You are on page 1of 12

Environmental Science and Pollution Research

https://doi.org/10.1007/s11356-021-13089-0

SELECTED CASE STUDIES ON THE ENVIRONMENT OF THE MEDITERRANEAN AND SURROUNDING


REGIONS

Assessing and building climate change resilience of farming systems


in Tunisian semi-arid areas
Jamel Ben Nasr 1 & Hatem Chaar 1,2 & Fadoua Bouchiba 1 & Lokman Zaibet 3

Received: 1 April 2020 / Accepted: 17 February 2021


# The Author(s), under exclusive licence to Springer-Verlag GmbH, DE part of Springer Nature 2021

Abstract
The agricultural sector plays a strategic role in the Tunisian economy, particularly in rural areas. Resilience and adaptation to
climate change are the main challenges facing this sector. This paper aims to analyze climate change resilience of agricultural
production systems in Tunisian semi-arid areas and to propose options for policy interventions. A path Structural Equation Model
(SEM) was used to predict the resilience of these systems using the partial least squares method (PLS). Results show that farming
systems in Tunisian semi-arid areas remain threatened against negative impact of climate change since 80% of farms in the
sample have shown low resilience levels. The most important determinants of agricultural systems’ resilience are farmers’
income and access to food, adaptive capacity, and access to productive and non-productive assets. Results indicate also that
integrated systems, income diversification, along with cooperation and collective action are the key options to enhance resilience
of rural households and farming systems. It is recommended to raise awareness of stakeholders and decision-makers about
climate change challenges and to develop integrated approaches to better engaging with local stakeholders and institutions in
adaptation programs and strategies development.

Keywords Climate change . Resilience . Adaptive capacity . Farming systems . PLS-SEM

Introduction over comparable time periods” (UNFCCC 2011). CC is ex-


pected to create new opportunities in some parts of the world;
Climate change (CC) refers to any change in climate state over Belyaeva and Bokusheva (2017) suggest that a number of
time, whether due to natural variability or as a result of human regions in Russia could actually benefit from future changes
activity (Roger and Pielke 2004; IPCC 2007; UNDP 2010; in the climate. According to Safonov and Sfonova (2013) the
CCAFS and FAO 2012).The United Nation’s Framework northwestern regions of Russia are expected to experience
Convention on Climate Change defines climate change as “a grain yield increases of 8–9% due to increasing temperatures
change of climate that is attributed directly or indirectly to scenario. However, CC is also expected to cause considerable
human activity that alters the composition of the global atmo- distress particularly in arid regions. In fact, it has substantially
sphere, and that is in addition to natural climate variability increased the frequency of extreme weather events, and it is
considered one of the most serious threats to sustainable de-
Responsible Editor: Philippe Garrigues velopment and to natural ecosystems around the world
(Kalungu et al. 2013; Bombi 2018; Yu et al. 2018).
* Jamel Ben Nasr According to Anwar et al. (2019), at the global level and
jamelnasr@yahoo.fr between 1997 and 2016, more than 524,000 people have died
as a direct consequence of over 11,000 extreme climate events
1
National Agronomic Institute of Tunisia (INAT), University of and economic losses are estimated to US$3.16 trillion.
Carthage, Tunis, Tunisia Climate change not only generates uncertainty and natural
2
National Agricultural Research Institute of Tunisia (INRAT). Field disasters, but also has different effects on daily life in different
Crops Laboratory, University of Carthage, Tunis, Tunisia regions. Farmers consider drought as the most important cli-
3
College of Agriculture and Marine Sciences, Sultan Qaboos mate factor affecting crop production (Kansiime and
University, Muscat, Oman Mastenbroek 2016; Kelvin et al. 2018), and resilience and
Environ Sci Pollut Res

adaptation to CC seems to be the main challenge for agricul- et al. 2003). High levels of vulnerability imply low levels of
ture. Due to its geographical characteristics, (e.g., typhoons resilience and vice versa (Villagran de Leòn 2006; Shah et al.
and earthquakes), Tunisia is one of the Mediterranean coun- 2018). The second approach, based on an inclusive relation-
tries particularly vulnerable to CC, since it has a climate ship, considers resilience as one of the embedded and nested
marked by pronounced seasonality and “unpredictable” prob- components and/or determinants within the broader concept
abilities of the occurrence of extreme events, which can often of vulnerability (McEntire 2001; Turner et al. 2003; Pelling
reach a catastrophic magnitude. The Tunisian climate has a 2003; Hoterova 2020). It is included in adaptive capacity,
high frequency of extreme events such as frequent droughts which represents the system ability to modify or change its
and catastrophic floods (the years 1969, 1973, 2002, 2018) characteristics or behavior in order to better cope with existing
causing enormous damage to the country’s economy and its or anticipated external stress (Adger et al., 2004). According
potential resources (water, soil, forest) (Nefzi 2012; MARHP to Adger (2006), Birkmann (2006) and Folke (2006),
2018). Adaptive capacity forms, exposure, and sensitivity of the sys-
The agricultural sector plays a strategic role in the Tunisian tem to hazard are some of the vulnerability components.
economy, particularly in poor rural areas. Nearly 34% of the Literature about CC vulnerability and resilience points to
population lives in rural areas and is directly or indirectly several factors that can influence vulnerability, adaptive ca-
dependent on agriculture. The sector contributes about 9.6% pacity, and resilience levels of agricultural systems. In this
to GDP (Gross domestic product) (World Bank, 2019). The respect, there is a clear link between agroecosystem resilience
climate projections for Tunisia according to the Had CM3 and social resilience, particularly for social groups or commu-
model predict a general increase in temperatures ranging from nities that are dependent on ecological and environmental re-
+0.8 °C to + 1.3 °C by 2020 and from +1.8 °C to +2.7 °C by sources for their livelihoods (FAO 2007, Kelvin et al. 2018).
2050 (MARHP and GIZ 2007; MARHP 2011; MARHP and Adger (2000) defines social resilience as the ability of groups
GIZ 2012; Touzi and Ben Zakour 2015). For precipitation, or communities to cope with external stresses and distur-
there would be decreases ranging from −5 to −8% in 2020, bances as a result of social, political, and environmental
and from −10 to −30% by 2050. These trends would affect all change. Rapid and uncertain changes in temperature and rain-
seasons. Such climate variability would aggravate droughts fall patterns increase the vulnerability of smallholder farmers
and water stress leading to more degradation of both natural and accentuate rural poverty (AGRA 2014; Kansiime and
ecosystems and agricultural production systems. Mastenbroek 2016; Fahad and Wang 2020). Resilience can
The consequences of CC on Tunisian agricultural produc- reflect both the individual’s ability to cope with a difficult or
tion systems would then be a change in land use, the disap- stressful situation. It depends also on collective action, espe-
pearance of some rainfed crops in arid area, and significant cially on how farming communities well develop their natural
declines in crop yields. There is an expected decrease in the and social capital. These capitals make farmers and their
cereal areas by 20% by 2020, causing a 40% fall in the pro- systems more or less vulnerable to climatic shocks. Altieri
duction of rainfed cereals and a decrease of 52% of olive oil et al. (2015) recognized this link and explained that the fact
production and 36–50% of animal production by 2030 of reducing social vulnerability through the extension and
(CIHEAM 2016).Without adaptation strategies, policies, and consolidation of social networks, both locally and at regional
support measures on a large scale, the effects of climate scales, can contribute to increases agroecosystem resilience.
change on crop production are irreversible. The poorest rural According to NapiWouapi and Sanni (2018), levels of re-
communities will suffer the most severe impacts and suffer silience, diversification, and risk management within the
disproportionately from its negative effects. smallholder production systems depend on agricultural
Resilience is now widely used by the scientific community management and enabling policy environments. Gil et al.
working on CC (Füssel 2007) with no consensus about the (2017) found that integrated agricultural systems were more
definition, sometimes leading to confusion (Gallopin 2006; resilient to climate variability and climate change than special-
Janssen and Ostrom 2006; O’Brien et al. 2005; Lindoso ized agricultural systems. Literature dealing with agro-system
2017) The links between the different concepts of vulnerabil- vulnerability and resilience has concentrated mostly on large-
ity, resilience, adaptation, and adaptive capacity are also the scale global and national level scenarios of climate change and
subject of intense debate (Cutter et al. 2008; Edmonds et al. on the recommended international-level policy options
2020; Yavuz et al. 2020). There are at least two approaches (Osbahr 2008; Kais and Islam 2016). Since CCs will have
that can currently be distinguished (Galderisi et al. 2010). The spatially and socially differentiated impacts experienced pri-
first approach considers resilience as the “opposite face” of marily at local levels, it is crucial to understand the
vulnerability (Folke et al. 2002; Villagran de Leòn 2006; community-level climate change vulnerability and resilience
Proag 2014); a negative connotation has been attributed to (Markkanen and Anger-Kraavi 2019; Thomas et al. 2019).
the vulnerability that is sought to reduce and a positive one This paper aims to assess resilience of agricultural production
credited to the resilience that seeks rather to increase (Klein systems to CC in arid and semi-arid rural areas of
Environ Sci Pollut Res

