1. Introduction
Climate change is one of the tangible and increasingly present effects of global warming. According to the Intergovernmental Panel on Climate Change (IPCC; Masson-Delmotte et al. 2019), human activities have caused global warming, with an estimate of approximately 1.0°C with respect to preindustrial levels and a probable range of 0.8°–1.2°C. Based on climate model simulations, global warming is likely to reach 1.5°C between 2030 and 2052 if it continues to increase at the current rate.
One of the regions that has suffered the greatest effects of climate change is the Latin American region. In particular, in the year 2021, the temperature was above the 1981–2010 average in all subregions, with the maximum anomaly of +0.59 (±0.1°C) recorded in the region of Mexico and Central America, which corresponds to +0.97 (±0.1°C) above the World Meteorological Organization (WMO 2022) reference period of 1961–90 for climate change.
Among the main effects identified by the WMO (2022) in the Latin American region are the increasingly generalized droughts that have had direct impacts on crop yields and food production and a significant impact on shipping routes. The Caribbean subregion is where the greatest effects are found with the greatest precipitation deficit, since it is located in the most water-stressed areas of the world. By contrast, heavy rains in the Central American area have caused landslides and flash floods in rural and urban areas of central and South America.
The presence of fires in the Amazon has caused significant impacts as 2020 surpassed 2019 and became the most active fire year in the southern Amazon. Additionally, this adds to the deforestation that has been experienced in the last 4 years, particularly in the Amazon River basin, which extends through nine South American countries and stores 10% of the global carbon. Another aspect highlighted by WMO (2022) is the increase in the number of tropical cyclones in much of Mexico and South America, as in 2020, there was an unprecedented record of 30 such storms in the Atlantic basin. Hurricanes in particular have affected more than 8 million people in Central America. Sea level rise in the region is also above the world average. With an average of 3.6 mm yr−1 between 1993 and 2020, sea level in the Caribbean has risen at a higher rate than the world average, which was 3.3 mm yr−1. Sea level rise, particularly in Latin America and the Caribbean, is an important variable in the analysis of the quality of life of inhabitants, since more than 27% of the population lives in coastal areas, and it is estimated that between 6% and 8% live in areas with a high or very high risk of being affected by coastal hazards (WMO 2022). In addition to the increase in atmospheric temperature, there is evidence that the sea surface temperature in the North Atlantic Ocean was significantly higher than normal throughout the year 2022. In the Caribbean, 2020 was the year with the largest sea surface temperature changes ever recorded. Beginning in May 2020, sea surface temperatures began to gradually cool in the equatorial Pacific and La Niña developed. This, together with the temperature increase in the Atlantic, contributed to a more active hurricane season than normal (WMO 2022). Finally, it should be noted that extreme weather events affected more than 8 million people in Central America, aggravating food shortages in countries that had already been paralyzed by economic crises, COVID-19 restrictions, and conflicts.
Taking into account the importance and impact that climate change has on the quality of life of the inhabitants, flora, fauna, and the ecosystem of the Latin American region, this paper aims to study the different temporal patterns for the maximum, minimum, and average climatic temperatures of the different countries of Latin America and the Caribbean, using fractional integration as the main methodology (Robinson 1994). Annual data from 1901 to 2021 are employed with the purpose of identifying whether or not there is the presence of significant trends.
One of the main characteristics of climate time series is the presence of nonstationary patterns, which implies changes in the behavior of statistical properties that fluctuate over time. Therefore, one of the advantages of fractional integration techniques is the possibility of working with nonstationary time series in a more effective way than traditional methods, since they allow modeling both short- and long-term persistence of the analyzed series, with mean-reverting behavior, as well as analyzing changing trends over time.
Another necessary aspect in the analysis of climate series is the presence of complex patterns in the long term; however, the main characteristic of fractional integration is to capture this type of extensive pattern essential for understanding and modeling climate phenomena. It is also interesting to point out the advantages of the application of fractional integration methods according to their capacity to handle extreme phenomena, the accuracy of long-term predictions of the series, their ability to model gradual changes more effectively than traditional methods, and, finally, their capacity to adapt to irregular data, since the modeling is robust, which implies that it can adapt to the complexities in the collection of climate data.
In technical terms, one of the main characteristics of fractional integration is that it allows the number of differences in the integration order of the series to generate fractional values, which allows greater flexibility than other methods that only allow for integer degrees of differentiation (Dickey and Fuller 1979; Phillips and Perron 1988; Elliott et al. 1996). Moreover, a positive degree of integration supports the long-memory hypothesis that is claimed by many authors as a statistical feature of climatological data.
