Urban resilience has been defined by a plethora of scholars, several times and in many different ways. Abdrabo and Hassaan defined urban resilience as “the ability of an urban system in all its dynamism to support, in the face of a hazard or pressure, the provision and accessibility to services and functions essential for the wellbeing of all residents, especially those lacking the means to buffer stresses” [11]. Batica and Gourbesville defined urban resilience as the ability of a system to function during and after a shock [1]. While, the United Nations University and Centre for Policy Research [27] defined urban resilience as “the ability to activate protective qualities and processes at individual, community, institutional and systems level to engage with hazards and stresses and cooperate with each other in order to maintain or recover functionality and prosper while adapting to a new equilibrium and minimizing the accumulation of pre-existing or additional risks and vulnerabilities”. Furthermore, The Rockefellers Foundation/Arup [28] defined urban resilience as the ability of an urban area to continue functioning such that its inhabitants more so the poor and vulnerable can be able to survive and thrive in the midst of shocks and stresses. This study will define urban resilience as the capacity of an urban system and all its constituent socio-ecological networks to absorb climate change shocks and stresses through preparedness and the capacity to recover and innovate when faced with climate change shocks and stresses [1] [10] [12].

Studies have developed several assessment indices that try to refine urban resilience into generally related dimensions such as the physical, natural, social, economic and institutional [1] [11] [29]. This study conducted a review of 17 indices that have been developed and used to measure urban resilience to climate change shocks and stresses (Table 1). These were grouped into indices that measure urban resilience to multiple climate change shocks and stresses, flood resilience indices and heat or drought resilience indices. The indices were reviewed to synthesize the dimensions used, the qualities observed by variables to monitor resilience, the scale of the urban area and the methods of data collection used.

A synthesis of dimensions used by indices (Table 1) revealed that the social dimension was used by 10 indices, economic dimension (8), physical (7), institutional (7), ecological (7), adaptive capacity (5), governance (2), emergency preparedness (1) and exposure (1). Only 7 indices used dimensions that closely conform to those proposed by UNISDR [31], Sendai framework for action 2015-2030 and IPCC [32]. This implies that most indices deviate from the generally and globally accepted dimensions of resilience. This could be because they were developed for specific purposes, context and field of study. Even among resilience indices developed to measure a particular climate change shock like floods, there exists variance in the dimensions and variables considered. This makes it difficult for researchers in the same discipline to formulate appropriate research questions and resolve resilience baselines [33].

A synthesis of qualities commonly used to observe resilience revealed that 7 out of the 17 indices bring out clearly the qualities of resilience observed (Table 1). The rest of the indices seem not to bring out clearly these qualities which are indeed important in monitoring a systems resilience [3] [6]. The most observed qualities were redundancy, inclusiveness, robustness, flexibility, resourcefulness and integrated. Redundancy is having spare capacity to accommodate disruptions, inclusive is broad consultation and engagement of stakeholders and robustness is having well-constructed and well managed physical assets that can withstand shocks and stresses [28]. Flexibility is the ability to change, evolve and adapt in response to a disturbance, resourcefulness is the ability to find different ways of achieving goals while integrated is exchange of information between systems supportive to common outcomes [12] [28] [29].

Analysis of the scale of the urban areas for which the resilience indices were developed and applied (Table 1) revealed that all indices were developed to assess resilience at city and metropolitan levels and none for municipalities and townships. In addition, there was limited capturing of data at household level. However, Joerin [10] who captured data at zonal level indicated that they could have got better results if the assessment was conducted at household level.

Analysis of the methods of data collection (Table 1) showed that out of the 17 indices, 9 used objective or secondary methods of data collection, 3 used subjective (primary) methods while, 5 used both objective and subjective methods. Subjective based methods are important because a person’s resilience does not only constitute his tangible objective livelihood assets but also psychological, such as social networks, sense of place, beliefs and cultural identity [22] [34]. They also help in identifying improvements in resilience over time using consistent assessors, measuring the marginal effect of a resilience enhancement strategy and promoting the use of local knowledge in climate change risks assessment [33] [35]. On the other hand, using objective based measurements is faster due to their use of robust socio-economic and bio-physical secondary data. They reduce inter-assessor bias and offer high level of generalization [28]. However, objective data sets are often lacking in most developing countries making subjective methods suitable in such localities. Both subjective and objective based methods have their own strengths and weaknesses [28].

Out of the 17 indices synthesized, 8 indices that have dimensions and variables related to the physical, social, economic and institutional factors were selected (Table 2) [36] [37]. The indices selected include: The CDRI [10], IRI [11], SI [12], FDRI [14], FRI [1], CHRIL [29], SDAMSR [9] and CRI [23]. The variables obtained from the indices selected were used to validate the variables that were obtained from community perceptions on the most important components in resilience building in Mbale municipality in the previous study [38]. These formed the basis for proposing a context specific municipality resilience index that was used to quantify resilience against climate change shocks and stresses in Mbale municipality in Eastern Uganda with limited robust bio-physical and socio-economic data.

