INTRODUCTION

Agriculture is an economic activity highly susceptible to climate variability and climate change (CC) (^{Jawid, 2020}). Changes in temperature, rainfall patterns and extreme weather events, such as floods and droughts, are having negative effects on food production and food security, with small producers being the most affected (^{Jawid, 2020}; ^{Malhi et al., 2021}). Any impact on the livelihoods of these farmers to produce food would significantly affect the rural economy and the food security of the population (^{Estrada et al., 2022}).

Small producers fundamentally depend on agriculture as their main source of income and subsistence (^{Toledo and Barrera-Bassols, 2017}). They often do not have access to resources and technologies needed to adapt to changing climatic conditions, making them particularly vulnerable to the effects of CC (^{Nor Diana et al., 2022}).

In Mexico, this type of agriculture is prevalent and constitutes 81.3% of the rural economic units in the country (^{Ramirez-Juárez, 2022}). This sector plays an essential role in the welfare of the rural population, where the majority of small farmers depend on this activity (^{Estrada et al., 2022}).

So, it is necessary to address the potential impact of CC in small producers’ activity to dimension the environmental and economic risks that CC represents in those vulnerable agroecosystems in order to generate strategies and public policies for sustainable development focused on adaptation and mitigation (^{Guerrero-Carrera et al., 2020}; ^{Raihan and Tuspekova 2022}).

Coffee farming is a good case study because it represents an example that encompasses the aforementioned problems (^{Guerrero-Carrera et al., 2020}). Coffee production is one of the agricultural activities most threatened by CC (^{Campbell, 2021}). Mexico is one of the ten main coffee-growing countries worldwide, and over half a million coffee growers depend on this activity in the country (^{Gumencindo-Alejo et al., 2021}). However, in recent decades, coffee growers have faced various problems in Mexico, such as low income due to the low profitability of coffee caused by international price fluctuations, which often result in a decline in prices (^{Avalos et al., 2023}). Moreover, in 2012, coffee rust damaged vast swaths of Mexican plantations, forcing replanting (^{Henderson, 2020}).

Globally, studies suggest a progressive CC impact on coffee production, which will affect from 30 to 90% of all coffee-growing areas (^{Läderach et al., 2017}; ^{Imbach et al., 2017}). Furthermore, CC facilitates the spread of coffee pests and diseases (^{Avalos et al., 2023}).

Then, it is necessary to understand how climate variations are affect production, profitability and land value (LV) in coffee farming, as well as to identify the adaptation measures that are implemented to reduce the negative impacts of CC (^{Mora et al., 2010}; ^{Guerrero-Carrera et al., 2020}). Because of the vulnerability of coffee growers, such research is needed to provide them with information to anticipate CC effects and to design local strategies, programs, and policies for CC adaptation and mitigation accordingly (^{IPCC 2007}; ^{Craparo et al., 2015}). Similarly, studies with an economic approach (such as the Ricardian model) are particularly relevant to assessing the potential impact of CC on agriculture (^{Mora et al., 2010}; ^{Jawid, 2020})

THEORETICAL FRAMEWORK

For last two decades, great interest has been expressed in analyzing and measuring the effects of CC in agriculture. In these efforts, several models are reported in the literature, ranging from agronomic, agroclimatic, and economic approaches (^{López and Hernández, 2016}). Within the latter are those that use a general equilibrium model, the production function model, and the Ricardian approach (^{Mora et al., 2010}).

The agronomic approach simulates crop growth during the plant life cycle and measures the effect of changed climatic conditions on crop yield and input requirements (^{Schlenker et al., 2006}). Agronomic models provide evidence that technological solutions would increase yields but they fall short of justifying whether farmers will actually choose these techniques or even whether these strategies would be economically beneficial responses (^{Reilly, 1999}). An important disadvantage is that economic aspects of the problem are not considered (^{Schlenker et al., 2006}).

Agroclimatic models use base scenarios and climate projections to estimate the effects of CC on crop production. According to ^{Candelaria et al. (2011)}, these models are used to simulate sustainability, under a holistic and systemic vision. Another important evolution has been to consider the producer as a subject who participates in modeling. ^{Adams et al. (1990)} carried out the first study to analyze the effects of CC on agriculture in the United States from an economic point of view. The results showed that CC leads to a slight reduction in the total cultivated area in the United States. The Computable General Equilibrium (CGE) approach models agriculture in relation to the other major sectors of the economy and allows resources to move between sectors in response to economic incentives.

In summary, agronomic models do not fully capture farmers’ adaptation and mitigation strategies to face CC and, on the other hand, CGE models are only appropriate for highly aggregated sectors of the economy (^{Schlenker et al., 2006}).

