprises the development and

implementation of the Climate Resilience Fuzzy In-

ference model that will be used to estimate resilience

levels, while the second part comprise the construc-

tion and implementation of the Climate Resilience

Fuzzy Controller that will be used to estimate re-

silience recovery times. Figure 1 shows a conceptual

diagram of the main functioning of the proposed mod-

eling scheme.

4.1 Development of the Resilience

Fuzzy Inference Model

Classical inferences are based on tautologies given

by propositional calculus, which aim is the study of

propositions formed by logical connectors. Through

the concept of formal system, propositional logic rep-

resents propositions by means of formulas depicting

natural language that can be used to form rules of in-

ferences. A rule of inference is then a function that

considers premises and, by analyzing their syntax, re-

turns a conclusion. Classical rules of inference can be

generalized for its implementation in the context of

fuzzy logic theory through the use of linguistic vari-

ables expressed as quantitative terms, and composi-

tional rules of inference. Ultimately, these two ele-

ments state the framework that holds the ﬁeld of ap-

proximate reasoning.

A fuzzy expert system tries to emulate the reason-

ing process of a human expert within a speciﬁc do-

main of knowledge; an expert system is then a fuzzy

inference system if the reasoning process is made over

a data set by means of fuzzy rules, which expresses

the actual knowledge about a particular problem.

We are following the fundamental steps required

by an expert fuzzy system to be implemented.

Mainly:

Step 1. Build a knowledge base, which is repre-

MSCCES 2016 - Special Session on Applications of Modeling and Simulation to Climatic Change and Environmental Sciences

378

Figure 1: Conceptualization of the proposed modeling scheme (A) A resilience level is estimated by the Climate Fuzzy

Inference System. (B) Control actions are taken trough the implementation of the Climate Resilience Fuzzy Controller onto

indicators’ performance.

sented by a set of fuzzy rules if-then type interrelat-

ing each selected resilience indicator in order to re-

ﬂect our knowledge about climate resilience forma-

tion and evolution. In this way, we would be capa-

ble to assemble compositional rules of inference over

the very same indicators that are assumed to create a

certain resilience level. Our main sources of informa-

tion would be based on a detailed resilience literature

review and subjective assessments coming from in-

ternational experts in the ﬁeld, which in turn can be

quantiﬁed by means of participatory questionnaires

properly pondered by Delphi or Hierarchical Analytic

Process techniques (Saaty, et al, 1991). By following

this scheme, we want to induce a multidisciplinary

discussion among climate and social science scholars,

in order to improve rule’s structure strength.

Step 2. In order to fuzzify the reported indica-

tor’s values, as well as their possible outcomes after

the inference, a database of fuzzy sets must be deﬁned

in terms of membership functions and a deﬁnition of

meaningful linguistic states for each indicator. Pos-

sible membership functions shapes and their respec-

tive universes of discourses would be consolidated,

as stated in the previous point, by subjective assess-

ments coming from international experts in the ﬁeld.

Our goal is to achieve a reliable membership func-

tions database considering a different range of shapes

to be implemented on the FIS model.

Step 3. In the fuzziﬁcation module, original data

will be converted into fuzzy sets by determining the

match between the input raw indicator data and its de-

ﬁned fuzzy set (membership function). These fuzzi-

ﬁed values will be used to evaluate those rules pre-

sented in the knowledge base.

Step 4. In the model’s inference engine, each

fuzzy rule stored in the knowledge base will be evalu-

ated in order to perform fuzzy inferences. In order to

evaluate each rule, the fuzzy inference engine deter-

mines a ﬁring strength for each rule considering the

degree of match and the fuzzy connectors used over

antecedents. Afterwards, an estimation of the out-

puts is made based on the estimated ﬁre strength and

the match with the deﬁned fuzzy sets for output vari-

ables in the consequent part of each rule. Once rules

have been evaluated, an aggregation process would be

made through a fuzzy aggregator operator.

Step 5. In the defuzziﬁcation module of the

model, the inference result or the aggregate outcome

(a fuzzy set), would be re-converted by means of

a defuzziﬁcation process into a crisp, scalar value;

weather to obtain a ﬁnal estimation or for further pro-

cessing.