Considering the amplitude of the time series under examination, it is feasible to find the presence of long-memory behavior, which generates a better coupling of fractional integration techniques. Therefore, the proposed methodology would seem to be appropriate for determining the nature of the shocks, being transitory if the order of integration is less than 1 or permanent if it is equal to or greater than 1.
2. Literature review
Although there is vast research on climate change in terms of applied statistics, there is no consensus regarding the deterministic versus stochastic nature of the trend in climatological data. Some studies, like Bloomfield and Nychka (1992) and Zheng and Basher (1999), suggest a significant deterministic trend. Conversely, Woodward and Gray (1993, 1995), Stern and Kaufmann (2000), and several others argue for stochastic terms in climate data, often employing unit root models. Additionally, some researchers propose representing climate change effects with long memory or fractionally integrated models, as seen in the works of Bloomfield (1992), Koscielny-Bunde et al. (1998), Gil-Alana (2023, 2005, 2008a,b, 2015), Claudio-Quiroga and Gil-Alana (2022), and others.
We should also mention studies using frequency domain semiparametric methods, such as Mangat and Reschenhofer (2020). Recently, Dagsvik et al. (2020) analyze temperature data from 96 selected weather stations worldwide and reconstructed Northern Hemisphere temperature data over the last two millennia. Using a nonparametric test, it was found that the stationarity hypothesis is not rejected by the data. Subsequently, other properties of the data were investigated using the fractional Gaussian noise (FGN) model. The investigation presents evidence that the FGN process is a good representation of the temperature process. The key feature of the FGN model mentioned in the research is its long-term dependence, which implies that the temperature series can show long trends and cycles. They reach similar conclusions in that the temperature of the planet presents long-memory patterns. However, it is necessary to point out that they differ in the memory capacity of the model since a considerable degree of memory is obtained in this work, including nonstationary characteristics and lack of reversion to the mean.
Studies focused on a global scale include Gil-Alana (2005), which is based on a fractional integration analysis of the temperatures in the Northern Hemisphere. Using seasonally adjusted monthly data from 1854 to 1999, the results of that research highlight the long memory as well as highly persistent behavior. Subsequently, Gil-Alana (2008b) investigated global and hemispheric temperature anomalies again using fractional integration techniques; however, this time using segmented trends, both with the parametric approach and with the nonparametric approach of Bloomfield (1973). In this case, the results confirm that the series have long memory with statistically significant time trends, but the results differ depending on the series considered and the model estimated. When a nonparametric approach is used assuming the existence of a single break in the data, the results show that the regime shift takes place in 1964 for global and Southern Hemisphere temperatures and in 1971 for the Northern Hemisphere. The above translates into an increase of 2.37°C (100 yr)−1 in the Northern Hemisphere and 1.30°C (100 yr)−1 in the southern temperatures, after the date of rupture. If breaks are considered, the results range from 0.46° to 0.51°C over the last 100 years.
The results agree with the sea surface temperature increase observed in Gil-Alana (2015). The interpretation of structural breaks suggests that being able to identify a single break or multiple breaks would indicate that the climate may not only change gradually but may also present changes abruptly (Hare and Mantua 2000; Alley et al. 2003; Ivanov and Evtimov 2010; Coggin 2012; Paul et al. 2014; Adedoyin et al. 2020; Yao and Zhao 2022; Ericsson et al. 2022).
Another group of research is linked to local scales, analyzing the effects of climate variability always taking into account the properties of long memory of the series: Franzke (2012) for daily temperatures at four stations: central England, Stockholm (Sweden), Faraday–Vernadsky (Antarctic Peninsula), and Alert (Canada); Percival et al. (2001) and Gil-Alana (2012) for monthly temperatures in Alaska; Kane and Usman (2013) for the Sokoto metropolis in Nigeria; or Franzke (2010), Bunde et al. (2014), and Ludescher et al. (2016) for monthly temperatures from different Antarctic stations. Gil-Alana (2003) studied the monthly temperature series of central England (CET), while Caballero et al. (2002) studied the multidecadal daily time series of CET, Chicago, and Los Angeles.