The production function approach estimates the impacts of CC on agriculture by varying one or a few input variables, such as temperature, precipitation, and carbon dioxide levels (^{Gay et al., 2006}). Numerous studies employing this methodology have predicted severe yield reductions for various crops across different regions as a consequence of global warming (^{Lobell et al., 2011}). However, these studies fail to consider the adaptive measures that farmers can implement in response to changing environmental conditions. By not accounting for these potential adaptations, previous studies may have overestimated the extent of damage caused by environmental changes (^{Mendelsohn and Dinar, 2009}).

^{Mendelsohn et al. (1994)} developed the innovative Ricardian approach, which corrects the bias in the production function technique by using cross-sectional economic data on LV. This approach examines how climate affects net income, allowing for a more comprehensive analysis of the effect of climatic and non-climatic variables on LV and farm income. Despite the emergence of new models and techniques in the years following its introduction, the Ricardian approach remains a valuable and widely used tool in research on CC and agriculture due to its ability to capture farmers’ adaptation and provide more realistic estimates of long-term economic impacts.

The Ricardian approach allowed for substitution in land uses. Therefore, when the CC, a piece of land becomes unsuitable for typical agricultural uses, but is still suitable for other valuable, non-agricultural uses, this will be reflected in the price of the land and should be picked up in the regression hedonic, which can be estimated using ordinary least squares (OLS) (^{Schlenker and Roberts, 2009}). In this case, an agronomically focused analysis that looked closely at agricultural uses would exaggerate the value loss induced by CC, while the Ricardian approach would correctly identify that substitution of appropriate non-agricultural uses reduced the value loss. Thus, the model reduces estimation biases; however, it does not take into account the social, economic, and environmental implications of such a change in land use (^{Schlenker et al., 2006}).

One aspect that makes the Ricardian model very attractive is that it takes adaptation into account, while the others do not. We know theoretically that adaptation reduces harm. The Ricardian approach finds less harm to American agriculture than the other approaches. This is empirical evidence of the power of adaptation (^{Schlenker et al., 2006}).

The Ricardian approach has been widely applied in various regions of the world to assess the impact of climate change on agriculture. For example, ^{Kurukulasuriya and Mendelsohn (2007)} applied this method in several African countries, finding that lowland farms are more sensitive to climate change than highland farms. In Asia, ^{Jawid (2020)} used the approach in Afghanistan, revealing significant impacts on rainfed agriculture. These studies demonstrate the versatility of the Ricardian method in different geographical and agricultural contexts.

Although the Ricardian approach has disadvantages and weaknesses like other approaches that evaluate the impact of CC on agricultural production, this model has characteristics of ease of estimation and greater possibilities of obtaining the required data, considering that this model can capture the potential adaptation measures that producers could incorporate in their production units due to the climatic conditions that could be in place.

Among the methodological improvements in the present study, the following aspects were included: a) a technological level index of the production units was constructed and incorporated as control variables in the specifications, b) this model jointly integrated the agricultural and non-agricultural incomes of the production units and their influence on LV. Additionally, c) georeferencing techniques of the production units were used in this study to obtain more precise locations and achieve more accurate climatic values in the developed geospatial interpolations. These contributions enhance the understanding of the potential impact of CC on coffee production.

This study tests two hypothesis:

Hypothesis (H1): CC has a negative economic impact on coffee production in Cuetzalan and Mazateca region in México.

Hypothesis (H2): Small coffee farmers are more vulnerable to CC than large coffee commercial farms by comparing the climate sensitivity of small and large farms in Mexico.

METHODS

This section addresses the specifications of the Ricardian model, the study area, sociodemographic data, climate and soil variables, as well as the climate scenarios and statistical analysis.

Study area

This study was conducted in the Mazateca region of Oaxaca and in the Cuetzalan Puebla region (Figure 1). These regions strongly depend on coffee-growing because they are two of the main Mexican coffee-producing regions (^{Moguel and Toledo, 1999}), and they are highly vulnerable to CC (^{Monterroso et al., 2014}).

Figure 1 Location of Mazateca region of Oaxaca and Cuetzalan Puebla. 

RESULTS

The average LV per hectare of coffee-growing area was $42,557 for the Mazateca region and $83,874 for the Cuetzalan region. The mean annual temperature was 19.1 °C for all PUs, 17.5 °C for the Mazateca region and 20.81 °C for the Cuetzalan region. The mean annual precipitation was 2,9930 mm for the entire sample, 3,301 mm for the Mazateca region, and 2,633 mm for the Cuetzalan region. The plot size also differed between regions, averaging 1.07 ha in the Mazateca region, and 1.46 ha in the Cuetzalan region.