northwestern Tunisia, where farms are traditional, soils are SEM is a general statistic modeling technique widely used
degraded, technology is poorly developed, and the population in the behavioral sciences (Hox and Bechger 1999; Tambwe
is poor. We use the RIMA approach (resilience index mea- and Mbise 2020). It can be viewed as a combination of factor
surement and analysis) as a systemic approach integrating analysis and regression or path analysis. The interest is on
individual farmer characteristics, formal and informal factors, theoretical constructs represented by the latent variables
internal and external factors, and social capital (FAO 2015) to (LV) or indicators. A LV is a variable that is not observable
analyze resilience and vulnerability of agricultural systems. and cannot be measured directly. Conversely, for a manifest
Structural equation modeling (SEM) has the advantage over variable (MV), a measure can be collected directly. LVs can
traditional multivariate techniques to incorporate simulta- be estimated from overt variables (LV or MV) by isolating
neously several latent variables and construct with a high va- their share of common variance. Structural equation models
lidity level. This approach allows the estimation of latent var- (SEM) consist of a system of equations that can be visualized
iables from observed variables and the explicit assessment of by graphical path diagram, and nodes represent variables in
measurement error with model testing (Novikova et al. 2013). the form of squares for MVs, and as rounds for LVs, arcs
Hence, SEM will be combined to RIMA approach to asses model causality. Each MV is associated with a single LV
and explain resilience of agricultural systems. and the LVs can be related to each other.
SEM is now commonly used by researchers in different fields
because it allows causal links between several variables, includ-
ing moderators and mediators, while incorporating measurement
Materials and methods errors (Fan et al. 2016; Tarka 2018). But it happens that the
prerequisites for its use are not perfectly respected and so is the
The RIMA approach to measuring resilience normality of the data which is a requirement of the covariance
and maximum likelihood analysis method (Bentler and Chou
Resilience index measurement and analysis (RIMA) is a quan- 1987; Malthouse 2001). The violation of normality can lead to
titative approach that provides a conceptual framework for erroneous or aberrant results (Chin et al. 2003). However, there is
measuring and interpreting resilience including the systemic a method of estimating parameters that is less demanding in these
factors that characterize resilience, which provides a complete conditions of use (no normality, small samples accepted), and
and detailed estimate of resilience. RIMA has been technically which, moreover, is more suitable for exploratory and predictive
improved based on its application in many countries (Kenya, research where the data are generally collected by surveys. It is
Sudan, Somalia, South Sudan, Ethiopia, Uganda, Niger, the partial least squares structural equation modeling (PLS-
Burkina Faso, Tanzania, Malawi, Nigeria, Palestine, etc.). SEM), developed, according to Chatelin et al. (2002), Wold
This innovative approach allows to explain why and how (1985), and Lohmöller (1987 and 1989). Considered as a soft
some systems cope with climate shocks and change better modeling, PLS-SEM has the advantage of checking several links
than others do (Alinovi et al. 2010; FAO 2010; FAO 2015). between several explanatory variables and explaining at different
To assess resilience of farms with RIMA approach, we levels. It is based on a partial least squares algorithm and analysis
need a combination of the six latent variables (LVs). The of variance. It has given rise to several computer applications,
structural model (Eq. 1) consists of the set of relationships including LVPLS by Löhmoller (1984), PLS Graph by Chin
between the dependent latent variable (LV), resilience, and (1998) and software developed at the University of Hamburg
explanatory LVs (Alinovi et al. 2010; FAO 2015). called Smart-PLS by Ringle et al. (2015).
The PLS-SEM consists of two parts: In the first part, a
Ri ; t ¼ f ðIFAi ; ABSi ; ACi ; SSNi ; AAi ; Si Þ ð1Þ
measurement model refers to the identification and estimation
In Eq. (1), the resilience (R) of a given system (i) at time (t) of LV from indicators that are observed or MVs. In the second
is the result of the combination of six latent variables part, a structural model refers to the determination of the
(Table 1). causal relationships between the latent variables and which
makes it possible to trace the meaning of the hypotheses com-
posing the research model to be tested. This model can be
Structural equation modeling and partial least represented by a directed graph or “path diagram” consisting
squares approach of a set of LVs, each one can be described by a set of MVs.
Path models are diagrams used to visually display the hypoth-
To assess resilience of farms with RIMA approach, we need a eses and variable relationships that are examined when SEM
combination of the six latent variables (Table 1). Hence, this is applied (Hair et al. 2011; Hair et al. 2017)
study uses structural equation modeling (SEM) to incorporate The PLS approach is based on an iterative algorithm that
all the influencing factors (latent and observed variables) and alternates a construction of LVs by an external estimate with
to analyze the structural relationship between variables. another construction by an internal estimate. These two
Environ Sci Pollut Res