Gil-Alana (2008a) demonstrates that the northern, southern, and world temperature anomaly series are fractionally integrated and that the three series exhibit an increase in warming effects after the breaks. He uses fractional integration but with segmented time trends. Moreover, Gil-Alana (2008b) demonstrated that the same three series (global, northern, and southern temperatures) are fractionally integrated around the order of 0.5 by using a nonparametric approach.
Gil-Alana (2018) examines temporal trends in global maximum and minimum temperatures—and the differences between them—using annual data from 1895 to 2017. His empirical results again support the global warming hypothesis. Finally, Gil-Alana and Sauci (2019a) investigated the trend time coefficients in the temperature across 48 states of the United States, revealing an increase in temperature anomalies in all except 10 states, and with an increase that is bigger than that found using other conventional techniques. They also discovered long memory in temperature data and increasing warming patterns across Europe in a closely related paper (Gil-Alana and Sauci 2019b). Other recent articles dealing with changes in temperatures include Di Luca et al. (2020), Vicente-Serrano et al. (2020), González-Pérez (2022), and Salnikov et al. (2023).
3. Methodology
In this context, depending on the value of d, different processes can be considered such as
-
antipersistence, if d < 0;
-
short memory or I(0) processes, if d = 0;
-
long-memory covariance stationary, if 0 < d < 0.5;
-
nonstationary and mean-reverting processes, if 0.5 ≤ d < 1;
-
unit roots or I(1) processes, if d = 1; and
-
long-memory patterns after first differentiation, i.e., I(d) with d > 1.
4. Data
The data used in this research come from the Climate Change Knowledge Portal belonging to the World Bank (2023). The Climatic Research Unit Gridded Time Series (CRU TS) is a widely used observational climate dataset. For all land domains with the exception of Antarctica, data are shown on a 0.5° latitude by 0.5° longitude grid. It is obtained through the interpolation of monthly climatic anomalies from vast networks of weather station readings.
The CRU TS version 4.05 gridded dataset is created from observational data and provides quality-controlled temperature and precipitation measurements from hundreds of weather stations worldwide, as well as derivative products including monthly climatologies and lengthy historical climatologies. This dataset was produced by the CRU at the University of East Anglia (UEA; Harris et al. 2020).
The data have an annual frequency and refer to the time period from 1901 to 2021. Tables 1–3 analyze the descriptive results of each country of Latin America and the Caribbean, respectively, for mean, minimum, and maximum temperatures.
Descriptive statistics: mean temperatures (°C).
Descriptive statistics: minimum temperatures (°C).
Descriptive statistics: maximum temperatures (°C).
Table 1 and Fig. 1 show the results according to the average temperatures in Latin American and Caribbean countries. The average value for the region shows that the region has an average temperature of 23.60°C. However, maximum cases within the average temperature evaluation are noticeable, as is the case of Aruba with 28.46°C, in contrast to the minimum average values in Chile with 9.27°C. On the other hand, the behavior of the standard deviation between countries highlights the British Virgin Islands with the greatest variation in the data, followed by Antigua and Barbuda.
Mean temperatures (°C) in Latin America. Source: prepared by the authors based on World Bank (2023).
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
On the other hand, Table 2 and Fig. 2 show the filtered data considering the minimum temperatures that have occurred in the region. Chile is the country that presents minimum temperatures of 4.32°C in contrast to Aruba that presents minimum scenarios around 24.97°C; worthy of mention are the cases of the British Virgin Islands and the Dominican Republic that present the greatest deviations from the mean.
Average minimum temperatures (°C) in Latin America. Source: prepared by the authors based on World Bank (2023).
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
Table 3 and Fig. 3 show the descriptive statistics on maximum temperatures, where there is an average temperature of 28.48°C with maximum values of 32.01°C in Aruba, followed by El Salvador with 30.98°C and Chile with 14.26°C in the minimum classification within the group of maximum temperatures. Also, the countries with the greatest deviation in their data are the British Virgin Islands and Antigua and Barbuda.
Average maximum temperatures (°C) in Latin America. Source: prepared by the authors based on World Bank (2023).
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
Finally, Fig. 4 shows the temperatures and correlograms of the Latin American region grouped as follows: Caribbean, Central America, South America, and separately Mexico. The graph shows that the subregions with the highest temperatures from 1901 to 2021 are the Caribbean and Central America, followed by South America and Mexico.
Temperatures and correlograms in subregions in Latin America from 1901 to 2021. Source: prepared by the authors based on World Bank (2023).