Regarding soil types, differences were observed between the study regions. In the Mazateca region, Alisol (18%) and Leptosol (82%) soils predominate, while in the Cuetzalan region, eutric vertisol is the dominant soil type (88%). In the overall sample, which includes both regions, the means for Alisol, Leptosol, and Vertisol_e are 0.09, 0.44, and 0.41, respectively, reflecting the contribution of each region to the total distribution of soil types.

The overall mean altitude of the plots was 1,066 masl, with 1,324 masl for the Mazateca region and 764 masl for the Cuetzalan region. The farmers’ mean age was 57 years for all PUs, 54 years for the Mazateca region and 60 years for the Cuetzalan region. The mean education of the household head was 5 years for the whole sample, 4.8 years for the Mazateca region and 5.3 years for the Cuetzalan region. The farmers’ coffee-growing experience was 25.9 years overall, 23.9 years for the Mazateca region and 28.2 years for the Cuetzalan region.

The specification results for the Mazateca region showed that the temperature variables in spring and autumn, as well as the precipitation variables in spring and winter, have a significant impact on the coffee LV (Table 2). Furthermore, it was found that other variables, such as soil type, age and experience of the head of household, their educational level, the presence of a computer in the home, and the technological index, also significantly influence the land value in this region.

On the other hand, in the Cuetzalan region, the specification results indicated that the precipitation variables in summer and autumn do not have a significantly different effect on the coffee LV. However, it was found that other variables do have a significant influence on the land value, such as altitude (masl), land area, age and gender of the head of household, and household income.

The absence of values for certain variables in the specifications is due to the use of model selection techniques, such as stepwise regression and best subset selection. These techniques identify the most relevant and significant variables that explain the variation in the coffee LV while minimizing multicollinearity and improving model parsimony. Variables that do not meet the predetermined statistical criteria are omitted in the final specification, ensuring that only essential and significant variables are included. The evaluation of the model shows normality of the error term, no heteroscedasticity, no collinearity, and no omitted variables.

By employing equation 8, the optimal values for the significant climatic variables (temperature and precipitation) were determined. It was found that an average spring temperature below 19.9°C is detrimental to LV in the Mazatec region, indicating that temperatures above this range are beneficial for profitability (Table 3).

In the Cuetzalan region, a different pattern was observed. An average spring temperature below 22.2°C was beneficial for LV. Similarly, for summer, an average temperature above 25.4°C negatively impacted LV, indicating that high temperatures in spring and summer could have unfavorable consequences for the coffee crop.

Regarding precipitation, in the Mazatec region, it was found that in spring, precipitation above 297 mm benefited LV. This indicates that an optimal water level during this season is favorable for coffee cultivation. On the other hand, precipitation above 266 mm in winter had a negative impact on LV.

In the Cuetzalan region, it was found that precipitation above 404 mm in spring negatively affected LV. As for winter, precipitation above 262 mm was beneficial for LV in this region, as water deficit during this period could affect coffee vegetative growth and timely flower development.

These results demonstrate that climatic requirements in coffee production can significantly impact land profitability according to each region. This is related to the crop type and specific local climatic characteristics of each production region.

In the Cuetzalan region, it was observed that production units (PUs) located at an altitude higher than 937 meters above sea level experienced an increase in LV.

Access to technology had a positive effect on coffee LV in the Mazateca region. Producer age showed a differentiated effect between regions: while in the Cuetzalan region older producers were associated with a higher LV, suggesting better adaptation to local conditions, in the Mazateca region greater producer age had a negative impact on LV. Finally, higher per cápita income evidenced a positive effect on coffee LV in both regions, although it was only statistically significant in the Cuetzalan region.

The marginal impacts of CC were striking in the Cuetzalan region but low in the Mazateca region. The temperature margins were -45,000 pesos (-54%) per degree centigrade for PUs in the Cuetzalan region. In the Mazateca region, by contrast, the temperature effects increase the LV by 1,203 pesos (2.83%) per degree centigrade. The losses of LV in the Cuetzalan region are associated with the autumn and winter temperatures. Given the decreased precipitation, the LV will decrease 0.91% per mm precipitation in the Cuetzalan region and increase 0.52% per mm precipitation in the Mazatec region (Table 4).

DISCUSSION

Regional contrasts reflected in socioeconomic indicators are a result of environmental conditions, historical trajectories, communication dynamics, and social contexts specific to each region (^{Rahn et al., 2018}; ^{Donatti et al., 2019}). These differences largely determine the differentiated impact of CC on LV in the Cuetzalan and Mazateca regions. These findings are consistent with previous studies that have analyzed the impacts of CC on Mexican agriculture using the Ricardian approach. Similar to ^{Galindo et al. (2015)} and ^{Mendelsohn et al. (2010)}, differentiated impacts are found across regions and types of production units (PUs), with southern regions and small-scale producers being the most affected. The magnitudes of the projected impacts are comparable to those reported by ^{Mendelsohn et al. (2010)}, who estimate losses between 42 and 54% of LV by 2100. However, as ^{Arellano (2018)} points out, the use of market values for land in the context of incomplete markets may underestimate the relationship between climate and agricultural productivity, suggesting that the impacts could be even greater.