Table 1 Latent variables


according to the FAO-RIMA Latent Definitions
model (FAO 2010). variables

R Resilience is the ability of a system, community or society exposed to the risks to resist, absorb,
accommodate, and correct the effects of a hazard in a timely and effective manner.
IFA Income and access to food provide an approximation of the status and the livelihood security.
ABS Access to basic services, showing the ability of a system to meet their basic needs, as well as
access and effective use of basic services. Examples of indicators: access to schools, drinking
water, electricity and health.
AC Adaptive capacity is the ability of a system to adapt to a new situation and develop new
livelihood strategies.
SSN The social safety net component measures the ability of households to access timely and reliable
assistance provided by international agencies, associations, and NGOs, as well as the help of
relatives and friends.
AA Assets are composed of productive and non-productive assets. Examples of indicators include
land, livestock, and durable goods. Other tangible assets such as home, vehicle, and house-
hold equipment reflect the standard of living and wealth.
S Degree to which a system is affected by a shock and the degree to which the system has been
affected in the recent past.

estimates are repeated until convergence. We start with an approach, the constructs of their resilience model estimated
external estimate of the LVs by linear combination of their with the PLS are formative, since the indicators (= MV) form
MVs. Each LV is estimated using the MVs of its block ac- (and not caused by) the construct (= LV). LVs of PLS model
cording to the following formative scheme: are the six fundamental pillars of resilience in the FAO’s
RIMA conceptual model: income and food access (IFA), ac-
ξ k ¼ ∑ j ωk j  Χ k j þ δ k ð2Þ
cess to basic services (ABS), assets (AA); social safety nets
(SSN), sensitivity (S), and adaptive capacity (AC). The indi-
where (Χkj) is the vector associated with the jth MV of the LV
cators of the resilience latent variables belong to three types
(ξk); (ωkj) is a weight associated with (Χkj), and (δk) is an error
(Table 2): type 1, the categorical indicators where the value
vector of the latent variable.
can be quantified according to a scale; type 2, the binary
In a second step, the LVs are estimated based on the inter-
indicators where the value is 0 or 1; and type 3, continuous
nal model, each LV being estimated using only the other LVs
indicators.
related to it:
ξk ¼ ∑i β ki  ξi þ δk ð3Þ
Study area and data
where (βki) represents the structural coefficient associated
with the relationship between the variables (ξk) and (ξi), and The study uses a survey data, collected from a sample of 60
(δk) is an error term associated with the endogenous LV (ξk). farms during 2018 in the Bargou-Siliana region, in northwest-
Structural equation modeling is recently used to analyze ern Tunisia. Its economy is based mainly on agricultural ac-
climate change vulnerability, risk perceptions and behavioral tivity (cereals, arboriculture, and sheep farming).
intentions, mitigation and adaptation (Dang et al. 2018; Lee The climate of this region is warm and temperate with high
et al. 2019; Lee and Lin. 2019; Lee 2018; Sha et al. 2017). spatial and temporal variability (Fig. 1). The annual average
Results of all these studies conducted in different countries, temperature is 16°C. The maximum temperature can reach
showing that is an appropriate model for this topic. 35°C in July. On average, the rainfall is 477 mm/year.
Structural equation model with PLS approach (PLS-SEM) Owing to the geographical conditions of Siliana, the sensitiv-
was specified in order to build a path model that can predict ity of the environment to natural hazards is high, particularly
the resilience of agricultural production systems in the study in Bargou area (ONAGRI, several years). According to Fig. 2,
area. PLS-SEM was fitted with SmartPLS 3.0 software there was a downward trend in annual precipitation over the
(Ringle et al. 2015). This path model provides empirical mea- study area. The decrease in precipitation, combined to in-
sures of the relationships between the constructs or LVs and crease in temperature would cause drought, which has become
their indicators (measurement models), as well as between longer, more frequent, and more intense in this area. Drought
LVs (structural model). The PLS path model can follow a preoccupies and affects the rural population, its activities, and
reflexive or formative construct. Referring to FAO’s RIMA its wellbeing
Environ Sci Pollut Res

Table 2 Latent variables (LV) or


constructs, their expected effect LV Effect Indicators Code Type Value
on resilience to CC, and their in-
dicators, Code type: 1, qualitative IFA (+) Self-consumption IFA1 3 Average 80.2 TD/month/farmer
ordinal; 2, binary (with yes = 1 or Total expenditure IFA2 3 Average 57.7 TD/month/farmer
no = 0); 3, continuous (value) Income diversity IFA3 3 Average Number = 3
ABS (+) Drinking water access ABS1 2 % of 1 71.6%
Electricity access ABS2 2 % of 1 96%
AC (+) Irrigated areas AC1 3 Average 3.23 hectares
Adaptation strategies AC2 3 Number % of Number = 2
available 1
Use of family labor AC3 2 48%
SSN (+) Member of an organization SNN1 2 % of 1 52%
Cooperation SNN2 2 % of 1 40%
AA (+) Own private land AA1 3 Average 40 hectares
House owner AA2 2 % of 1 90%
Buildings: storage, stable AA3 2 % of 1 35%
Vehicle owner AA4 2 % of 1 31.6%
Tractor owner AA5 2 % of 1 33.3%
Irrigation equipment AA6 2 % of 1 70%
S (−) Conflict with other farmers S1 2 % of 1 65%
Abandonment of some S2 2 % of 1 33%
crops
Yield over the past S3 1 Scale (1–3) 3: decrease in yield = 80%

4040000,000000
420000 ,000000 450000 ,000000 480000 ,000000 510000 ,000000 540000 ,000000 570000 ,000000

7,000000 8,000000 9,000000 10,000000 11,000000 12,000000


38,000000

38,000000
37,000000

37,000000

BOU ARADA
4020000,000000

4020000,000000
EL KRIB EL AROUSSA
36,000000

36,000000

GAAFOUR
35,000000

35,000000
4000000,000000

4000000,000000
BOU ROUIS
34,000000

34,000000

BARGOU
SILIANA NORD
33,000000

33,000000
3980000,000000

3980000,000000
SILIANA SUD
32,000000

32,000000

MAKTHAR
31,000000

31,000000
3960000,000000

3960000,000000

Tunisia KESRA
30,000000

30,000000

Legend
7,000000 8,000000 9,000000 10,000000 11,000000 12,000000
Climate
ER-ROUHIA
3940000,000000

