The role of perception in subsistence farmer adaptation in Africa: Enriching the climate finance debate Open Access

African communities and economies are very vulnerable to climate change (Boko et al., 2009) and at recent UN climate summits the continent is prioritized to receive international climate change adaptation finance (UNFCCC, 2009; decision 3/CP.15 and UNFCCC, 2010; 95/CP.16). Annual climate finance should approach US$100 billion from 2020 onwards. How to use the share that goes to adaptation efficiently and effectively is still under discussion (Chaum et al., 2011; Denton, 2010; Schalatek, 2011). This paper focuses on subsistence‐based agriculture, which has a key role in adaptation in Africa.

The agricultural sector contributes on average 29 per cent of the GDP of African countries and is largely subsistence‐based (UNECA, 2001). Whilst the share of the population active in agriculture is decreasing, its absolute number climbs towards 600 million in 2020 (FAO, 2011). As agriculture is generally rain‐fed (FAO, 2011) it is very sensitive to precipitation anomalies. With agricultural systems already under performing and environments degraded (Magadza, 2003), subsistence farmers need to adapt to climate change. This means:

Although changing climatic conditions are often perceived by subsistence farmers (Bryan et al., 2009; Deressa et al., 2009), information provision on climate change in rural Africa is generally low (van Aalst et al., 2008; Nhemachena and Hassan, 2007), which is indicative for subsistence farmers' knowledge on the matter. How can international adaptation finance support millions of subsistence farmers to adapt to climate change if these people hardly know or understand what it is?

Rather than looking at climate change, this paper therefore focuses on perception of climate related hazards (CRHs: droughts, extended dry seasons, floods, etc.). Understanding present‐day effects of and autonomous responses to CRHs is a prerequisite for studying the effects and responses to climate change (Adger et al., 2003; Hulme et al., 2000; van Aalst et al., 2008). Second, the focus on perception of subsistence farmers, based on the idea that understanding the way people perceive climate risks is essential to say anything meaningful about their autonomous adaptive behaviour (Grothmann and Patt, 2005). Understanding the perception of subsistence farmers can be helpful for policymakers in promoting climate change adaptation, for example through knowledge sharing, adaptation projects and risk management strategies (Patt and Schröter, 2008; Deressa et al., 2009; Bryan et al., 2009; Plapp and Werner, 2006; Nyanga et al., 2011; Kreibich, 2011). This paper goes beyond policymakers and draws initial lessons for international climate finance institutions.

This paper first provides a short summary of existing knowledge on subsistence farmers' adaptation to CRHs in Africa. It shows the increased emphasis on autonomous adaptiveness, which takes place in complex environments. It shows some standard adaptive measures in Africa, and that subsistence famers' perception plays an important role in adaptation to CRHs.

This paper builds on that knowledge based on 227 interviews in subsistence‐farmer communities in Botswana and Ghana. Results are put in the perspective of the respective countries' agricultural, environmental and economic development. In a synopsis I compiled the Botswana and Ghana data and draw conclusions on the role of the respondents' motivation, experience and means (e.g. workforce, money, assets) in adaptation to CRHs. In the conclusion lessons are drawn for international climate finance institutions.

“Local adaptiveness” is described by Smit and Skinner (2002) as a climatic change related regular part of farmers' ongoing management that takes place either spontaneous or planned. It has been part of a scientific debate long before “global warming”; the Sahel drought of the 1970s already initiated a great increase in studies. At that time, the national or regional oriented desertification narrative generalized area‐specific environmental phenomena (Slegers and Stroosnijder, 2008). Farmer's adaptiveness was considered as low. Communities were seen as inherently vulnerable and were considered dependent on strong governments and aid from Western countries (Mortimore and Adams, 2001; Roncoli et al., 2001).

In the following three decades, researchers moved away from the diagnostic‐prescriptive view on local adaptiveness towards a more long‐term perspective on development and the human‐environment interaction in particular. Climate variability was acknowledged; the Sahelian environment became known as unstable but resilient. Indigenous knowledge and local innovations were revisited as key for farm‐level adaptation to climatic risks (Stigter et al., 2005).

