Taxonomy and Ex Ante Metric of Climate Change Adaptation Projects Recorded in the Nationally Determined Contributions (NDCs) as Updated for Conference of the Parties-26 (COP-26)

Abstract

The authors have proposed a method of reiterating the statistical analysis of the Nationally Determined Contributions (NDCs) of the UNFCCC Parties, which were updated at Conference of the Parties-26. The present analysis confirms the taxonomy developed in 2020, based on 2475 adaptive solutions recorded in 2022 NDCs, and discusses the differences observed. An ex ante adaptation metric is proposed, which allows monitoring of adaptive solutions over time and comparisons between projects in time and space. The fitness coefficient evaluates the ex ante relevance of these adaptive projects in relation to the climate challenges of each country. The authors have proposed a program of continuous improvement instead of a definitive calculation. The authors have developed an algorithm to automate the text analysis and minimize the subjectivity of the analysis. The objective is to assign a level of vulnerability to each project for each hazard in the country. The correspondence analysis was used to derive the most representative dimensions of project category dispersion and vulnerability intensities from a contingency table for each hazard. This coefficient can be made available to experts, project developers, and funders for ex ante evaluation and selection of candidate projects for funding before more in-depth analyses are carried out.

1. Introduction

Adaptation to climate change is a critical challenge facing humanity in the present and future [1,2]. Climate change acts as a “threat multiplier”, exacerbating resource scarcity and increasing stress on social-ecological systems, particularly due to the growing world population, increasing demand for food, water, and energy, and decreasing natural resources [3,4]. Climate change is known to interact with other anthropogenic environmental impacts, causing severe weather events such as floods, storms, droughts, heat waves, and fires, as well as land and forest degradation and increased groundwater salinization [5,6].

The 2nd volume of the 6th IPCC report published on 28 February 2022 emphasized the urgency of adaptation in reducing vulnerability to climate change, contributing to sustainable development, and limiting the increase in global average temperature to below 2 °C above pre-industrial levels [7]. The Paris Agreement requires signatory parties to prepare, communicate, and update their successive nationally determined contributions (NDCs) [8], and many countries include adaptation projects in their NDCs, contributing to the global adaptation target [1,9].

Measuring, Reporting, and Evaluating (MRE) adaptation is a challenging task for policy makers and countries due to the context-dependent and multifactorial nature of adaptation projects [10,11]. Evaluation tools for adaptation actions can be divided into two categories: ex post tools, which provide information on the scope and effectiveness of ongoing or past actions, and ex ante methods and indicators that assist in financing or choosing an adaptation project [12]. It is widely agreed that a universal ex post metric is not possible [13,14].

Boutang et al. (2020) [2] developed a taxonomy and ex ante metric to assess the effectiveness of adaptation projects in NDCs, using the concept of “fitness”, calculated from a contingency table representing the frequency of a project category. After the update of NDCs at COP-26, governments registered new adaptation projects with varying presentation styles. Some countries presented projects as a list of actions without specifying sectors, while others presented a breakdown by sector and included a description of climate change challenges in the sector [15]. Despite the varying quality and format of adaptation projects in NDCs, this work provides a decision-support tool for policymakers, assuming the choice of projects is based on meeting a given challenge. However, if the choice of projects is influenced or imposed by powerful stakeholders, the relevance of this metric may be limited [16].

This article uses the same methodology as Boutang et al. (2020) [2] to analyze the adaptation components of the updated NDCs submitted by parties to the Paris Agreement at COP-26 [17]. Parties must periodically (every 5 years) communicate their updated NDC or a communication on adaptation and provide information on mitigation, adaptation, and international support commitments and actions [9]. The NDCs analyzed in this article were collected and analyzed using the methodology developed by Dixit et al. (2022) [18] and are better aligned with national adaptation plans and policies than the original NDCs [19]. The methodology used in this article is not the weakness of the work, but rather the quality of the projects recorded in the NDCs by governments.