3940000,000000

sub-humide
semi aride sup
semi aride moy
semi aride inf
0 5 10 20 30 40 Siliana governorate
Kilometers aride

420000 ,000000 450000 ,000000 480000 ,000000 510000 ,000000 540000 ,000000 570000 ,000000

Fig. 1 Geographical location and bio-climate of the study area


Environ Sci Pollut Res

Fig. 2 Irregularity and decrease


in annual precipitations in the
study area (ONAGRI n.d., several
years)

Results The blindfolding procedure was executed in order to assess


the predictive relevance of the path model. The Q2 value of the
Structural model resilience endogenous construct (0.236), assessing the
model’s predictive relevance, was above zero. This result pro-
Once the construct measures have been confirmed as being vides clear support for the model’s predictive relevance re-
reliable and valid, the next step addresses the assessment of garding the endogenous latent variables.
the structural model. The first step consists of checking the
structural model for collinearity among the predictor con- Measurement model
structs by examining VIF values of all sets of these constructs.
In the next steps, the model’s predictive capabilities as well as The formative measurement model was first assessed for col-
the relationships between the constructs are evaluated (Hair linearity issues. High levels of collinearity between formative
et al. 2017). The most important evaluation metrics are R2 indicators could be a crucial issue since they have an impact
(explained variance), f 2 (effect size), Q2 (predictive rele- on the estimation of model weights and their statistical signif-
vance), and the size and statistical significance of the structur- icance (Hair et al. 2018). The level of collinearity could be
al path coefficients. assessed by computing the tolerance (TOL) or its reciprocal
The VIF values of all combinations of endogenous con- (VIF = 1/TOL). TOL represents the amount of variance of one
structs and corresponding exogenous constructs were clearly indicator not explained by the other indicators, measuring the
below the threshold of 5 (Table 4). Therefore, collinearity same construct (Hair et al., 2017). VIF values of 5 and higher
among the predictor constructs seems not to be a critical issue respectively show a potential collinearity problem (Hair et al.,
in the structural model. 2011). The two indicators, AA2 and AA4, had a high level of
The coefficient of determination (R2) was equal to 0.652, collinearity (VIF > 5) and were removed from the AA con-
indicating a moderate predictive power accuracy of the endog- struct, whereas the remaining indicators (AA1, AA3, AA5,
enous resilience construct. The R2 adjusted was equal to 0.613. and AA6) still sufficiently capture the construct’s content of
Except for the exogenous construct S, all the other variables latent variable AA and were not redundant. For the other for-
had effect size (ƒ2) values higher than 0.02 indicating signifi- mative model constructs (ABS, IFA, S, and SSN), there were
cant effects on resilience construct. The AC construct had a no problems of multicollinearity since VIF< 2.5 for remaining
large effect (f2 = 0.37). The path coefficients of three constructs, indicators (Table 4).
AA, AC, and IFA were significant at 0.05 level, whereas the The measurement model was then assessed by evaluating
path coefficient of ABS, S, and SSN were not (Table 3). the significance and relevance of the formative indicators. The

Table 3 Model path coefficients


Formative indicators Path coefficients t value p value 95% BCa confidence interval

AA → resilience 0.224 1.918 0.055* [−0.069, 0.398]


ABS → resilience 0.130 1.096 0.273 [−0.087, 0.365]
AC → resilience 0.342 2.121 0.034* [−0.010, 0.584]
IFA → resilience 0.398 2.319 0.021* [0.113, 0.807]
S → resilience −0.054 0.572 0.567 [−0.228, 0.140]
SSN → resilience 0.113 1.101 0.271 [−0.081, 0.330]

*Significant at 5% level
Environ Sci Pollut Res

contribution of a formative indicator, and thereby its rele- households and farming systems in this semi-arid bioclimatic
vance, is evaluated by testing the significance of its outer area remain threatened by negative impacts of CC.
weight by means of bootstrapping procedure (Table 5). Based on estimated resilience (R) values (Fig. 3), three
Outer weights of all formative indicators were significant at groups of production systems can be distinguished
a 5% level, except those of AA1, AA3, AA5, ABS1, AC3, (Table 6): the first group with low resilience value R ≤ 0.4,
IFA1, IFA3, R2, S1, and SSN2 (Table 5). Furthermore, outer the second group of systems with medium resilience (0.4 < R
loadings of AA5, IFA1, IFA3, and S1 were statistically sig- ≤ 0.7), and the third group corresponds to systems with high
nificant, and those of R2 and SSN2 were relatively high and level of resilience to climate change (R > 0.7)
should be retained. The most resilient systems to CC seem to have the highest
Moreover, prior research provides support for the relevance AC, IFA, and AA (Fig. 4). AC corresponds to adaptive capac-
of AA1, AA3, ABS1, and AC3 indicators for capturing some ity which represents system ability to adapt to the changing
parts of AA, ABS, or AC dimensions (according to the RIMA environment in which it operates. This is a multidimensional
approach (FAO 2015)). Thus, these indicators were also concept being determined by complex inter-relationships
retained in the formative constructs, even though their outer among a number of factors at different scales (Vincent,
weights and outer loadings were not significant. 2007). AC is positively correlated with the observed variables
In our model, R1 and R2 are two indicators which help to irrigated areas (AC1), Adaptation strategies available (AC2),
calibrate the estimated value of resilience. In this model, we and use of family labor (AC3). The third group, the most
use two categorical variables: farmer’s perception of income resilient at the CC, contains medium-sized farms (Table 6),
level (R1) and the level of risk perception (R2). which have access to irrigation water. They integrate irrigated
crops or they practice complementary irrigation in the dry
periods. The production system is relatively diversified among
these farmers. This crop diversification is adopted as an adap-
Discussion tation strategy. In fact, unlike group1 and group 2, where
cereal farming accounts for 92%, group 3 presents relatively
The most important variables influencing resilience were IFA diversified production systems with the integration of olive
(Income and food access), AC (Adaptive Capacity), and AA trees, apple trees, peach trees, and vegetable. All of these
(Own Private Land) (Table 3). farms group do not adopt technologies and innovation. Our
Resilience value of these production systems varies be- results are in agreement with Seo (2010), Martin et al. (2016),
tween 0 and 1. Results show that 80% of the farms have and Gil et al. (2017) who conclude that integrated farming
resilience levels below 0.7. This indicates that rural systems are more resilient under global warming than special-
ized farms. The use of family labor can also be considered as a
Table 4 Collinearity test
of measurement model Construct Indicators VIF positive factor of adaptive capacity, especially during peak
periods such as sowing and harvesting periods. Thus, 48.3%
AST AA1 1.286 of surveyed farmers use family labor.
AA3 1.918 The latent variable IFA positively affects the resilience lev-
AA5 1.975 el of agricultural production systems. Since agriculture is the
AA6 1.116 main source of income for 63.3% of farm managers and
ABS ABS1 1.073 households, high level and diversification of income with
ABS2 1.073 safety food access of household enhance farms resilience
AC AC1 1.188 level. This is confirmed by a positive and significant
AC2 1.283 correlation of IFA with the two observed variables IFA2,
AC3 1.120 total expenditure as a proxy of income level, and IFA3
IFA IF3 1.355 which represents the income diversification. This is in
IFA1 2.025 accord with Opiyo et al. (2014) who conclude that households
IFA2 2.428 without livelihood diversification activities are likely to be
S S1 2.179 more vulnerable than households with complementary
S2 1.052 sources of income and less dependents.
S3 2.117 The model result confirmed that productive and non-
SSN SSN1 1.024 productive assets are observed to be significant determinants
SSN2 1.024 of farming system resilience. In fact all indicators and proxy
R1 1.008 (own private land (AA1), house owner (AA2), buildings
R2 1.008 (AA3), vehicle owner (AA4), tractor owner (AA5), and irri-
gation equipment (AA6) are positively correlated with the
Environ Sci Pollut Res