Climate change adaptation literature also has increased attention for the farmer, although initially distant analyses were made here too. For example, in their overview of potential adaptive measures in the agricultural sector, Smit and Skinner (2002) recognize that little research is done on the adoption of measures. Kandlikar and Risbey (2000) called the tendency to cast adaptation as a physical and market based process and the lack of attention for political and social dimensions “problematic”. Without detailed knowledge of a community it is essentially impossible to develop appropriate and applicable adaptation measures for them (Smit and Pilifosova, 2003).

The increasing focus on local scale adaptation to CRHs in Africa is described in a growing body of literature. A summary of this discussion is provided in the two subheadings below. More detailed discussions are provided by Bryan et al. (2009), Eriksen et al. (2011), Maddison (2007), Slegers (2008) and Smit and Skinner (2002).

Although the successfulness of measures to adapt to CRHs depends on the local context, some general adaptive measures prevail. These include improved soil and water management (Smit and Skinner, 2002; Stigter et al., 2005), crop diversification (Nhemachena and Hassan, 2007), small‐scale water harvesting and irrigation (Pauw et al., 2008; Deressa et al., 2009), seasonal forecasts (Patt and Gwata, 2002), education (Maddison, 2007), livelihood diversification (Thomas and Twyman, 2004), and migration (McLeman and Smit, 2006; Mertz et al., 2010).

Individual farmers apply different adaptive measures (Ziervogel et al., 2006) and apply measures differently (Pauw et al., 2008). People's most important concern is their daily life and livelihood (van Aalst et al., 2008). The climate is just one factor in the complex environment in which subsistence farmers operate (Westerhoff and Smit, 2008; Ziervogel et al., 2006; Bryan et al., 2009). Adaptation is influenced by socio‐economic, cultural, political, institutional, ecological and geographical factors that shape the human‐environment interactions (Eriksen et al., 2011; Nyanga et al., 2011). The application of adaptive measures depends for instance on the type of climate stimuli, the type of farming, a farmer's means (Dolan et al., 2001); access to markets, availability of credit and savings, knowledge (Gbetibouo, 2009); education and closeness to roads (Maddison, 2007). In such a complex setting it is difficult for researchers, policymakers and adaptation finance institutions to determine whether an activity constitutes adaptation and whether it can be financed as such.

Apart from looking at the local level, research also tries to reflect “from” the local level, through insights in people's perception of CRHs. This is useful for policymakers as perceptions are reflected in behaviour (Patt and Schröter, 2008; Deressa et al., 2009; Bryan et al., 2009; Plapp and Werner, 2006; Meze‐Hausken, 2004; Nyanga et al., 2011; Kreibich, 2011), and because differing views between governments or researchers and local people can have far‐reaching implications. For example, a lack of consensus among Tanzanian farmers and scientist on the causes of declining harvests limited the transition towards more sustainable use of natural resources (Slegers, 2008); and an expensive relocation programme in Mozambique failed because policymakers and the rural population perceived flood risks differently (Patt and Schröter, 2008).

Perceptions of climatic changes do not necessarily result in autonomous adaptation. For example, whilst majorities of farmers in several African countries perceived changes in temperature and rainfall, many farmers did not adapt to these changes (e.g. 42 per cent (Deressa et al., 2009); 44 per cent (Nyanga et al., 2011); 37‐62 per cent (Bryan et al., 2009)). Explanations are found for instance in a lack of knowledge of adaptive measures, no tenure security, or restricting socio‐economic conditions (Bryan et al., 2009). The hypothesis of this paper is that apart from perception of climatic changes additional factors play a role in subsistence farmers' perception of CRHs and adaptation, such as perceived vulnerability and motivation to adapt.

Five villages in the Dayi River Basin (Volta region) in Southeastern Ghana were selected for surveying based on discussions with the Ministry of Food and Agriculture (MoFA) regional office and the NGO Development Institute, and the willingness of village leaders to cooperate. The villages represent three distinct ecological zones in the basin: have and Woadze represent the mountainous downstream area, and Koloenu the mid‐stream transitional Savannah, where forests are largely cleared and turned into arable land. Lipke Kukurantumi and Lipke Abrani represent the upstream (still) forested area. The villages varied in size from 42 to 3,508 inhabitants (Ghana Statistical Service, 2005); in general more interviews were done in larger villages.