2. Literature Review

2.1. Concept of Climate Change Adaptation Metrics

The delay in adaptive responses can be explained by the polysemy and complexity of the notion of adaptation. It is a “framework concept” that aims at the resilience of developments and activities in vulnerable areas. Understanding adaptation often means adopting an ecosystem approach and an “inter-cognitive” approach in response to complex global societal and climatic changes [20,21]. In biology, the concept of adaptation is old [22]. In February 2022, the IPCC experts described the concept of anthropogenic adaptation to climate change in their Part II report [7]: “Adaptation in ecological systems involves autonomous adjustments through ecological and evolutionary processes. In human systems, adaptation can be anticipatory or reactive, as well as incremental and/or transformational. Firstly, an adaptive metric serves to answer the question “do we need to adapt?” by taking stock of the situation under study, i.e., highlighting the effects of climate change. Once the need for adaptation actions has been established, a metric can be used to answer the question, “how do we prioritize among identified needs?” In this case, it is a matter of comparing the negative effects of climate change on several sectors of activity (such as agriculture or industry) and choosing, via climate forecasts and simulations, which sector should be prioritized in the face of the current and future effects of climate change. Finally, the third and last function of an adaptive metric is to answer the question “how to undertake the MRE of adaptation?”, i.e., to evaluate the effectiveness of adaptation projects vis-à-vis a particular vulnerability over time and ex post [3,12,23]. In this paper, we focus on the second and third applications.

By the term metrics, Leiter et al. refer to two types of tools: indicators (a single variable) and indices (a combination of several indicators) [21]. They also raise important questions about the nature of the tools used. The first concerns the geographical horizon of the indicator. Depending on whether the indicator considers a local, regional, national, or global scale, it will not have the same precision, and its degree of comparability will not be the same. In this respect, studies differ widely. For example, Jacob et al. (2022) constructed an index using data from 139 Quebec municipalities to evaluate the propensity to adapt municipal, regional, or city entities [11]. St-Laurent et al. (2022) propose a series of global evaluation criteria organized around four axes: use of information, project management, ecological and social results, and progress in the field of adaptation [24]. Here we propose a national-level metric that would retain comparative qualities.

According to the Union for the Environment [13] adaptation metrics are valuable in their informative quality. However, they cannot provide explanations for the figures they present [13]. Furthermore, by focusing on a specific aspect, a metric gives importance to a data item without this necessarily being justified in a more global context. Leiter et al. (2019) emphasize the rigor associated with data collection as well as the calculation methods used [21]. To meet this need for transparency, several countries provide information sheets in their NAPs presenting data sources, calculation methods, and reference values used for interpretations.

The temporal dimension of indicators is also considered by experts [21,25]. This involves assessing the moment at which an indicator is used, i.e., whether it is used before (ex ante) or after (ex post) the implementation of adaptation projects. Most of the MRE work associated with adaptation falls within the framework of “ex-post” analysis to evaluate its effectiveness. This is, as explicitly stated in the example of Albania, a process of learning from the past, a “learning by doing” approach. Although few researchers have nevertheless attempted to construct adaptation metrics within the ex ante framework. For example, Ricalde et al. (2021) [26] proposed, based on a study carried out in the water sector in Chile, a comparative analysis using modeling and simulations.

Adaptation is the central topic of Charles Darwin’s 1859 theory of evolution by natural selection [22]. Richard Dawkins (2018) proposed a program of work to reconstruct ancient environments from the adaptive characteristics of fossil living things. “The ‘adaptations’ of an animal, its anatomical details, instincts, and internal biochemistry, are a series of keys that exquisitely fit the locks that constituted its ancestral environments” [1]. We take this proposal and apply it to adaptive projects and environments degraded by climate change by analogy with living things. As with other living beings, humans adopt adaptive keys to the consequences of climate change (environmental locks). The keys that are most effective in addressing the issues at stake will ultimately be chosen by humanity and recorded in their NDCs.

Here again, we attempt to estimate the statistical value of the correspondences between keys and locks (solutions and hazards). By comparing different portfolios of institutional and infrastructural interventions that can be built under different hydrological scenarios, the authors have built a decision support tool that can be used before choosing an adaptation action [27].

2.2. Concept of Taxonomy of Adaptive Projects

Metrics assess the relevance or effectiveness of specific solutions. However, can these expected qualities also be assessed for categories that group together homogeneous issues? Organizing adaptive solutions into homogeneous and related categories, or taxonomies, helps with adaptation decisions. For example, a taxonomy of climate risks was created at the initiative of the UN Office for Disaster Risk Reduction (UNDRR) and the International Science Council (ISC) [28]. This is the Sendai Framework, which identifies 74 risks divided into the following 6 categories: biological hazards, environmental hazards, geological or geophysical hazards, hydrometeorological hazards, technological hazards, societal hazards, and other uncategorized hazards. The European Union has created a taxonomy dedicated to adaptation activities in 2019. The objective of this taxonomy is to help investors, companies, issuers, and project developers implement a transition to a low-carbon, resilient, and resource-efficient economy. This taxonomy distinguishes between two types of substantial contributions to adaptation objectives: “adaptive activities” (i.e., projects deployed to respond to a particular situation) and “activities enabling the adaptation of an economic activity” (i.e., projects that are intended to foster the adaptation of an entire sector of activity) [28,29].