Summarily, most urban resilience indices either focus on the physical assets viewpoint while disregarding the intangible assets such as the social capital and safety nets [22], while others focus on adaptive capacity which views an urban area as a system encapsulating the physical, social, economic and institutional dimensions [11]. This study views an urban area as a complex system comprising of the social, economic, physical and institutional dimensions and thus, the complex adaptive systems theory was used to anchor this piece of work to make it possible to estimate resilience of Mbale municipality [39] [40]. Further still, household resilience capacities are process-oriented and context specific [22] [34]. Therefore attempts to measure resilience of households need to recognise people’s intangible assets and their perceptions of resilience in the face of shocks and stresses while recognizing the existing indices [41]. This can offer deeper explanations and thus solutions than currently possible.

The study was conducted in Mbale municipality in Eastern Uganda from July to August 2017 (Figure 1). Mbale municipality comprises of three divisions i.e. Wanale, Industrial and Northern divisions constituted by 14 wards/parishes with a total population of 92,857 [42]. The municipality was selected based on its location in the mountainous disaster prone area characterised with recurring flash floods, prolonged droughts and high rate of urban population growth and density which impose pressure on land and social service provision [42] [43].

The municipality experiences a humid tropical climate with bimodal rainfall from March to June and from September to November, while the dry seasons spans between December to February and in July [44]. The average total annual rainfall was 1500 mm and the average annual temperature was 23˚C for the period 1960-2009 [44]. Climate change is expected to increase the frequency and intensity of extreme events such as prolonged droughts, flash floods and heat waves in the municipality [44].

The study adopted the methodology of building a composite index put forward by Nardo and his colleagues [36] and OECD [37]. The methodology has severally been used [11] [12] [13] [23]. This methodology involves selection of an appropriate spatial scale, appropriate dimensions, indicators and variables to be measured.

The scale of a municipality was chosen since it is an administrative unit with normative competencies [13]. To come up with appropriate dimensions, indicators and variables that could measure urban resilience to climate related shocks and stresses in the context of Mbale municipality, the study draws from the previous study that used a participatory approach to come up with priority variables perceived to be important in enhancing community resilience in Mbale municipality [38] (Table 2). The expert views obtained from the reviewed indices helped to validate the variables that had been obtained from community perceptions. The selection of variables was based on accessibility, relevance and most repeated variables, while getting rid of those that are synonyms to obviate repetition [13] [37]. The variables were also selected basing on the following qualities of resilience that they measure: redundancy, inclusive, robustness and integrated [6] [12] [15] [28] as explained in this study sub section 1.2. The variables were then categorized into indicators and dimensions (Table 2).

The MRI constitutes 46 variables, 14 indicators and four dimensions (physical, social, economic and institutional) (Table 2). The physical dimension is selected because redundant, robust, flexible and inclusive physical infrastructure such as water, electricity, sanitation and transportation facilitates the well-functioning of an urban area, enhances people’s ability to respond, absorb and recover easily when faced with climate change shocks and stresses [28]. The social dimension is selected because 1) robust and inclusive education sector can promote human development and enables people to easily understand early warnings which creates awareness about potential shocks and stresses. 2) Robust and inclusive health sector can support citizens to recover during and after shocks and stresses, while social safety networks among urban residents shows peoples connectedness to manage disturbances [10]. The economic dimension shows the ability of people to earn income to be able to meet the basic needs of life. It also shows the ability to save which provides the spare capacity that can be used during shocks and stresses [10]. The institutional dimension looks at the local and national government functioning to manage shocks and stresses. It constitutes good governance, population awareness and crisis management. Transparent, inclusive and integrated government planning helps to capture the diverse needs of the urban area for appropriate strategic planning. It also enables protection of important ecosystems which are essential in enabling the urban area to thrive during climate change chocks and stresses [28].

The MRI seeks to accommodate the context of Mbale municipality in Uganda where limited secondary data exists at household and municipal level. To reduce the challenge, 389 structured household interviews were conducted to get responses to the 46 variables (Table 3) in Industrial, Northern and Wanale divisions of Mbale municipality. The unit of analysis was a household. Household survey was conducted based on a multi-stage sampling procedure. This was due to its high level of flexibility, cost effectiveness, accommodating more than two sampling stages and sampling methods required [45]. The first stage involved a

Y = yes, N = no.

purposive selection of seven wards out of 14 wards in the three divisions. These wards were selected taking into consideration sensitivity and fragility of the wards to climate change shocks and stresses and population characteristics [44] [42]. The second stage involved random sampling of households while using a structured household interview questionnaire. Random sampling was deemed appropriate at this stage because every household in the selected wards are likely to be affected by climate change shocks and stresses, thus assumed to be uniform and selecting any household would be representative of the total population. The study sample size was statistically determined using Yamane’s procedure [46] at a 5% level of precision.