The fieldwork region has a tropical climate with a bimodal wet season peaking in June and September.

Local meteorological data shows that the annual rainfall decreased from 1,700 mm/year in 1975 to the present 1,400 mm/year and that the rainy season shortened. The mean annual temperature increased by 1°C in the period 1961‐2000 (Westerhoff and Smit, 2008), and is projected to keep rising (2°C in 2050) (Agyemang‐Bonsu et al., 2008).

Precipitation projections are difficult to make for the case study region. The 21 global models used by the IPCC project an overall precipitation increase but are inconsistent (Christensen et al., 2007). The more detailed medium scenario of the regional NCAP projection shows a further decline in rainfall (−10.9 per cent in 2050) and a shortening rainy season. This would be detrimental for agriculture and farmers have to adapt (Griebenow and Kishore, 2009; Agyemang‐Bonsu et al., 2008)

In a country with such diverse climatic zones and cultural diversity, the villages cannot represent Ghana as a whole. However, the large and growing agricultural labour force and arable land (FAO, 2011); constrained socio‐economic engagement with water and land resources; limited irrigation (MoFA, 2006); correlation between agricultural yields and rainfall and weather hazards (Perret and Bossard, 2008; Agyemang‐Bonsu et al., 2008); and the high levels of deforestation (Griebenow and Kishore, 2009) are key indicators of socio‐ecological interactions common both in the villages and the country.

In an integrated manner, these issues are addressed by the Dayi River Basin Board (established in 2010) and in national policies and plans such as the “Draft National Irrigation Policy” and the “Agriculture and Sustainable Land Management Strategy and Action Plan”. Activities described in the latter include building institutional and technical capacity, awareness raising, knowledge development and the creation of incentives for farmers for sustainable land management (MoFA, 2008).

Research took place in two villages in the Motloutse River Basin (Central district), near Botswana's border with South Africa and Zimbabwe. Gobojango has 1,631 inhabitants and Tsetsebjwe 3,457 (CSIR, 2003). They were selected in cooperation with the SADC Groundwater and Drought Management project leaders based on their representativeness for drought‐prone rural villages in Botswana and the willingness of village leaders to cooperate.

Botswana has one rainy season annually (November‐March). With 300‐400 mm of rainfall per annum, the case study area is arid to semi‐arid just like most of the country (Adedoyin and Mphale, 2002). The amount of rainfall is highly variable and characterized by a high degree of inter‐annual and inter‐decadal rainfall variability (Schulze, 2000) and declined by 5‐15 per cent over the twentieth century (Hulme et al., 2000). The temperature has a mean minimum of 6‐10°C in winter and a mean maximum of 28‐34°C during summer (Government of Botswana, 2000); the latter increased by 1.3°C in the period 1958‐1998 (Adedoyin and Mphale, 2002).

The IPCC A1B scenario projects the mean temperature to rise by 3.5°C towards 2100 (Christensen et al., 2007). Changes in rainfall are difficult to predict in Southern Africa. For 2100, mean decreases in rainfall of −16 per cent (March‐May), −34 per cent (June‐August) and −31 per cent (September‐November) are projected by Shongwe et al. (2009), which would shorten the rainy season. Their model is not statistically significant for the three wettest months of year (December‐February). Under climate change, water stress is very likely to increase; adding pressure to rain‐fed agriculture and reducing available vegetation for free range livestock.

Botswana is a middle‐income country but poverty is still widespread in rural areas. The agricultural sector is of key economic importance, but yields are constrained by inter alia poor farming practices, poorly organized markets, degraded lands and water shortages (Government of Botswana, 2002; Government of Botswana and United Nations, 2010). The per capita annual food production declined by 60 per cent compared to the production peak in 1971 (FAO, 2011).