As demonstrated by Boutang et al. (2020) [2], this type of classification can be built from the observation of NDC documents as well as climate databases. We update this taxonomy here according to issues and projects revised at the end of 2022.

3. Materials and Methods

3.1. Data Base

Many UNFCCC Parties have revised their NDCs for COP-26 with a higher level of ambition. Where a country’s NDC did not provide detailed information on adaptive projects, we used that country’s National Adaptation Plan (NAP). The main objectives of a NAP are to identify medium- and long-term adaptation needs and develop and implement associated strategies and programs.

The database consists of 3643 Excel rows (3643 projects) from 142 countries. However, not all the projects identified were considered “usable”. A project is considered “usable” for the purposes of our analysis if: (1) When a hazard can be assigned, i.e., when the NDCs and/or NAPs provide sufficient information to understand the climate issue of the project in question. They were classified as “unknown hazards” when they were not sufficiently well formulated to be classifiable in the taxonomy or as “multi-hazards” when their formulation was too general and did not allow for the identification of specific hazards. From a statistical perspective, this sorting avoids the bias that would have resulted from a forced or arbitrary assignment of a key (project) to a lock (hazard). (2) Where information on the intensity of the hazard in question is held in the database that was again chosen in this paper, namely DARA (see Section 3.2 and Section 3.4) for the country initiating the project. Projects from the following countries were excluded from our analysis: Andorra, the State of Palestine, Monaco, and St. Kitts and Nevis. These countries are not included in the DARA database.

By applying these two criteria, 2475 projects are considered usable, i.e., 67.9% of the total identified.

3.2. Categories of Physical Risks

Adaptation cannot be detached from the impacts of the climate change-induced hazards it claims to reduce, which is why the authors have paid great attention to climate risk assessment [30]. While vulnerability is defined for a specific system, sector, or region (as a degree of sensitivity and adaptive capacity to a risk), physical risk applies to all systems, sectors, and regions [31]. We have therefore chosen, as was done by Boutang et al. in 2020 [2], to associate adaptive projects with physical risks (hazards), which allow international comparisons. A database of climate prediction risks for each country is provided by DARA, independent of the assumptions made by the countries in drafting their NDCs. DARA is a non-profit organization that aims to improve the performance of humanitarian aid by conducting evaluations, research, and policy studies. In total, a list of seven (7) physical hazards was compiled from the DARA database [32] and the study by Mirza et al. [12]: drought, flood, heat wave, sea level rise, storm, temperature increase, and rainfall. However, the names associated with the temperature increase and heat wave hazards are not obvious in DARA. For these two hazards, the DARA correspondences “heating and cooling” and “heat and cold illnesses” were used, respectively.

Each of the 2475 selected projects was associated with one or more hazards. This project-hazard association makes it possible to construct a taxonomy.

3.3. Taxonomy of Adaptive Solutions for Each Type of Hazard

To associate this project with hazards, we developed in 2020 an algorithm (R code) to automate the text analysis and to minimize the subjectivity of the analysis. This algorithm is available as supplementary material on request. It allows to highlight the projects attributed to each hazard, and then perform a lexical analysis by breaking them down into word pairs. Trivial elements, such as linking words and numbers, have been eliminated. The results are a list of word pairs ranked in descending order. We retained the occurrences greater than or equal to four, which seemed more significant. The limitation of the threshold of occurrences to four was preferred to decrease the plethoric volume of word pairs obtained, to avoid the repetition of several word pairs belonging to the same project, and finally to facilitate the classification. The classification of these occurrences and the links from the most important to the weakest made it possible to create project categories. Selectively, we sorted the word pairs by category and relationship. Some word pairs that did not express a specific adaptation action were discarded, such as “climate change, adaptation action, climate resilience, […]”. The taxonomy was then organized from the selected word pairs, starting from the most general association to the most specific pair. The number of categories and projects per category is available in the results section and in Appendix A. The taxonomy by hazard is visualized in tree form with the software “Xmind. 22.09.3168” (Appendix A).