Table 5 Formative constructs


outer weights significance testing Formative Formative Outer t p 95% BCa confidence Outer
results constructs indicators weights value value interval loadings

AA AA1 0.007 0.023 0.982 [−0.554, 0.536] 0.377 **


AA3 0.035 0.156 0.876 [−0.394, 0.499] 0.324 **.
AA5 0.380 1.653 0.099 [−0.003, 0.858] 0.584 ***
AA6 0.833 * 5.983 0.000 [0.585, 1.047] 0.917 *
ABS ABS1 0.111 0.272 0.786 [−0.687, 0.899] 0.363 **
ABS2 0.965 * 2.718 0.007 [0.224, 1.128] 0.994 *
AC AC1 0.490 * 2.561 0.011 [0.109, 0.862] 0.767 *
AC2 0.675 * 3.893 0.000 [0.291, 0.950] 0.889 *
AC3 0.064 0.307 0.759 [−0.386, 0.432] 0.370 **
IFA IFA1 −0.001 0.001 0.999 [−0.874, 1.339] 0.709 ***
IFA2 0.978 * 1.772 0.077 [−0.341, 1.789] 0.999 *
IFA3 0.043 0.201 0.841 [−0.328, 0.530] 0.542 ***
Resilience R1 0.919 * 9.749 0.000 [0.705, 1.004] 0.948 *
R2 0.320 1.574 0.116 [−0.246, 0.570] 0.403 ***
S S1 −0.185 0.391 0.696 [−0.996, 0.843] 0.568 ***
S2 0.647 * 2.055 0.040 [−0.073, 1.011] 0.638 *
S3 0.893 * 1.977 0.048 [−0.083, 1.492] 0.776 *
SSN SSN1 0.847* 3.754 0.000 [0.313, 1.029] 0.911 *
SSN2 0.416 1.058 0.290 [−0.676, 0.946] 0.547 ***

*Significant indicator’s outer weight; ** 0.1<outer loading<0.5 but absolutely important indicator. ***No sig-
nificant indicator’s outer weight but its outer loading is high (above 0.50) and significant

latent variable AA. Assets are so important to the equilibrium Conclusions


of system and rural household when they contribute directly to
the income generation process (productive assets). Since This study confirms that rural areas in semi-arid Tunisia are
farms and household access to productive assets are characterized by overall vulnerability (more than 80%). It
different, the shocks can have different consequences and showed that income and access to food (IFA), adaptive capac-
lead to different system behaviors. This was confirmed by ity (AC), and productive and non-productive assets (AA) are
finding of FAO (2016) that income and assets should be part key determinants of farm’s resilience levels. Irrigation water
of the resilience pillars to climate change. Also by Opiyo et al. access and diversification of farmers’ activities are the main
(2014) who confirmed that own private land combined to adaptation strategies adopted by farmers. However, the lack of
credit access have a positive impact on farming system resil- technological progress and innovations reduce the adaptation
ience level. Tatwangire (2011) established a significant rela- of these farming systems and increase farms vulnerability.
tionship between welfare increasing and access to additional Understanding environmental change, vulnerability, and
livestock holdings and productive farm equipment. resilience of agricultural production systems is necessary for
policy makers and for different stakeholders to develop

Table 6 Production system


category and resilience values Production system Group 1 Group 2 Group 3

Resilience R ≤ 0.4 0.4 < R ≤ 0.7 R >0.7


Production systems (Number) 27 21 12
Percentage 45% 35% 20%
Average resilience 0.21 0.51 0.83
Farm size 1 ha < area <100 ha* >50 ha* 10 ha < area <50 ha*
Area average 17 ha 75 ha 35 ha

*ha hectare
Environ Sci Pollut Res

Fig. 3 Resilience path model estimated with Smart-PLS

adaptation programmers against negative impacts of climate Strategic actions depend on social capital and the capacity of
change. By assessing resilience level of farming systems in rural people to invert opportunistic behavior. We recommend
Tunisians semi-arid areas, and by analyzing determinants of such an approach of working with farmers and engaging with
this resilience, this study aims to help to formulate and define local stakeholders and institutions based on cooperation and
effective strategies for adaptation to climate change and en- collective actions to enhance resilience of rural households
hance adaptive capacity. and farming systems.
Our study points to the need to increase awareness of stake- While this approach and the results provide a starting point,
holders and decision-makers about climate change challenges. particularly for stakeholders and policy-makers to understand
Based on the findings, cooperation and collective actions are vulnerability of agricultural systems and sources of resilience
key strategies to deal with climate change. Indeed, climate at local levels, it is important to note that these results concern
change requires cooperation and collective actions in certain a small scale of analysis, which represents a major limitation
areas; especially in the case of management of groundwater of this study. Since climate change is a global phenomenon,
and other common resources in rural communities. It is time to the scope of this work should be extended to cover the country
look at strategic actions that minimize resource degradation level. Therefore, The resilience index measurement and anal-
including water and soils which are fundamental assets. ysis approach (RIMA) could be extended to the resilience