Temperature affects LV more than precipitation due to the narrow optimal ranges for key physiological processes such as flowering, fruit development, and the impact of pests and diseases (^{Camargo, 2010}; ^{Guerrero-Carrera et al., 2020}). Cuetzalan is likely to face temperatures above these maximum thresholds more frequently, exacerbated by the associated water deficit, negatively impacting coffee plantation LV even in the short term (^{Thioune et al., 2020}).

The severe and even total reduction of LV in Cuetzalan towards the medium and long term is consistent with previous projections of extreme damage in areas unsuitable for coffee under warming scenarios in Latin America (^{Läderach et al., 2017}; ^{Imbach et al., 2017}). Considering the narrow ecophysiological thresholds documented for the optimal development of arabica coffee, it is highly likely that irreversible limits will be exceeded, beyond which incremental adaptation will not be sufficient, compromising the sustainability of coffee-based livelihoods in this region (^{Villers et al., 2009}; ^{Harvey et al., 2018}).

Our results on the impact of climate change on coffee production in Mexico are consistent with studies conducted in other producing regions. For example, ^{Ovalle-Rivera et al., (2015)} projected a significant decrease in suitable areas for coffee cultivation in Tanzania and Vietnam by 2050. Similarly, ^{Bunn et al,. (2015)} found that climate change could reduce the suitable area for coffee cultivation in Ethiopia by half. These studies, like ours, underscore the vulnerability of small producers and the need for region-specific adaptation strategies.

Additionally, ^{López-Feldman and Mora-Rivera, (2019)} highlight the distributive implications of CC, finding significant increases in poverty and inequality, particularly in the southeast. This is relevant given the context of marginalization of many coffee producers and suggests that the estimated impacts could exacerbate pre-existing socioeconomic vulnerabilities, as is the case in the Mazateca region. Moreover, impacts are expected to be more acute in small PUs, with the potential to worsen due to inequalities, prices, market conditions, high productive dependence, and lack of specific policies (^{Rahn et al., 2018}; ^{Campbell, 2021}). Thus, the adaptive capacity of small-scale producers could be seriously compromised due to the potential deterioration of LV, difficulties in accessing financing, lack of adequate technologies, scarcity of technical education, resource limitations, and weak institutional support in climate risk management by public policies (^{Donatti et al., 2019}; ^{Avelino et al., 2015}).

When designing adaptation strategies, it is crucial to consider the relationship between adaptation and resilience. Although transitioning to other productive activities could be an option in extreme cases where coffee cultivation becomes unviable, this change can have profound implications for the resilience of households and communities that depend on coffee, eroding the social capital, traditional knowledge, and cultural identity associated with this activity (^{Bacon et al., 2014}; ^{Campbell, 2021}). Therefore, adaptation policies should seek to strengthen the resilience of coffee systems, promoting strategies that allow coffee-based livelihoods to be maintained as much as possible, including climate-resilient agricultural practices, productive diversification, and reduction of structural vulnerabilities that limit the adaptive capacity of households (^{Campbell, 2021}). A holistic approach that integrates adaptation and resilience will be key to facing the challenges of CC in the coffee sector, particularly for small-scale producers in highly vulnerable regions such as the Mazateca region and the Cuetzalan region.

CONCLUSIONS

This study provides a rigorous assessment of the potential impacts of climate change on coffee production in the Mazateca and Cuetzalan regions of Mexico using the Ricardian approach. The study integrates climatic, geographic, and socioeconomic variables in a robust statistical analysis to evaluate their interaction with land value. The research contributes to the existing literature by demonstrating the differential effects of climate change on land value according to region and farm size, highlighting the vulnerability of small-scale coffee producers, especially in the Cuetzalan region. The results warn that changes in temperature and precipitation can have significant and potentially irreversible impacts on coffee production, with damages reaching 100% of land value in some scenarios.

A crucial aspect to consider is the adaptive measures that coffee producers could adopt in the face of climate change impacts. Some producers may choose to completely abandon coffee cultivation and switch to other productive activities, while others may decide to switch to more resistant coffee varieties or diversify their crops. The choice of adaptive measures will depend on the economic, social, and ecological resilience of coffee-growing households, as well as the implementation of effective and timely public policies that support this activity.

This study opens avenues for future research, such as the exploration of databases that integrate additional climatic variables relevant to coffee production and the expansion of the geographic scope of the study. The results have significant implications for public policy, suggesting a systemic approach that reduces the vulnerability of producers and promotes sustainable management practices.