3.4. Risk Level for Each Hazard-Country Combination

The objective is to assign a level of vulnerability to each project for each hazard in the country. For several hazards, the Climate Vulnerability Monitor assigns a level of hazard intensity to each country by 2030. The present study uses the 2030 vulnerability level, which is closer to 2022 than the 2010 horizon also proposed by DARA. Four levels of “vulnerability” (actually “risk”, as defined above) determined by DARA have been selected: LowModerate, High, Severe, and Acute. Thus, each project is no longer associated with a hazard but with one of the four levels of that hazard.

3.5. Contingency Tables, Correspondence Analysis, and Adaptive Fitness Coefficient

For each type of hazard, we constructed a contingency table with the adaptive projects determined and the four levels of intensity (see the result section and Appendix A). These contingency tables present the frequency of project categories (keys) in relation to hazard intensities (locks). The distribution of variables between the intensity columns is the basis for statistical correspondence analysis (CA) between the qualitative hazards and project variables. This CA also allows the study of the possible relationship between adaptation project categories and hazard intensities.

The CA calculates the vector distances between all project categories and hazard intensity levels within a hazard class (e.g., temperature increase). The three dimensions (coordinates in vector space) most representative of the variance between distances are selected. Then, a second time and thanks to the distances obtained, a score between 0 and 1 is achieved by normalizing the minimal and maximal distances observed in the whole sample. A distance between each project category and each hazard intensity is calculated. The vector distance between an environmental key and a hazard intensity is calculated as follows [2]:

$$\mathrm{Dij}=\sqrt{{\left(I_1-J_1\right)}^2+{\left(I_2-J_2\right)}^2+\left(I_3-J_3\right)²}$$

(1)

where $I_1$ is the coordinate of the environmental key i for the first dimension; $J_1$ is the coordinate of hazard intensity j for the first dimension and Dij is the vector distance between environmental key i and hazard intensity j.

At this stage, we obtain vector distances, which are expressed numerically. The smaller the distance, the stronger the link between project class and hazard intensity.

Then we can calculate the fitness coefficient with Equation (2). The closer the fitness coefficient is to 1, the smaller the vector distance between the project class and the level of “vulnerability”, and therefore the more statistically “relevant” the adaptive action is. Indeed, we define the relevance of an action to risk as the degree of statistical correspondence between the most recent human choices (the choice repository) and levels of climate hazard intensity [2].

Fitnessij = 1 − D′ij

(2)

with

$$\mathrm{D}\prime \mathrm{ij}=\frac{\mathrm{Dij}-\mathrm{Dmin}}{\mathrm{Dmax}-\mathrm{Dmin}}$$

(3)

where D’ij is the normalized vector distance between environmental key i and hazard intensity j. The maximum (Dmax) and minimum (Dmin) vector distances are taken for a hazard, along with the four intensity characteristics that characterize it.

These fitness measures are different from the frequency alone, although obviously frequency is at the root of the fitness metric. This is because fitness is a statistic, measured partly from the frequency of projects (this is the true content of the contingency tables) and partly from relative distances. As a result of this, the difference in value between the frequency and the fitness measure is that a key (project category) may have a low frequency at a given intensity but a high fitness measure or it may have the same frequency but different fitness values.

4. Results

4.1. Observation of Contingency Tables and Taxonomy

By hazard, the distribution of projects operated is very broad. Rainfall, temperature increase, drought, and flood are the most represented hazards (

Table 1. Distribution of projects by hazard.

Hazards	Number of Projects and %
Flood	1022	28.0%
Drought	1173	32.2%
Storm	163	4.5%
Heatwaves	570	15.6%
Sea level rise	296	8.1%
Temperature increase	1387	38.1%

). The least represented hazards are storm, sea level rise and heat waves. It should be noted that a project may belong to one or more hazards.

Table 2. Number of project categories created for each risk class.

Risk Category	Number of Projects Categories
Flood	53
Drought	49
Storm	16
Heat waves	30
Sea level rise	21
Temperature increase	64

shows the number of categories in the taxonomy for each hazard. The taxonomy of a single hazard is represented at different levels of category depending on the parentage or affiliation of the projects (Figure 1 and Appendix A). Here the term “category” refers to a group of adaptive projects or solutions with the same environmental or societal target. Most adaptive projects are found in the lower levels of the categories (Figure 1 and Appendix A). This distribution of projects in these lower levels could indicate the specificity of the projects, as the more specific the projects are, the lower the levels (branches) of the taxonomy.