The Botswana Government has developed several programmes to increase the agricultural output and to reduce effects of droughts on rural communities, such as the Labour Intensive Public Work Relief Program (still known as its predecessor “Drought Relief Programme” (DRP)) and the Integrated Support Programme for Arable Agricultural Development (ISPAAD). DRP was established in 1982 and funded through mineral exploitation revenues. It provides a safety net in times of droughts by financing bottom‐up initiatives and creating temporary jobs this way. During its peak it reached 84,000 participants, providing assistance to 45 per cent of the family farm labour force and equalling 11.1 per cent of the total rural household income (Valentine, 1993). ISPAAD is smaller and started in 2008 to address challenges in the arable sub‐sector of agriculture, to combat poor technology adoption by farmer and low productivity of arable farming (Government of Botswana, 2011). ISPAAD provides farmers with free seeds, fertilizers and draught power (www.agriculture.co.bw).

Face‐to‐face interviews were administered in subsistence farmer communities in Ghana (n=107) and Botswana (n=120) in 2009. The survey method is based on literature on surveying perceptions of natural hazards (Bird, 2009) and household surveying in developing countries (Grosh and Glewwe, 2000). The survey started with general questions on the socio‐economic and demographic situation of the respondents (e.g. education, age, agricultural practices, and income). Second, questions on inter alia perceived probabilities and consequences of CRHs, vulnerability, dependency, experience, successfulness, means and motivation to adapt to CRHs provided an overview of the respondents' perception of CRHs and adaptation. These concepts derive from literature (Grothmann and Patt, 2005; Patt and Schröter, 2008; Sjöberg, 2000; Maddison, 2007; Plapp and Werner, 2006) and were not predefined to the respondents to allow them to answer from their own perspective. Respondents were asked how successful they perceive certain adaptive measures that were collected from survey test rounds and literature (Agyemang‐Bonsu et al., 2008; Nhemachena, 2008; Ziervogel et al., 2006; Bryan et al., 2009). However, respondents neither always brought adaptive measures into practice, nor applied them in the light of CRHs alone.

The response format mainly consisted of closed questions. Many questions used a Likert scale from 0 (e.g. extremely low, not important) to 6 (e.g. extremely high, extremely important). Likert scales are “quite useful” when offering limited response categories (i.e. 7) and have many desirable quantitative properties (Sjöberg, 2000). Ordinal scales could have a non‐linearity bias, but in this context it is not very different from an interval scale (Wright, 1997) and potential mild deviations from strict linearity can be safely disregarded (Sjöberg, 2000). Regression and χ²‐tests were used to analyse the data.

Households were randomly picked for interviewing to ensure representativeness of views and experiences. Face‐to‐face interviewing gave respondents the chance to ask for clarification and ensured answering of all questions. Interpreters translated in local languages if necessary. Face‐to‐face interviewing and the spatial probability sampling can lead to deviations in the sample. For instance, men in Botswana are usually out in the field with their cattle during the day or even during the week, forcing the interviewers to work intensively in the weekends.

Interviews lasted for 45‐75 min. Response rates were 100 per cent, with 50 per cent (Ghana) and 79 per cent (Botswana) female respondents. Respondents were often but not always head of the household, with average ages of 50 (Ghana) and 53 (Botswana).

This section provides the survey results based on:

• the socio‐economic conditions in which respondents adapt to CRHs; and

• their perceptions of and ways to adapt to CRHs.

Unfortunately, the role of local institutions that inhibit or facilitate adaptation (markets, extension, NGOs, etc.) was not part of the study.

The average household size in the Ghana case study was 6.1 persons. On average respondents went to school for 11 years. Respondents complained about high levels of poverty. For example, whilst over 90 per cent lived in a brick built house, could eat three times a day and could afford their children's school fees, a majority had neither health insurance nor means of mobility. 66 per cent of the respondents mentioned an income decrease over the past five years (vs 22 per cent increase).

Of the households, 87 per cent earns more than half of their income in agriculture, mostly through common rain‐fed crops such as maize and cassava. Households cultivated five acres on average. More cultivated land led to a higher harvest value (total harvest multiplied by its market value): R²=0.255. Only six farmers irrigated land. Land tenure restricts many farmers from doing long‐term investments, such as tree‐planting (Rozema, 2009). Respondents stated that poverty causes deforestation, leading to short‐term individual gains, but long‐term adverse impacts on the community.

In Botswana, the household size averaged 5.3 persons. Respondents went to school for seven years on average. Like in Ghana, they complained about high levels of poverty. Most respondents lived in a brick built house (82 per cent) and had means of mobility (56 per cent). However, 80 per cent had no health insurance, 65 per cent did not eat three times per day, and 55 per cent could not pay the school fees of their children.