Fig. 4 Resilience level and


determinants
Environ Sci Pollut Res

analysis and policies approach (RAP with RIMA II) which is Development in Transition Economies (IAMO), Halle (Saale),
http://nbn-resolving.de/urn:nbn:de:gbv:3:2-69113.
more appropriate at the national level (FAO 2016). This ex-
Bentler PM, Chou CP (1987) Practical issues in structural modeling.
tended model should improve the relevance of decision sup- Sociol Methods Res 16(1):78–117. https://doi.org/10.1177/
port tools in the policy process of climate change mitigation 0049124187016001004
and adaptation. Birkmann J (2006) Measuring vulnerability to promote disaster-resilient
societies: conceptual frameworks and definitions. In J. (Ed.)
Measuring vulnerability to natural hazards – towards disaster resil-
ient societies. United Nations University Press. Hong-Kong. New
Author contribution Conceptualization and methodology: Jamel BEN York. 9-54 pp.
NASR., Hatem CHAAR and Fadoua BOUCHIBA. Bombi P (2018) Potential impacts of climate change on Welwitschia
Investigation and data collection were performed by Jamel BEN mirabilis populations in the Namib Desert Southern Africa.
NASR and Fadoua BOUCHIBA Journal of Arid Land 10(5):663–672. https://doi.org/10.1007/
Analysis and visualization were performed by Jamel BEN NASR, s40333-018-0067-1
Hatem CHAAR and Lokman Zaibet. CCAFS and FAO (2012) Training guide - gender and climate change
The first draft of the manuscript was written by Jamel BEN NASR. research in agriculture and food security for rural development.
All the authors commented on previous versions of the manuscript www.fao.org/climatechange/micca/gender/en
especially Hatem CHAAR and Lokman Zaibet. Chatelin YM, Vinzi VE, Tenenhaus M (2002) State-of-art on PLS model-
All the authors contributed to review and editing according to re- ling through the available software. HEC Business School. Jouy-en-
viewer’s comments. Josas
All the authors read and approved the final manuscript. Chin WW (1998) The partial least squares approach to structural equation
modeling. In: Marcoulides GA (ed) Modern methods for business
Data availability The datasets used and/or analyzed during the current research. Lawrence Erlbaum Associates, Mahwah, pp 295–358
study are available from the corresponding author on reasonable request. Chin WW, Marcolin BL, Newsted PR (2003) A partial least squares
latent variable modeling approach for measuring interaction effects:
results from a Monte Carlo simulation study and an electronic-mail
Declarations emotion/adoption study. Inf Syst Res 14(2):189–218
CIHEAM (2016) Avec le changement climatique. quel avenir de
Ethics approval and consent to participate Not applicable l’agriculture en Tunisie?. Centre International de Hautes Etudes
Agronomiques Méditerranéennes. Lettre n°37. Septembre 2016.
Consent for publication Not applicable. 1.4 pp.
Cutter SL, Barnes L, Berry M, Burton C, Evans E, Tate E, Webb J (2008)
Competing interests The authors declare that they have no competing A place-based model for understanding community resilience to
interests. natural disasters. Glob Environ Chang 18(2008):598–606
Dang HL, Li E, Nuberg I, Bruwer J (2018) Vulnerability to climate
change and the variations in factors affecting farmers’ adaptation:
a multi-group structural equation modelling study. Clim Dev 10(6):
509–519. https://doi.org/10.1080/17565529.2017.1304885
References Edmonds HK, Lovell JE, Lovell CAK (2020) A new composite climate
change vulnerability index. Ecol Indic 117:106529. https://doi.org/
Adger WN (2000) Social and ecological resilience: are they related? Prog 10.1016/j.ecolind.2020.106529
H u m G eo g r 2 4 ( 3 ): 3 4 7 – 3 6 4 . h tt p s : / / d o i . o r g / 1 0 . 1 1 9 1 / Fahad S, Wang J (2020) Climate change, vulnerability, and its impacts in
030913200701540465 rural Pakistan: a review. Environ Sci Pollut Res 27:1334–1338.
Adger WN (2006) Vulnerability. Glob Environ Chang 16:268–281. https://doi.org/10.1007/s11356-019-06878-1
https://doi.org/10.1016/j.gloenvcha.2006.02.006 Fan Y, Chen J, Shirkey G, John R, Wu SR, Park H, Shao C (2016)
Applications of structural equation modeling (SEM) in ecological
Adger WN, Brooks N, Bentham G, Agnew M and Eriksen S (2004) New
studies: an updated review. Ecol Process 5:19. https://doi.org/10.
indicators of vulnerability and adaptive capacity, Tyndall Centre for
1186/s13717-016-0063-3
Climate Change Research. Technical Report 7, Norwich
FAO (2007) Adaptation to climate change in agriculture. forestry and
AGRA (2014) Africa agriculture status report: climate change and small fisheries: perspective. framework and priorities. Inter-Departmental
holder agriculture in sub-Saharan Africa. Alliance for a Green Working Group On Climate Change. Food and Agriculture
Revolution in Africa, Nairobi, Kenya. (AGRA) Organization of the United Nations (Rome. 2007)
Alinovi L, Mane E, Romano D (2010) Measuring household resilience to FAO (2010) Measuring resilience: a concept note on the resilience tool.
food insecurity: Application to Palestinian Households. In: Available at http://www.fao.org/3/al920e/al920e00.pdf
Benedetti R, Bee M, Espa G, Piersimoni F (eds) Agricultural survey FAO (2015) Analyzing resilience for better targeting and action.
methods. Wiley, Chichester. https://doi.org/10.1002/ Resilience analysis in Senegal 2005. By Resilience Analysis and
9780470665480.ch21 Policies team. Agricultural Development Economics Division
Altieri MA, Nicholls CI, Henao A, Lana MA (2015) Agroecology and the Food and Agriculture Organization of the United Nations. available
design of climate change-resilient farming systems. Agron Sustain in http://www.fao.org/3/a-i4456e.pdf
Dev 35(3):869–890. https://doi.org/10.1007/s13593-015-0285-2 FAO (2016) Analysing resilience for better targeting and action.
Anwar MA, Zhou R, Sajjad A, Asmi F (2019) Climate change commu- Resilience Index Measurement and Analysis II RIMA-II.
nication as political agenda and voters’ behavior. Environ Sci Pollut Resilience Analysis and Policies team. Agricultural Development
Res 26:29946–29961. https://doi.org/10.1007/s11356-019-06134- Economics Division Food and Agriculture Organization of the
Belyaeva M and Bokusheva R (2017) Will climate change benefit or hurt United Nations. available in http://www.fao.org/3/a-i5665e.pdf
Russian grain production? A statistical evidence from a panel ap- Folke C (2006) Resilience: the emergence of a perspective for social-
proach, Discussion Paper, No. 161, Leibniz Institute of Agricultural ecological systems analyses. Glob Environ Chang 16:253–267
Environ Sci Pollut Res