4.2. Analysis of Projects in Relation to Vulnerability Level Intensities

The adaptive projects for each hazard were distributed according to the level of vulnerability of their country of adoption. The distribution of projects varies widely across vulnerability intensities within and between hazards (Figure 2). For example, in the case of the “Drought” and “Heat waves” hazards, adaptive measures are more concerned with the “High” intensity, whereas in the case of the “Temperature increase” and “Sea level rise” hazards, the measures taken are for the “Low/Mod” intensity (Figure 2). This shows that in progressive risk situations (gradual temperature or sea level rise), the “Low/Mod” intensity presents a major part of the projects, contrary to catastrophic situations such as “Drought” and “Heat waves”, where the importance of the projects is on the “High” intensity.

Of the total, the “Low/Mod” and ‘“High” intensities have the largest number of projects, 1644 and 1381, respectively, while the ‘“Severe” intensity has only 609 projects (Figure 2).

The contingency tables present a detailed analysis of projects according to intensities. As an example, the contingency table below illustrates the distribution of storm projects in relation to vulnerability levels. Here, more than 76% of the projects are on the “low/mod” vulnerability level (

Table 3. Contingency table for the “Storm” class.

Storm Keys	Low/Mod	High	Severe	Acute	Meta-Category	Total
Coastal management	18	0	1	4	Coastal management	23
Food security	2	0	0	1	Coastal management	3
Human settlement	1	0	0	1	Coastal management	2
Insurance scheme	4	0	0	3	Coastal management	7
Knowledge management	5	0	1	3	Coastal management	9
Mangrove	6	0	0	4	Coastal management	10
Water management	11	0	1	0	Coastal management	12
Capacity building	3	3	0	3	Coastal protection	9
Coastal protection	10	1	0	3	Coastal protection	14
Infrastructure control	8	0	0	0	Coastal protection	8
Infrastructure development	6	0	0	0	Coastal protection	6
Marine protection	6	0	0	1	Coastal protection	7
Monitor system	3	0	0	1	Coastal protection	4
Public awareness	7	0	1	0	Coastal protection	8
Warning system	30	2	0	3	Coastal protection	35
Weather forecast	4	0	0	2	Coastal protection	6
Total	124	6	4	29		163

). This implies that most of the adaptive measures are taken only in low to moderate storm risk situations. For example, of the 35 warning system measures, more than 85% are taken at a low to moderate vulnerability level.

4.3. Fitness Coefficient or Relevance Value of Project Categories

Correspondence analysis (CA) was used to derive the most representative dimensions of project category dispersion and vulnerability intensities from a contingency table previously constructed for each hazard (Table 3 and Appendix A).

All examples given below are just picked up for clarification and illustration, not because they are specific in any sense.

Consider the fitness of the storm hazard categories (

Table 4. Adaptation fitness coefficient (between 0 and 1) calculated for adaptive project categories associated with “Storm” risk. The colors indicate the relevance of the keys according to the risk intensities. Highest values are marked by dark orange, and lowest values by light blue. Intermediate orange intensities are put according to the degree of relevance of the adaptive solution. The highest the fitness value is, the darkest the orange color. Very low fitness values are colored in light blue.

Storm Fitness Keys	LowMod	High	Severe	Acute
Coastal management	0.98	0.14	0.47	0.73
Food security	0.83	0.16	0.32	0.88
Human settlement	0.66	0.13	0.25	0.98
Insurance scheme	0.74	0.15	0.29	0.96
Knowledge management	0.73	0.11	0.58	0.76
Mangrove	0.76	0.15	0.30	0.94
Water management	0.88	0.09	0.56	0.53
Capacity building	0.33	0.84	0.00	0.44
Coastal protection	0.90	0.31	0.33	0.79
Infrastructure control	0.93	0.11	0.34	0.55
Infrastructure development	0.93	0.11	0.34	0.55
Marine protection	1.00	0.14	0.36	0.69
Monitor system	0.92	0.16	0.34	0.80
Public awareness	0.78	0.06	0.67	0.49
Warning system	0.97	0.25	0.34	0.66
Weather forecast	0.83	0.16	0.32	0.88