Most of the respondents grew crops (77.5 per cent) and kept livestock 58 per cent). Their yearly income varies greatly, was difficult to quantify and is unknown. Respondents complained that the surroundings are very dry and getting dryer (Bogaert, 2009). For them CRHs were droughts. They stated that present‐day harvests are lower than in the past, and that they keep smaller herds than before as less vegetation is available. This in its turn caused food shortages and reduced incomes. Some respondents were less affected as they lived close to groundwater sources that were provided by the government. A complex factor in Botswana was that respondents of all ages hold the government responsible for adaptation to CRHs. According to 55.3 per cent of the respondents the government (local, national, the chief or the president) is responsible for adaptation. Another 12.2 per cent consider God responsible. Only 3 per cent of the respondents consider themselves responsible. In terms of financing adaptation to CRHs, 69 per cent considers the government responsible. People who rely on the efficacy of public adaptation will probably take less action autonomously (Grothmann and Patt, 2005).

Ghanaian respondents stated that “drought”, “heat”, “rainfall variability” and “prolonged dry seasons” are typical CRHs with a high probability to happen (average>4). The perceived high climate‐dependency (average 4.3) and vulnerability to CRHs (average 3.8) (Figure 1) correlated significantly (R²=0.31). Respondents feared that CRHs would lead to hunger (mentioned by 91 respondents) and even death (nine times), yet in practice the most important consequence was outmigration of youth. Lack of water and failing harvests ranked second and third.

The respondents had little means to adapt to CRHs (average 1.9) and stated that financial constraints prevent control over CRHs. The motivation to adapt to CRHs was higher than in Botswana (average 4.2 vs 3.0), and had positive linear relations to their perceived experience in farming (R²=0.183), the perceived cost‐effectiveness of measures (R²=0.130), means (R²=0.088) and perceived successfulness in adaptation to CRHs (R²=0.051). The latter also positively related to farmers' means to adapt (R²=0.221) and perceived experience (R²=0.218).

A second order polynomial trend line in farmers' motivation and years of experience/age showed that farmers around with 24.2 years of experience/the age of 55.2 were most motivated to adapt to CRHs. The age and years of experience themselves also correlated (R²=0.347). Whilst the age correlated better with motivation than years of experience (R²=0.059 vs R²=0.037), the years of experience correlated better to the perceived means (R²=0.105 vs R²=0.027) and perceived vulnerability (R²=0.099 vs R²=0.001) (Figure 2).

Education, diversification of crops and irrigation were perceived as the most successful autonomous measures to adapt to CRHs; and nature conservation, irrigation and creating buffer zones around rivers as the most successful community measures. Outmigration and changing profession were considered as very ineffective (Figure 3), which needs to be seen in a socio‐cultural context (e.g. the perception of failing as a farmer, or leaving elderly people behind). In practice outmigration is important in resilience of societies (Adger, 2000; McLeman and Smit, 2006) and it reduces the regional population growth from 2.5 per cent to 1.9 per cent (Ghana Statistical Service, 2005).

Diversifying crops was a successful way to increase income. Harvest values had more significant positive correlation with the number of crops grown (R²=0.401) than with the cultivated acreage (R²=0.255). Irrigation also increases the net revenue per land unit (Kurukulasuriya et al., 2006). Many respondents wanted to irrigate, but did not have the means.

Annual income and perceived means to adapt positively correlated (R²=0.216). Respondents considered adaption to CRHs very cost‐effective (average 4.6). However, as 66 per cent of them stated that their income decreased in the past five years, the available resources for adaptation are expected to decline.

In Botswana, the respondents' very high perceived vulnerability to droughts (average 5.2) and very high fear for droughts (average 5.0) correlated significantly (R²=0.15). Respondents perceived to be highly climate‐dependent (Figure 1). Inactiveness towards CRHs is not an option according to 89 per cent of the respondents.

The perceived climate‐dependency decreased with increasing cultivated acreage (R²=0.17) and increasing crop diversity (R²=0.31). Yet neither agriculture nor livestock related climate‐dependency correlated significantly with vulnerability, because the results of perceived dependencies were bipolar (Figure 1). This could be explained by the scattered groundwater provision by the government.