4.2 Development of the Resilience

Fuzzy Controller

The general aim of a fuzzy controller is to process

fuzzy and non fuzzy information in a fuzzy scheme

of reasoning in order to determine if the current sys-

tem utility performance is in line with predetermined

standards and if necessary, to take remedial actions

over system’s variables in order to assure that they

A Proposal for Climate Change Resilience Management through Fuzzy Controllers

379

Figure 2: Conceptualization of the resilience fuzzy controller closed-loop operation. Once a resilience level is estimated by

a FIS by means of a non-linear aggregation of a set of n resilience indicators, input control variables are selected (such as

the resilience target level, and the values for constrained variables). Then a subset of fuzzy controllers C

will be activated

depending on the relevance of their related indicator for the considered climate impact and its capacity to be managed through

control actions (As an example, in the ﬁgure the selected indicators are: 1, 2, 3, and the activated fuzzy controllers are:

C1,C2,C3 respectively. The rest of indicators might be considered as constants or to be associated to a rate of change on their

own). These selected controllers will perform control actions over the action rates of their respective resilience indicators in

order to achieve the target resilience level. Before each cycle starts again, an exact appraisal on the investment cost needed to

achieve that particular resilience level will be estimated.

are being used in the most effective way in achieving

planned objectives. The proposed Resilience Con-

troller will use two system’s states: a desired re-

silience level that must be selected by the ﬁnal user,

and an observed/measured resilience state, which will

be estimated in advance by the Climate Resilience

Fuzzy Inference model. The fuzzy controller works

through comparing the desired resilience state to the

estimated resilience state and then adjusting the nec-

essary variables by considering the difference be-

tween the two mentioned resilience levels and an a

speciﬁed control strategy.

The input variables of the fuzzy controller will be

the outputs of the Resilience Fuzzy Inference model,

which are already fuzzy sets. In this scheme we’re

proposing, an important feature comes with those

variables on which control actions are going to be

taken (control-output variables) which are those of the

resilience indicators already deﬁned in the Climate

Resilience Fuzzy Inference model on which control

actions can be managed. Each of these indicators

should have associated a particular ratio value, for

example: number of physicians reported per year,

CO2 production per year, percentage of population

with higher education reported per year, unemploy-

ment rate per year, etc. It is through these temporal

snapshots that the so-called action range pertaining to

each resilience indicator can be established and then,

control actions can be designed. The domains of all

control variables will be divided in fuzzy sets, which

allow their handling in a linguistic manner.

By following this scheme we are ensuring that the

MSCCES 2016 - Special Session on Applications of Modeling and Simulation to Climatic Change and Environmental Sciences

380

controller remains engaged to a qualitative scale rep-

resenting only realistic rates of change that can be

achieved by each indicator per unit of time. More-

over, the same rate can be used as an alternative scale

in terms of monetary value, since all ratio values can

be translated to production costs and therefore, all es-

timations made by the fuzzy controller can be easily

translated into currency investments.

The Climate Resilience Fuzzy Controller will use

fuzzy logic rules to relate input variables to those

changes that would need to be made over resilience

indicators in order to achieve a particular resilience

level. Controller’s rules are in some how different

from the previous mentioned fuzzy inference rules

since the aim now is to design a controlled perfor-

mance over output variables. Therefore the main

structure of a control rule relates conditions to con-

trol actions, in this sense if conditions are in the form

of linguistic variables representing two process state

variables, a control rule implemented by a fuzzy im-

plication, will lead into an action over control vari-

ables.

As we mentioned before, once the changes over

output variables have been established, they will be

used again as inputs for the aforementioned Climate

Resilience Fuzzy Inference model in order to achieve

ﬁnal resilience estimations. Working together, the

Fuzzy Inference model and the Resilience Fuzzy Con-

troller can establish a loop-based framework suitable

to perform resilience estimations while maintaining

control actions over indicators performance. It is im-

portant to mention that although each indicator aims

to describe a particular system’s feature or condi-

tion, not all of these features will be controllable

and in any case, the relevance of each indicator to

inﬂuence a potential resilience level will be related

with the type of climate impact under scope (whether

droughts, ﬂoods, etc.) Therefore the proposed con-

trolling scheme will be based on a careful selection

of which variables/indicators are potentially control-

lable and may inﬂuence as well a possible outcome.

For example, in a drought event, the control actions

would be related with indicators depicting food secu-

rity and water resources, while in the case of a climate

impact in the form of ﬂoods, the settlement / infras-

tructure sensitivity indicator would be more suitable

to be used as a guide to implement control actions.

Once the Fuzzy Resilience model has been com-

pleted, we intent to implement it over at least two

cities by following a scenario-building approach.

Therefore, in order to deﬁne a set of different sce-

narios we will use explicits combinations of values

belonging to constrains variables, along with the po-

tential responses of ofﬁcial bodies in terms of: bud-

get allocation and the political willingness to embrace

and implement adaptation-resilience strategies. Oth-

ers types of variables, not described by the aforemen-

tioned set of indicators but critical to describe partic-

ular social conditions (such as: social development,

risk perceptions, learning capacities, communication

between stakeholders, etc.) can be included in this

stage.

In this way, applying the ﬁnal resilience controller

would consist in ﬁrst deﬁning a particular target re-

silience level and then selecting those input param-

eters aimed to describe a particular user’s condition,

either social or political. The target resilience level

would be completely conﬁgurable and will be de-

pending on those particular goals requested by the ﬁ-

nal user. Since the time lapse to achieve this level de-

pends entirely on real indicator’s rates of changes, and

a direct cost of the experiment can be automatically

obtained, a ﬁnal user could re-conﬁgure if needed,

whether input parameters or the ﬁnal resilience level

in order to be consequent with his own resources and

or particular resilience strategic planning.