). For this hazard class, the vulnerability levels “Low/Mod” and “Acute” have high scores for almost all adaptive project categories (Table 4). This means that adaptation projects, when there is a risk of storms, are more likely to be undertaken at low/moderate and acute vulnerability levels. For example, categories such as “warning system”, “coastal protection”, “infrastructure control”, and “marine protection” represent very high fitness scores of 0.97, 0.90, 0.93, and 1, respectively, on the “Low/Mod” vulnerability level. These are the most relevant measures a priori related to this risk intensity. Similarly, on the “Acute” intensity, we find the highest fitness coefficient scores in the categories of “insurance scheme”, “human settlement”, and “knowledge management”, with 0.96, 0.98, and 0.76, respectively. These are the adaptive measures that, ex ante, appear to be the most relevant at this level of storm risk vulnerability.

5. Discussion and Limitations

5.1. Project Descriptions and Taxonomies

5.1.1. Brief Description of the Projects

The sample size, representing 68% of all projects described in the revised NDCs and NAPs, is comparable to or slightly smaller than the sample in Boutang et al. (2020) (74.0%) [2]. Projects that lack clarity in their risk objectives, are vague in the nature of the adaptive project, or are off-topic have been discarded. One limitation of the taxonomy is that projects are treated as independent, when in reality, actions are often interdependent (Dixit et al., 2022 [18]). Transformative adaptation in NDCs is analyzed to determine the extent to which adaptation actions align with transformative approaches, such as actions that aim to create systemic change through innovation, location shifts, or scale expansion in response to climate change. However, the analysis does not provide qualitative descriptions or a sense of priority, and it is not within its scope to assess the possible maladjustments or transformative power of these actions. The analysis is comparative within the same hazard class and internationally. It cannot be used to judge issue priorities directly, unlike the Adapt Now report from the Global Commission on Adaptation [33]. However, the DARA country risk database can be used for this purpose [32]. The taxonomy used in this study is “under construction,” with meta-categories of projects (such as coastal management) being a grouping of projects with generic descriptions, rather than the sum of underlying projects. The underlying projects relate to the generic description without being a subset of it. As projects are better described, the number of projects in meta-categories will decrease, and the term meta-category will become the set to which all underlying projects relate. The set of terms is designed to be sustainable, with a gradual shift of projects to finer branches of the tree. However, despite efforts to rationalize and automate the reasoning behind the taxonomy, the ordering of categories into several linked levels is based on expert opinion, as no algorithms have been identified to replace human judgment.

5.1.2. Vulnerability and Hazard

Future studies should use climate forecasts for regions affected by adaptive projects rather than entire nations. One limitation is that solutions are linked to hazards and not vulnerability in the strict sense. Vulnerability involves factors such as topography, geology, land use, and infrastructure. For example, water vulnerability from a precipitation hazard is assessed through complex modeling and in situ observations. A systematic comparison of 2475 projects between solutions and local vulnerability is currently not feasible. The study uses the DARA database, but the method can be replicated using any more reliable, complete, or up-to-date climate hazard databases. No statistics on the “rainfall” hazard were found in DARA.

5.1.3. The Fitness Coefficient

Our analysis allows for a comparative assessment of the suitability of adaptive solutions for a given vulnerability intensity within each hazard. The fitness coefficient was obtained through a comparative assessment and expresses the largest proportion (+75%) of variance. A project category with a high fitness coefficient (e.g., >0.80) for a given level of intensity is in line with commonly adopted solutions worldwide for a given hazard intensity. The fitness coefficient values may or may not confirm the results from contingency tables. For example, the adaptive measure “warning system” in the “storm” risk class has a high fitness and is also more numerous. The coefficient enhances the intuition from the count by revealing the relevance of choices not often found in contingency tables, such as at the “acute” level for the same risk. No orphan or rare cases of risk intensity versus project pairs were encountered, allowing for the calculation of the coefficient.

5.2. Comparison of Fitness Coefficient Values between Two Editions of NDCs

The fitness coefficient values were compared with those obtained by Boutang et al. (2020) [2] for all hazard categories. The evolution of the number of projects for each hazard and intensity between the two NDC editions is presented in Figure 3. The number of projects for the hazards “Drought” and “Heat Waves” has doubled in the new edition of NDCs (32.2% and 15.7%, respectively, compared to 13.1% and 2.6% in the previous edition). There has also been a clear increase in the number of projects for the hazards “Flood,” “Rainfall,” and “Temperature Increase.” However, the number of projects for the hazard “Storm” has decreased by 3.3% (from 7.8% in 2019 to 4.5% in 2022). This distribution of projects according to hazards highlights the pressing hazards and the adaptation needs of humanity.