The respondents perceived education, changing crop types and varieties, and irrigation as the most effective autonomous measures to adapt to CRHs (Figure 3); and improving water availability, dialogues among water users, insurance and the DRP as the most effective community measures. 11 respondents partly relied on the DRP during the survey. Outmigration and changing profession were not considered successful to adapt to CRHs, yet in practice many respondents already diversified their livelihood. Despite respondents' preferences irrigation is uncommon in the area.

Respondents were only moderately motivated to adapt to CRHs (average 3.0). This probably relates to the all significantly correlating perceptions of little means to adapt (average 2.4), moderate ability to recover (2.9) and low successfulness in adapting (2.3). The low perceived experience (average 1.9) also correlated significantly to the perceived ability to recover (R²=0.197) and perceived present and future successfulness in adaptation (R²=0.142 and R²=0.095, respectively)^[1]. The moderate motivation could also be explained by the acceptance of droughts as the norm (Slegers, 2008) and the reliance on the government.

Wealthier farmers are better able to adapt. Both the perceived present‐day successfulness in adaptation and the recovery capacity increase slowly with an increasing means to adapt (R²=0.07 and R²=0.08, respectively).

Both the respondents in Botswana and Ghana considered droughts as the most important CRH, perceived to be highly to very highly vulnerable to CRHs and climate‐dependent for their livelihood. A combination of the largely complementary datasets is therefore valuable, despite many cultural, climatic, economic and social differences between the case studies. Unlike the data of the individual countries, size of the full dataset (n=227) allowed for a χ²‐test on Likert scale questions. Some key factors were used that influence respondents' adaptation to CRHs, including the fear for future droughts and the means and motivation to adapt (Deressa et al., 2009; Kreibich, 2011; Plapp and Werner, 2006).

Adaptation to CRHs is nothing new (Adger et al., 2003) and has been successful in the past (Nyong et al., 2007). Therefore, the intention was neither to discover new adaptive measures, nor to identify their successfulness. A χ²‐test was conducted to identify how respondents' perception of the effectiveness of individual adaptive measures relates to the overall adaptation to CRHs. Although education is perceived as an effective adaptive measure (Figure 3) it is at the same time influencing farmer adaptation indirectly (Maddison, 2007) and therefore not included in the analysis.

A Bonferroni correction was applied to counteract the problem of multiple comparisons. After the correction (df=45) a p‐value of p≤2.2×10⁻⁴ indicates a highly significant relation at α=0.01 level between a key factor in the perception of adaptation and an adaptive measure. In this case no results are only significant at α=0.05 level (corresponding to p≤1.1×10⁻³). The χ²‐tests show no relation between either perceived overall successfulness in adaptation or recovery ability with any of the individual adaptive measures (Table I). There is for instance no relation between perceived effectiveness of changing planting dates and overall successfulness of adaptation (p=3.3×10⁻³). Means and motivation were only associated with one adaptive measure (“changing planting dates” (p=7.8×10⁻⁶) and “migration” (p=6.3×10⁻⁶), respectively). The factors “dependency on agriculture”, “fear”, “vulnerability to CRHs”, and “experience in adaptation” were often related to the perceived successfulness of on‐farm adaptation measures (changing crops and changing planting dates) and hardly with little‐practiced irrigation and off‐farm adaptation.

A second χ²‐test was conducted on the relation between the nine key perception factors. After a Bonferroni correction (df=36), 20 tests were highly significant at α=0.01 level (corresponding to p≤2.7×10⁻⁴) and two more were significant at α=0.05 level (p≤1.4×10⁻³) (Table II). The perceived means and climate‐dependency are key for adaptation to CRHs: the former is highly significantly related to six factors (p≤2.7×10⁻⁴) and significantly related to one factor (p≤1.4×10⁻³); the latter is highly significantly related to seven factors (p≤2.7×10⁻⁴). Perceived means is only unrelated to the perceived drought probability. So even though “perceived means” showed association with the perceived successfulness of only one individual adaptive measure (Table I), it still plays a key role in the overall perception of adaptation to CRHs. Regression analysis indeed showed that respondents' motivation increased with increasing perceived means (R²=0.042) and that increasing means lead to increased perceived successfulness in dealing with CRHs. In a second order polynomial trend line (R²=0.117), this trend flattened to “moderately successful” (Figure 4).