Furthermore, the UBOS census figures of the seven sampled wards were used to proportionately distribute the 389 households in order to derive the exact figures to be sampled in each ward or parish (Figure 1). Consequently, in Malukhu ward 46 households were sampled, in Namatala ward 129, in Nabuyonga ward 50, in Namakwekwe ward 79, in Nkoma ward 58, in Boma ward 11 and in Busamaga west 19 households. Closed ended questions were used because they provide an expeditious instrument in data collection, coding, interpretation and quantification [22]. The data collection tool was pre-tested on 15 households after obtaining their consent. The data collected constituted of households’ perceptions, knowledge and opinions regarding the availability and access to resilience variables. Two scales were used i.e., a binary scale where variables offer a choice between “0” (not available) and “1” (available) and a 5 point Likert scale that offers choices between “1” (strongly disagree), “2” (disagree), “3” (undecided), “4” (agree) and “5” (strongly agree).

To compute the resilience index, the study draws from the method used by Freudenberg [47] and Nardo and his colleagues [36]. This method has been used by climate disaster resilience studies [14] [48]. First, it involves normalizing the variables since different measurement units and scales were used. Normalization enables variables to have a common basis and avoids the problems in mixing measurement units [47] [36]. Several normalization methods exist [36]. This study uses a rescaling “Min-max” normalization method that adjusts observed variables to take a value of 0 to 1 [36]. This normalization method has severally been used [14] [12] (Equations (1)).

Normalized Value = ( E i − E min ) / ( E max − E min ) (1)

where: E_i is the observed variable, E_min is the minimum value of the observed variable, E_max is the maximum value of the observed variable.

Second, Principal Component Analysis (PCA) was used to assign weights to the normalized variables. This helps to reflect the contribution of each variable to the overall composite [48] [36]. In order to generate the resilience index, all variables are considered to have the same weights after normalization and therefore all are equally important in the final outcome of the MRI [14] [47]. Therefore, Equation (2) was used to combine variables to generate the resilience index of the dimensions [47] [14].

y i = ∑ ​ ( X 1 W 1 + X 2 W 2 + X 3 W 3 + ⋯ + X n W n ) (2)

where: y is the dimension of the resilience index, X is any normalized variable, W is the weight of any variable, n is the number of the variable or weights considered and i is the number of variables.

Third, to generate the overall Municipality Resilience Index of Mbale municipality, Equation (3) proposed by Mayunga [48] was used.

M R I = ( ∑ ( W 1 P D i + W 2 S D i + W 3 E D i + W 4 I D i ) ) / n (3)

where: PD_i is the index of the physical dimension, SD_i is the index of the social dimension, ED_i is the index of the economic dimension, ID_i is the index of the institutional dimension, W is the weight, n is the number of variables in the dimensions and i is the variable number.

Measuring resilience requires complex indicators and variables and thus it is often a challenge for a single index to be able to capture all variables [33]. This index gives emphasis to using subjective methods which may limit generalization of results. However, attempts need to be made to visit secondary sources to obtain objective variables where data cannot be obtained from households. Despite the limitations, the study presents and tests an assessment index to understand urban resilience in a developing country. This is important because it can allow government or other organisations that seek to improve resilience over time using consistent assessors [33].

Overall, the resilience index of Mbale municipality is 0.2. This implies a low resilience of the municipality to climate related shocks and stresses. This confirms reports that the municipality is located in a low socio-economically developed region of Uganda characterized by high level of vulnerability to climate related shocks and stresses [49]. In terms of the three divisions in the municipality (Figure 2), results revealed that, Industrial division has the highest average resilience score (1.7), followed by Northern division (0.9) and Wanale division (−1.9). The relatively higher resilience score by Industrial division contradicts findings of the previous study in the municipality that focused on only the social dimension and reported that Industrial division was the least resilient [50]. However, the study finding confirms studies that argue that resilience can be well understood by exploring its multidimensional aspects rather than just a single dimension [10] [14].