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
5. Empirical results
We display the results for the mean, maximum, and minimum temperatures. Tables 4 and 5 focus on the mean values, while Tables 6 and 7 focus on maximum and minimum values, respectively. Similar tables for the log-transformed values are reported in appendix B.
Estimates of the differencing parameter d under the assumption of autocorrelated (Bloomfield) errors. Series: mean values. The table indicates the estimates of d and the 95% confidence intervals (in parentheses) for the two scenarios of (i) with a constant (in column 2) and (ii) with a constant and a linear time trend (in column 3). In bold is the selected specification for each series in relation to the deterministic terms.
Estimated coefficients of the selected models in Table 4. Series: mean values. The values in column 2 are the estimated d (and 95% confidence bands) on the selected models. In columns 3 and 4, the values are the estimated intercept and time trend coefficients. In parentheses are the corresponding t values.
Estimated values of the differencing parameter d along with the intercept and the time trend, under the assumption of autocorrelated errors. Series: maximum values. The values in column 2 are the estimated orders of integration d (and 95% confidence bands) for each series. In columns 3 and 4, the values are the estimated intercept and time trend coefficients. In parentheses are the corresponding t values.
Estimated values of the differencing parameter d along with the intercept and the time trend, under the assumption of autocorrelated errors. Series: minimum values. The values in column 2 are the estimated orders of integration d (and 95% confidence bands) for each series. In columns 3 and 4, the values are the estimated intercept and time trend coefficients. In parentheses are the corresponding t values.
Bloomfield (1973) demonstrated that the log of the above expression can be well approximated by the log of Eq. (5) when p and q are small values, and thus, it does not require the estimation of such many parameters as in the ARMA models, which always results tedious in terms of estimation, testing, and model specification. In addition, the model was stationary across all its values unlike what happens in the autoregressive (AR) case [see Gil-Alana (2004) for its accommodation in the context of fractional integration].
Taking into consideration the results of Tables 4 and 5 and the map in Fig. 5 referring to the average values of temperatures in Latin America and the Caribbean, we observe that there is evidence of reversion to the mean for all countries, since the estimates of the differencing parameter are significantly below 1 in all cases. This mean-reverting property is particularly noticeable in 15 countries where we cannot reject the null hypothesis of d = 0 or short-memory processes. These countries are Antigua and Barbuda, Aruba, Argentina, Bolivia, the British Virgin Islands, Colombia, the Dominican Republic, Ecuador, Grenada, Haiti, Panama, Trinidad and Tobago, the Turks and Caicos Islands, Uruguay, and Venezuela. Note that for all these countries, the confidence interval reported in the second column in Table 5 includes the value 0. On the other extreme, countries such as Guatemala, and particularly Mexico and Brazil, display the highest degrees of integration with values of d around 0.5.
Differencing parameter “d” under the assumption of autocorrelated errors. Series: mean values. Source: prepared by the authors based on World Bank (2023). The colors of the countries are analyzed from the lowest (blue) to the highest degree of differentiation (red).
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
On the other hand, the countries with the highest magnitude for the time trend are Antigua and Barbuda, Aruba, and the British Virgin Islands, while there are two countries not displaying significant trends: Bolivia and Paraguay.
In the remaining cases, the time trend coefficient is found to be significantly positive, supporting thus the hypothesis of warming temperatures in Latin American and Caribbean countries.
Tables 6 and 7 refer to the maximum and minimum temperatures, respectively. However, instead of presenting the whole battery of results for d depending on the specification of the deterministic terms, we simply focus on the selected models for each country.
Starting with the maximum values (Table 6 and Fig. 6), we observe 17 countries where the short memory or l(0) hypothesis cannot be rejected, and among the 15 countries where long memory is detected, the highest estimates of d correspond to Guatemala and Brazil. The time trend coefficients are significantly positive in all cases except those of Bolivia and Paraguay, and the highest trend coefficients are found in the British Virgin Islands and Antigua and Barbuda.
Differencing parameter d under the assumption of autocorrelated errors. Series: maximum values. Short memory and long memory. Source: prepared by the authors based on World Bank (2023). The graphs have been filtered so that they clearly show, using colors, the countries according to their degree of differentiation. Countries are separated into two panels: (a) countries with short memory and (b) countries with long memory. Gray areas mean that for a given graph these countries are not filtered out.
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
Focusing next on the minimum values (Table 7 and Fig. 7), the values of d are generally higher, and we notice only seven countries where the I(0) hypothesis cannot be rejected. The highest degree of persistence is observed in Brazil and Mexico (0.50) and Jamaica (0.52), and once again, only Bolivia and Paraguay show no evidence of warming temperatures.