Respondents with a very low and a very high perceived climate‐dependency were least motivated to adapt to CRHs. Motivation decreased with increasing perceived vulnerability. “Experience” almost linearly increased motivation to adapt (Figure 5) and was highly significantly related (p≤2.7×10⁻⁴) to five factors but not to the respondents' perceived vulnerability or dependency (Table II). In rural farmer communities, Very experienced farmers can still be vulnerable and climate‐dependent, and inexperienced people can be little vulnerable and hardly dependent on the climate for their livelihood. This adheres to the “complex environment” discourse (Section 2).

Researching subsistence farmers' perception of and adaptation to CRHs offers insight in the circumstances under which they can adapt to climate change. It shows the vulnerability and climate dependency of subsistence farmers, and their desperation to maintain their livelihood in agriculture. This study is based on only 227 interviews in Ghana and Botswana and thus has limited inductive validity. However, three outcomes could be interesting particularly for the international adaptation finance community.

First, both case studies show an analogy between:

• respondents' perception of CRHs and existing environmental degradation; and

• respondents' measures to adapt to CRHs and the way environmental and economic problems are addressed in national development plans and policies.

Respondents react to natural resource‐based development constraints (e.g. deforestation, drought impacts, soil erosion) through measures that can be considered as adaptation to CRHs (e.g. crop diversification, irrigation, outmigration). Therefore, the international climate finance community should build on existing development plans and policies in developing countries that already address climate‐related environmental problems.

Second, the IPCC states that “African farmers have developed several adaptation options […] but such adaptations may not be sufficient for future changes of climate” (Boko et al., 2007). This paper concludes that it is not the “options” (here: “measures”) that are most important for subsistence farmer adaptation to CRHs. The χ²‐test showed no association between the perceived overall effectiveness in adaptation, and the effectiveness of individual adaptive measures. Two examples are outmigration and irrigation. Outmigration happened in both countries, although respondents perceived it as very ineffective to adapt to CRHs. Irrigation was perceived as very effective. However, budget constraints prevent respondents from applying it. The perceived successfulness of adaptation is more dependent on the perception of experience (R²=0.165), recovery (R²=0.172) and means to adapt to CRHs (R²=0.101). Based on the χ²‐tests and in line with Nyanga et al. (2011), this paper concludes that adaptation is more than the sum of adaptive measures. Financing adaptation thus requires more than financing adaptive measures. For example, this paper shows that increasing farmers' experience and means increases their motivation to adapt to CRHs. “Soft” adaptation (e.g. knowledge sharing, building institutions) as well as tackling development constraints that prevent successful adaptation (e.g. poverty and limited water availability) should be integrated in adaptation finance (Denton, 2010).

Finally, the respondents' motivation to adapt to CRHs decreased with increasing perceived vulnerability. In combination with the argument that adaptation measures of African farmers may not be sufficient for future climatic changes (Boko et al., 2007), the message to international climate finance institutions is that ex post adaptation (incl. disaster relief) will remain important.

The research was part of the ADAPTS project (www.adapts.nl), funded by The Netherlands Ministry of Foreign Affairs. The author would like to thank the project partners in Ghana and Botswana for their devoted cooperation. Special thanks go to the students for their excellent fieldwork in Ghana (Jaap Rozema) and Botswana (Linda Bogaert and Florian Expert). The author would like to acknowledge the comments and suggestions of Dr Susan van ‘t Klooster, Prof. Dr Guus Borger, Isabel van de Sand and two anonymous reviewers. All remaining errors are the fault of the author.

Pieter Pauw is a researcher at the Deutsches Institut für Entwicklungspolitik/German Development Institute (DIE‐GDI) in Bonn, Germany. His work focuses on institutional aspects of climate change adaptation finance. Previously, Pieter Pauw worked at the Institute for Environmental Studies (IVM) in Amsterdam, The Netherlands, on a variety of climate change adaptation projects in The Netherlands, Europe and developing countries. Pieter Pauw can be contacted at: pieter.pauw@die‐gdi.de