The whole scheme of the planned Resilience In-

ference Model, working together with the set of Fuzzy

Controllers can be seen in Figure 2.

5 CONCLUSIONS

In this study, we propose a decision making assistance

tool in the form of fuzzy controllers. This tool will es-

timate the amount of time needed to recover a certain

resilience level related to the actions of climate im-

pacts and global change while considering a coerced

social and political landscape. We propose the use of

climate resilience indicators, vastly reported in liter-

ature. By following a resilience framework aimed to

measure both, vulnerability and adaptive capacities,

our intention is to estimate ﬁrst an initial resilience

level by means of fuzzy composite indices, where in-

dicators aggregation is made assuming a non-linear

interdependence among selected indicators, contrary

to more traditional composite indices methodologies.

This resilience level can then be used as initial input

of a set of fuzzy controllers that implements control

actions over indicators’ performance ratios in order to

achieve a resilience level that can be selected by the ﬁ-

nal user. At the same time that the so called resilience

target level is achieved, an estimation of the required

investment costs to achieve such level is calculated,

considering the rate of production reported for each

indicator.

Resilience fuzzy controller’s conﬁguration op-

tions allows manageable projections considering di-

A Proposal for Climate Change Resilience Management through Fuzzy Controllers

381

verse circumstances that might be very helpful to an-

swer questions such as: if a certain climate resilience

level is required, what are those achievable changes

that needs to be done onto climate resilience indica-

tors? Or: in how long these changes would be re-

ﬂected in a real resilience improvement?

Finally, and since the Climate Resilience fuzzy

controller comprise just an element of the integral risk

to global change, an immediate future extension of

our current work would involve the development of a

full climate risk fuzzy controller.

REFERENCES

Astles K. L., Gibbs P. J., Steffe A. S., Green M. (2009). A

qualitative risk-based assessment of impacts on ma-

rine habitats and harvested species for a data deﬁcient

wild capture ﬁshery. Biological Conservation ,142,

2759-2773.

Bruneau, M., and others (2003). A framework to quanti-

tatively assess and enhance the seismic resilience of

communities. Earthquake Spectra, 19(4) 733-752.

Carre

no, M.L., Cardona, O.D., Barbat, A.H., (2012). New

methodology for urban seismic risk assessment from a

holistic perspective Bull Earthquake Eng,10, 47-565.

Cardona, O.D., (2001). Holistic evaluation of the seis-

mic risk using complex dynamic systems (in Spanish).

PhD Thesis Technical University of Catalonia,10, 47-

565.

Carpenter, S., B. Walker, J. M. Anderies and N. Abel

(2001). From Metaphor to Measurement: Resilience

of What to What? Ecosystems,, 4(8), 765-781.

Hobday AJ, Lough J (2011). Projected climate change in

Australian marine and freshwater environments. Mar

FreshwRes, 62, 1000-1014.

IPCC, (2005). IPCC Special Report on Carbon Dioxide

Capture and Storage. Prepared by Working Group III

of the Intergovernmental Panel on Climate Change.

Cambridge University Press, 12(3), 442.

Ibarraran ME, Malone EL, Brenkert AL. (2010). Climate

Change Vulnerability and Resilience: Current Status

and Trends for Mexico Environment, Development

and Sustainability, 12(3) 365.

Jennifer Ellis, (2014). Trust’s Communities Adapting to Cli-

mate Change Initiative and the Columbia Basin Rural

Development Institute, Columbia Basin Trust.

Miles, S.B., and S.E. Chang (2006). Modeling commu-

nity recovery from earthquakes. Earthquake Spectra,

22(2) 439-458.

Mamdani, E.H., Assilian, S., (1975). An experiment in lin-

guistic synthesis with a fuzzy logic controller. Intern.

J. of Man-Machine Studies,7(1), 1-13.

Malone, E.L., (2009). Vulnerability and Resilience

in the Face of Climate Change: Current Re-

search and Needs for Population Information. Pop-

ulation Action International: Washington DC.,

www.populationaction.org.

Moss R.H., Brenkert A.L., Malone E.L., (2001). Vulnera-

bility to Climate Change: A Quantitative Approach.

Paciﬁc Northwest National Laboratory PNNL-SA-

33642.

T.L. Saaty, L.G. Vargas. (1991). Prediction, Projection, and

Forecasting: Applications of the Analytical Hierarchy

Process in Economics, Finance, Politics, Games, and

Sports. Kluwer Academic Publishers, Boston, MA.

MSCCES 2016 - Special Session on Applications of Modeling and Simulation to Climatic Change and Environmental Sciences

382