The taxonomy has evolved between the two NDC editions in terms of project categories. The categories for “Heat Waves” and “Temperature Increase” have increased by a factor of 2, and the categories for “Drought,” “Flood,” and “Rainfall” have nearly doubled (as presented in

Table 5. Number of projects categories created for each risk class in 2019 and 2022.

Risk Category	Number of Projects Categories (2022)	Number of Projects Categories (2019)	Difference
Flood	53	27	+26
Drought	49	25	+24
Storm	16	21	−5
Heatwaves	30	7	+23
Sea level rise	21	20	+1
Temperature increase	64	27	+37
Rainfall	67	45	+22

). Conversely, the categories for “Storm” have decreased (as presented in Table 4). This disparity in the growth of taxonomy categories between hazards indicates the level of risk posed and the direction of human adaptation. In 2020, the EU Technical Expert Group on Sustainable Finance developed a taxonomy of adaptive solutions that closely resembles ours (as described in [29]).

Most of the new categories are located at levels 3 and 4 of the taxonomy trees (as presented in Appendix A). Research has shown that the more specific and detailed the measures for coping with risk, the greater the control and understanding of the risk (as described in [1,34,35]). It is noteworthy that the new taxonomy includes new categories, and some categories have been removed compared to Boutang et al. (2020) [2]. Despite these changes, the core of the taxonomy remains unchanged between the two publications.

5.3. Improvement Prospects

The prospects for improvement of the proposed metric and taxonomy are promising.

• Robust Taxonomy: As the specificity of projects increases, the taxonomy will become stronger, featuring a more extensive branching structure. This will result in a shift from projects categorized in meta-categories to more specific underlying levels;
• Reliability of Statistical Calculation: The use of regional climate models can improve the statistical calculation by enabling the analysis of the relationship between risk intensity in the project region and the adaptive solutions specified in the NDCs. Furthermore, a clear description of adaptive NDC projects will permit the calculation of the fitness coefficient on a larger sample of NDCs, increasing its reliability;
• Convergence or Divergence in Adaptive Choices: As demonstrated in this study, between the NDCs submitted for COP-21 and the revised NDCs submitted for COP-26, there was a substantial increase in the number of projects related to the hazards of “Drought”, “Flood”, “Heat wave”, “Rainfall”, and “Temperature increase” (Figure 3). This suggests a convergence in adaptive choices, but further analysis is required to determine if this trend will continue or diverge.

6. Conclusions and Take-Away Messages

The lack of universal metrics to compare different adaptive projects within the same class of climate risk or vulnerability is a significant obstacle to the establishment of sustainable and structured financing mechanisms for adaptation. To address this issue, we propose a fitness coefficient as a statistical evaluation of the a priori relevance of the adaptive solutions recorded in countries’ NDCs (Nationally Determined Contributions) in relation to various climate risks.

Our method involves categorizing 2475 projects from around the world and comparing them using a number of assumptions and methods. A statistical analysis of the relationship between project classes and hazard intensities allows us to calculate a normalized fitness coefficient between 0 and 1, indicating the mathematical correspondence between adaptation solutions and climate challenges. This normalization is performed across the risk class for all predicted hazard intensities.

This ex ante metric can be used by project developers, development banks, and experts to assess the relevance of an adaptive solution to a particular climate challenge. Project leaders can use the fitness coefficient when applying for funding, while funders can use it to shortlist candidate projects for further analysis.

We anticipate that the taxonomy of solutions and the proposed metric will continue to evolve and improve. As projects become more specific, the taxonomy will become more robust, and the fitness coefficient will become more reliable with the availability of regional climate models and a clearer description of adaptive NDC projects.

Additionally, it would be interesting to update the calculation of the fitness coefficient at each new NDC submission round under the Paris Agreement. Calculating the standard deviation for each of the six hazards (flood, drought, storm, heat waves, sea level rise, and temperature increase) could judge the difference in relevance of humanity’s choices over time and indicate the convergence of adaptive choices regarding climate issues.