Differencing parameter d under the assumption of autocorrelated errors. Series: minimum values. Short memory and long memory. Source: prepared by the authors based on World Bank (2023). The graphs have been filtered so that they clearly show, using colors, the countries according to their degree of differentiation. Countries are separated into two panels: (a) countries with short memory and (b) countries with long memory. Gray areas mean that for a given graph these countries are not filtered out.
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
Finally, in Tables 8 and 9 and in the map in Fig. 8 we examine the range, i.e., the difference between the maximum and minimum temperatures for each country. Once again, we start by reporting the estimates of d under the three potential scenarios in relation to the constant and the time trend. Table 8 focuses on the estimated coefficients, and we observe that short memory occurs now in eight countries (Barbados, Belize, Costa Rica, Dominica, Mexico, Nicaragua, Trinidad and Tobago, and Uruguay), and there is a group of countries with values of d above 0.5. They are Aruba (d = 0.55), Brazil (0.65), Colombia (0.51), Cuba (0.76), Ecuador (0.62), Haiti (0.63), Panama (0.64), the Turks and Caicos Islands (0.64), and Venezuela (0.56). In these cases, shocks in the range will take longer time to disappear than in the rest of the cases.
Estimates of the differencing parameter d under the assumption of autocorrelated (Bloomfield) errors. Series: range (maximum–minimum) values. The table indicates the estimates of d and the 95% confidence intervals (in parentheses) for the two scenarios of (i) with a constant (in column 2) and (ii) with a constant and a linear time trend (in column 3). In bold is the selected specification for each series in relation to the deterministic terms.
Estimated values of the differencing parameter d along with the intercept and the time trend, under the assumption of autocorrelated errors. Series: range (maximum–minimum) values. The values in column 2 are the estimated orders of integration d (and 95% confidence bands) for each series. In columns 3 and 4, the values are the estimated intercept and time trend coefficients. In parentheses are the corresponding t values.
Differencing parameter d under the assumption of autocorrelated errors. Series: range values. Source: prepared by the authors based on World Bank (2023). The graphs have been filtered so that they clearly show, using colors, the countries according to their degree of differentiation. Countries are separated into two panels: (a) countries with short memory and (b) countries with long memory. Gray areas mean that for a given graph these countries are not filtered out.
Citation: Journal of Applied Meteorology and Climatology 63, 10; 10.1175/JAMC-D-23-0141.1
In Table 9, looking at the time trend, this is significantly positive for Mexico and Nicaragua; significantly negative for Chile, Costa Rica, the Dominican Republic, Peru, Suriname, and Uruguay; and insignificant in the rest of the cases.
6. Concluding comments and recommendation
We have examined in this paper the main statistical features of the temperature time series in the Latin American region. In particular, we have been interested in testing the existence of significant positive trends, supporting thus the hypothesis of climate warming. However, instead of relying on classical assumptions that suppose that the errors are well behaved or stationary and integrated of order 0, fractional degrees of differentiation are permitted. In doing so, we permit for long memory if the order of integration is positive, a hypothesis that is supposed to be satisfied by climatological data.
The main findings of the research can be summarized as follows: When taking the average data, there is evidence supporting reversion to the mean in all countries; however, the rate of convergence is different across countries. Thus, there are 15 countries where the hypothesis of short memory or I(0) behavior cannot be rejected, implying a fast process of reversion in the presence of shocks. On the other hand, countries such as Guatemala, Mexico, and Brazil display orders of integration close to 0.5, implying longer processes of convergence to the mean value of the time series. This makes it necessary to reflect on current public policies and their real impact on warming, particularly taking into account that these are the countries where the effect of climate change may last the longest if they do not act promptly and with appropriate evaluation. Particularly in the context of Guatemala, the impact may be even more profound due to its close relationship with agriculture as an important part of subsistence for a large proportion of the population, causing greater socioeconomic vulnerability. Considering the region known as the dry corridor, which covers a large part of Guatemala and therefore the north of Central America, droughts are recurrent, causing crop failures that seriously affect rural livelihoods and increase food insecurity in the region [Food and Agriculture Organization of the United Nations (FAO), 2023]. It is also important to point out the trends underlying the results. A greater positive trend in temperatures is observed in Caribbean islands such as Antigua and Barbuda, Aruba, and the British Virgin Islands. Bolivia and Paraguay are the only countries which do not present a significant positive trend. For the rest of the countries in the Latin American and Caribbean regions, there is a statistically significant positive trend in climate temperatures, which is confirmed by the results of WMO (2022) in which it is highlighted that the warming trend in Latin America and the Caribbean continued in 2021. The average rate of temperature increase was approximately 0.2°C per decade between 1991 and 2021, compared to 0.1°C per decade between 1961 and 1990.