Findings in Figure 3 show that the physical dimension performed quite well in Northern divisions with an average resilience index of 3.1 and so poorly in other divisions. The economic dimension performed quite well in Wanale division (3.7) and so poorly in other divisions. The institutional dimension performed quite well in Industrial division (4.7) and so poorly in other divisions. However, the social dimension performed poorly in all the divisions with an average resilience index of 1.7 in Industrial, 0.3 in Northern and −2.4 in Wanale. This findings imply that Northern division has achieved more of the physical factors such as access to water, electricity, housing and transport as compared to other factors while Industrial and Wanale divisions may not have achieved enough in providing the physical factors. Whereas, Wanale division has achieved more in the economic factors such as savings and credit as compared to other factors, it is the institutional dimension that has been well achieved in Industrial division. This could be because of the various government and non-governmental institutions that have been strengthened in Industrial division and this have been appreciated by the population. From these findings, it is clear that the contributions of the four dimensions towards the resilience index of Mbale municipality is spatially differentiated. They indicate non uniform strengths and weaknesses in performance in the three divisions. This is likely to be due to differing priorities given to different sectors by the local governments in the three divisions. The overall poor performance of the social dimension in all the divisions corroborates with the results by Kamh [14] who reported that the social dimension performed poorly towards the resilience index of Jakarta and Alexandria.

Findings on the performance of the four dimensions of the MRI (Figure 4) revealed that the physical dimension scores 1.3, economic (1.2), institutional (1.1) and the social (0.7). The relatively high score by the physical dimension as compared to other dimensions implies that household in the municipality have currently achieved more of the elements in the physical dimension i.e. access to electricity, water, transport, housing and sanitation. Furthermore, the least score by the social dimension may imply that livelihoods in the urban environment are less defined by the social variables. This relates to the finding by FAO-UNICEF-WPF [51] who reported the highest score by the physical dimension and the lowest score by the social dimension towards the resilience of urban livelihoods in Dolow in Somalia. However, this study finding contradicts the finding of Kotzee and Reyers [12] in Eden district in South Africa who reported that it was the social factors which had a high resilience score as compared to other factors. This variations could be a result of how much the different governments and other stake holders have injected towards the development of the different sectors and thus points towards where more efforts need to be vested to achieve a holistic and multidimensional resilience.

The performance of the variables under the physical, social, economic and institutional dimensions are illustrated in Figures 5-8. In terms of the physical dimension (Figure 5), results generally revealed that access to water, ownership of toilets, diversity of waste management and housing highly contributed to the current household resilience while safety and alternative capacity of water, supply and reliability of electricity, connection to reliable sewage network, ownership of shelter, means of communication and adequate nutritious food were low contributors of current household resilience. This results give a picture on the contribution of each variable as shown by their factor loading. Factor loading indicate the relative statistical contribution of the observed variables in the resilience dimensions [51]. A high factor loading suggests that the variance of the latent variable is highly correlated to the absolute value of the observed variable. “A variable with a high factor loading within a dimension that has a high factor loading in the resilience index is a variable which, if improved by one standard deviation unit would affect the resilience index most positively” [51]. While a low factor loading suggests a low correlation between the observed variable and the latent variable at given point in time. These indicate weak contributors to the current resilience index and may indicate variables which have consistently low values; however, the impact of an improvement on resilience could be important. For example, so few people in the sample have reliable electricity would not be correct to conclude that electricity is irrelevant [51].

Among the social variables (Figure 6), networks with: NGO’s, relatives, friends and government have high factor loading as compared to the level and quality of education, access to health and health stockpiles to manage emergencies. Social networks have often been regarded as first responders during times of shocks and stresses because they can help provide gifts, loans, emotional support and early warnings which are important in enhancing peoples resilience [52] [53]. Networks with friends and relatives from the rural communities played an important role towards household resilience to food insecurity in Harare in Zimbabwe [54]. In the same vein, the resilience of urban refugees in Nairobi, Kenya was enhanced by establishing networks with NGO’s particularly religious organisations [55].

The performance of the variables under the economic dimension (Figure 7) revealed that monthly income, savings and access to credit have a high factor loading implying that households have achieved more in these variables as compared to alternative income sources, employment, remittances, pension and insurance cover. The high contribution by monthly income, savings and access to credit indicate that attempts by government and other stake holders to promote savings and credit have yielded progress and have been appreciated by households in improving their resilience. The importance of income and credit is always seen as stimulator for investment in income generating activities while savings act as a spare capacity to withstand shocks and stresses. Several studies have indicated the importance of income, access to credit and savings in enhancing household and community resilience [10] [29].

Under the institutional dimension (Figure 8), existence of an authority dealing in climate change, stakeholder involvement, climate awareness programs and early weather warnings relatively scored high, while ecosystem protection, representation of marginalized groups, vulnerability assessment and emergency facilities have very low scores. This reflects awareness about climate change shocks and stresses has been created and the population acknowledges this achievement while areas of good governance are still staggering in enabling the creation of a resilient Mbale municipality.