Folke C, Carpenter SR, Elmqvist T et al (2002) Resilience and sustainable Lee Y, Tung C, Lin S (2019) Attitudes to climate change, perceptions of
development: building adaptive capacity in a world of transforma- disaster risk, and mitigation and adaptation behavior in Yunlin
tion. Ambio. 31(5):437–440 County, Taiwan. Environ Sci Pollut Res 26:30603–30613. https://
Füssel HM (2007) Vulnerability: a generally applicable conceptual doi.org/10.1007/s11356-018-1358-y
framework for climate change research. Glob Environ Chang Lindoso D P (2017) Vulnerability and resilience: potentials, conver-
17(2):155–167 gences and limitations in interdisciplinary research. Ambient. soc.
Galderisi A, Ferrara F, Ceudech A (2010) Resilience and/or vulnerabili- [online]. 2017, vol.20, n.4 [cited 2020-12-01], pp.127-144.
ty? Relationships and roles in risk mitigation strategies. In Ache P. Available from: <http://www.scielo.br/scielo.php?script=sci_
Ilmonen M. Space Is Luxury. Selected Proceedings 24th Annual arttext&pid=S1414-753X2017000400127&lng=en&nrm=iso>.
AESOP Conference. ISSN 1809-4422. https://doi.org/10.1590/1809-
Gallopin GC (2006) Linkages between vulnerability. resilience. and 4422asoc0248r1v2042017.
adaptive capacity. Glob Environ Chang 16(3):293–303 Lohmöller J B (1984) LVPLS Program Manual, Version 1.6,
Gil JDB, Cohn AS, Duncan J et al. (2017) The resilience of integrated Zentralarchiv für Empirische Sozialforschung, Universität zu
agricultural systems to climate change. Wiley Periodicals. Inc. Köln, Köln
Volume 8. July/August 2017. WIREs Clim Change 2017. 8:e461. Lohmöller J B (1987) PLS-PC. Latent Variables Path Analysis with
https://doi.org/10.1002/wcc.461. Partial Least Squares. Version 1.8 for PCs under MS-Dos.
Hair JF, Ringle CM, Sarstedt M (2011) PLS-SEM: indeed a silver bullet. Lohmöller JB (1989) Latent variable path modeling with partial least
J Mark Theory Pract 19(2):139–151 squares. Heildelberg: Physica-Verlag.
Hair JF, Hollingsworth CL, Randolph AB, Chong AYL (2017) An up- Malthouse EC (2001) Checking assumptions of normality before
dated and expanded assessment of PLS-SEM in information sys- conducting factor analyses. J Consum Psychol 10:81
tems research. Ind Manag Data Syst 117(3):442–458 MARHP (2011) Atelier régional forets. parcours et changement
Hair JF, Sarstedt M, Ringle CM, Gudergan SP (2018). Advanced issues climatique dans la région du moyen orient”. Rapport national:
in partial least squares structural equation modeling (PLS-SEM). Tunisie. Ministère de l’Agriculture et de l’environnement :
Thousand Oaks: Sage Direction Générale des Forêts-Tunisie. 22 septembre 2011.
MARHP (2018) Ministère de l’agriculture des ressources hydrauliques et
Hoterova K (2020) Comparative analysis of the resilience and vulnera-
de la pêche. Les statistiques de l’Onagri. Tunisia.
bility of the railway infrastructure. MEST Journal 8(2):100–106.
MARHP and GIZ (2007) Stratégie nationale d’adaptation de l’agriculture
https://doi.org/10.12709/mest.08.08.02.10
tunisienne et des écosystèmes aux changements climatique. Tunisia
Hox J and Bechger T (1999) An introduction to structural equation
MARHP and GIZ (2012) Stratégie Nationale sur le Changement
modeling. Family Science Review 11:354–373. http://joophox.net/
Climatique SNCC : Rapport de la stratégie. Rapport final. Octobre
publist/semfamre.pdf
2012.
IPCC (2007) Climate change 2007: synthesis report. Contribution of
Markkanen S, Anger-Kraavi A (2019) Social impacts of climate change
Working Groups I. II and III to the Fourth Assessment Report of
mitigation policies and their implications for inequality. Clim Pol
the Intergovernmental Panel on Climate Change [Core Writing
19(7):827–844. https://doi.org/10.1080/14693062.2019.1596873
Team. Pachauri. R.K and Reisinger. A. (eds.)]. IPCC. Geneva.
Martin G, Moraine M, Ryschawy J, et al. (2016) Crop–livestock integra-
Switzerland. 104 pp
tion beyond the farm level: a review. Agronomy for Sustainable
Janssen MA, Ostrom E (2006) Resilience vulnerability and adaptation: a Development, Springer Verlag/EDP Sciences/INRA, 2016, 36 (3),
cross-cutting theme of the International Human Dimensions pp.53. 53 https://doi.org/10.1007/s13593-016-0390-x
Programme on Global Environmental Change. Glob Environ McEntire DA (2001) Triggering agents vulnerabilities and disaster reduc-
Chang 16(3):237–239 tion: towards a holistic paradigm. Disaster Prev Manag Vol.
Kais SM, Islam MS (2016) Factors influencing farmer’s choice of crop 10(Issue: 3):189–196
production response strategies to climate change and variability in NapiWouapi HA, Sanni M (2018) Enhancing resilience of livelihoods
Narok East Sub Community capitals as community resilience to and production systems to climate variability and other related risks
climate change: conceptual connections. Int J Environ Res Public in Africa. In: Leal FW (ed) Handbook of Climate Change
Health 13(12):E1211 Resilience. Springer, Cham
Kalungu W, Filho WL, Harris D (2013) Smallholder farmers perception Nefzi A (2012). Evaluation économique de l’impact du changement
of the impacts of climate change and variability on rain-fed agricul- climatique sur l’agriculture : étude théorique et application au cas
tural practices in semi-arid and sub-humid regions of Kenya. J de la Tunisie. Thèse de doctorat en Economie de l’environnement et
Environ Earth Sci 3(7):129–141 des ressources naturelles. Soutenue le 29-03-2012 à Paris.
Kansiime MK, Mastenbroek A (2016) Enhancing resilience of farmer AgroParisTech . Dans le cadre de Ecole Doctorale Agriculture.
seed system to climate-induced stresses: Insights from a case study Alimentation. Biologie. Environnement. Santé (2000-2015 ; Paris)
in West Nile region. Uganda. J Rural Stud Volume 47:–230. https:// Novikova SI, Richman DM, Supekar K, Barnard-Brak L, Hall D (2013)
doi.org/10.1016/j.jrurstud.2016.08.004 Chapter three - NDAR: a model federal system for secondary anal-
Kelvin J, Kinuthia S, Inoti K et al (2018) Factors influencing farmer’s ysis in developmental disabilities research. In: Urbano RC (ed)
choice of crop production response strategies to climate change and International Review of Research in Developmental Disabilities,
variability in Narok East Sub-county. Kenya. J Nat Resour Dev 08: vol Vol. 45. Academic Press, Cambridge, pp 123–153. https://doi.
69–77 org/10.1016/B978-0-12-407760-7.00003-7
Klein RJT, Nicholls RJ, Thomalla F (2003) Resilience to natural hazards: O’Brien K L, Eriksen S, Schjolden A and Lygaard L (2005) What’s in a
how useful is the concept? Environ Hazards 5:35–45 word? Interpretations of vulnerability in climate change research.
Lee Y-J (2018) Relationships among environmental attitudes, risk per- Climate Policy.
ceptions, and coping behavior: a case study of four environmentally ONAGRI, (n.d.)several years. Observatoire National de l’Agriculture.
sensitive townships in Yunlin county, Taiwan. Sustainability Statistical directory « Annuaire Statistique ». Available in http://
2018(10):2663 www.onagri.nat.tn/statistiques.
Lee Y, Lin S (2019) Vulnerability and ecological footprint: a comparison Opiyo FE, Wasonga OV and Nyangito MM (2014) Measuring household
between urban Taipei and rural Yunlin Taiwan. Environ Sci Pollut vulnerability to climate-induced stresses in pastoral rangelands of
Res 27:34624–34637. https://doi.org/10.1007/s11356-019-05251-6 Kenya: Implications for resilience programming. Pastoralism
Environ Sci Pollut Res

Springer Berlin Heidelberg. https://doi.org/10.1186/s13570-014- Interdiscip Rev. Clim Chang 10(2):e565. https://doi.org/10.1002/
0010-9 wcc.565
Osbahr H (2008) Building resilience: adaptation mechanisms and Touzi S and Ben Zakour M (2015) Expérience tunisienne pour faire face à
mainstreaming for the Poor. UNDP; New York. NY. USA: 2008. la Variabilité et au Changement Climatique en Zones Côtières.
Background Paper for UNDP Human Development Report. Rapport national dans le cadre du projet ClimVar (Med
Pelling M (2003) The vulnerability of cities: natural disasters and social Partnership). Available in https://planbleu.org/sites/default/files/
resilience. Earthscan. London. UK. publications/rapport_national_tunisie_october_2015.pdf
Proag V (2014) The concept of vulnerability and resilience. Procedia Turner BL, Kasperson RE, Matson PA et al (2003) A framework for
Econ Finance 18:369–337 vulnerability analysis in sustainability science. Proc Natl Acad Sci
Ringle CM, Wende S and Becker J-M (2015) SmartPLS 3. 100(14):8074–8079
Boenningstedt: SmartPLS GmbH, http://www.smartpls.com. UNDP (2010) Gender. Climate Change and Community-Based
Roger A and Pielke J (2004) What is climate change?. Issues in Science Adaptation. New York. New York. United States. (Also available
and Technology. VOL. XX. NO. 4. at: www.beta.undp.org/undp/en/home/librarypage/environment-
Safonov G and Sfonova Y (2013) Economic analysis of the impact of energy/climate_change/gender/gender-climate-change-and-
climate change on agriculture in Russia, National and Regional community-based-adaptation-guidebook-.html
Aspects. Oxfam Research Reports. Available at https://www. UNFCCC (2011) Report of the global environment facility to the confer-
oxfam.org/en/research/economic-analysis-impact-climate-change- ence of the parties. Framework Convention on Climate Change.
agriculture-russia. Also available at: https://unfccc.int/resource/docs/2011/cop17/eng/
Seo SN (2010) Is an integrated farm more resilient against climate 07.pdf
change? A micro-econometric analysis of portfolio diversification
Villagràn de Leòn J.C. (2006) Vulnerability: a methodological and con-
in African agriculture. Food Policy Vol.35(1):32–40. https://doi.org/
ceptual review. Source, n° 4, UNU-EHS.
10.1016/j.foodpol.2009.06.004
Sha Z, Xie Y, Tan X, Bai Y, Li J, Liu X (2017) Assessing the impacts of Vincent K (2007) Uncertainty in adaptive capacity and the importance of
human activities and climate variations on grassland productivity by scale. Glob Environ Chang 17(1):12–24. https://doi.org/10.1016/j.
partial least squares structural equation modeling (PLS-SEM). J gloenvcha.2006.11.009
Arid Land 9:473–488. https://doi.org/10.1007/s40333-017-0022-6 Wold H (1985) Partial least squares. In: Kotz S, Johnson NL (eds)
Shah AA, Ye J, Abid M, Khan J, Amir SM (2018) Flood hazards: house- Encyclopedia of Statistical Sciences, vol 6. Wiley, New York, pp
hold vulnerability and resilience in disaster-prone districts of Khyber 581–591
Pakhtunkhwa province, Pakistan. Nat Hazards 93(1):147–165 World Bank (2019) https://donnees.banquemondiale.org/indicateur/NV.
Tambwe M, Mbise ERC (2020) Determinants of farmers’ entrepreneurial AGR.TOTL.ZS?locations=TN. The reference was last accessed on
intention in Tanzanian agricultural sector: a structural modelling 06/03/2019
approach. Adv Soc Sci Res J 7:376–396. https://doi.org/10.14738/ Yavuz C, Kentel E, Aral MM (2020) Tsunami risk assessment: economic,
assrj.710.9256 environmental and social dimensions. Nat Hazards 104:1413–1442.
Tarka P (2018) An overview of structural equation modeling: its begin- https://doi.org/10.1007/s11069-020-04226-y
nings, historical development, usefulness and controversies in the Yu Y, Pi Y, Yu X, Ta Z, Sun L, Disse M, Zeng F, Li Y, Chen X, Yu R
social sciences. Qual Quant 52:313–354. https://doi.org/10.1007/ (2018) Climate change. Water resources and sustainable develop-
s11135-017-0469-8 ment in the arid and semi-arid lands of Central Asia in the past 30
Tatwangire A (2011) Access to productive assets and impact on house- years. J Arid Land 11(1):1–14. https://doi.org/10.1007/s40333-018-
hold welfare in rural uganda. Thesis number 2011: 27 ISBN: 978- 0073-3
82-575-0933-0. Norwegian University of Life Sciences, Ås.
Available at: http://hdl.handle.net/11250/2434107 Publisher’s note Springer Nature remains neutral with regard to jurisdic-
Thomas K, Hardy RD, Lazrus H et al (2019) Explaining differential tional claims in published maps and institutional affiliations.
vulnerability to climate change: a social science review. Wiley

You might also like