When looking at the maximum and minimum temperature series, starting from the analysis of maximum temperatures, practically 53% of the countries analyzed have a time series cataloged as short-memory or short-duration impact processes; on the other hand, there are about 48% of countries with long-memory processes that can produce and generate a longer recovery time albeit with reversion to the mean. Such are the cases of Guatemala and Brazil in particular. Similarly, to the mean case, when analyzing the time trends, it is highlighted that all Latin American countries except Bolivia and Paraguay have a clearly visible trend underlying their results. According to WMO (2022), there is a practically common trend for the region, with heatwaves and prolonged droughts. In the middle of the year, some areas of Central America were affected by a weak-to-moderate drought. In the Caribbean, several countries experienced some level of moderate drought, especially Haiti, the Dominican Republic, Puerto Rico, and parts of Cuba. The U.S. Virgin Islands had their fourth driest year on record, with the lowest annual rainfall since 1965, and groundwater levels were at historic lows, similar to 2016.
Taking into account the minimum values, the orders of integration are slightly higher though the same conclusion as with the maximum temperatures holds: The evidence of short memory is found in 22% of the countries evaluated compared to 78% of countries with evidence of long memory, with special cases such as Mexico, Brazil, and Jamaica with high degrees of differentiation greater than d ≥ 0.5. Again, as mentioned above only Bolivia and Paraguay present trends which are insignificant.
It is interesting to note that in the Latin American region there have been cold wave scenarios, particularly in the southern part of South America. For example, according to WMO (2022) in Argentina, the province of Catamarca recorded its lowest minimum temperature, −6.2°C (the previous record of −5.8°C was on 15 June 1961). In Brazil, the temperature in Vilhena, in the state of Rondônia, reached 8.2°C (compared to a monthly average of 19.2°C for 1981–2010). In Itatiaia National Park, in the highlands of Rio de Janeiro, a minimum temperature of −9.9°C was reached.
In the final part of the manuscript, we have examined the range, i.e., the difference between the maximum and minimum temperatures for each country. Short memory occurs now in eight countries (Barbados, Belize, Costa Rica, Dominica, Mexico, Nicaragua, Trinidad and Tobago, and Uruguay), and there is a group of nine countries with values of d above 0 (Aruba, Brazil, Colombia, Cuba, Ecuador, Haiti, Panama, the Turks and Caicos Islands, and Venezuela). In these cases, shocks in the range will take longer to disappear than in the remaining cases.
The above findings are complemented by the publication of the WMO (2023), detailing the most recent climatic conditions in the Latin American and Caribbean regions. It reveals an average rising trend of roughly 0.2°C decade−1 between 1991 and 2022 (greater in Mexico and the Caribbean) since 1900 in the Latin America and the Caribbean (LAC) region (compared to the previous 30-yr periods 1900–30, 1931–60, and 1961–90). Of the four subregions, Mexico experienced the greatest degree of warming, almost 0.3°C decade−1, in the period 1991–2022, particularly in central and eastern Mexico and in the Yucatan Peninsula. Temperature anomalies between +1° and +2°C were recorded in Guatemala and El Salvador.
In the Caribbean, positive temperature anomalies of +1°–+2°C were recorded in the Dominican Republic, Puerto Rico, and the small Caribbean islands. In South America, above-normal temperature anomalies of +1°–+1.5°C were observed in eastern Amazonia, the central and southern Andes of Peru, Bolivia, central Chile, and central Argentina.
Exceptionally high temperatures, low air humidity, and severe drought also led to periods of record wildfires in many South American countries. In January and February 2022, Argentina and Paraguay recorded an increase of 283% and 258%, respectively, in the number of hotspots detected compared to the 2001–21 average, and January–March wildfire carbon dioxide (CO2) emissions were the highest in the last 20 years.
Although it is not possible to conclusively summarize the main results while attempting to create a pattern in the behavior of the outcomes, considering that the research involves analyzing averages, minima, and maxima, it is possible to classify Caribbean countries as a common denominator. These countries have the ability to return to the mean value more easily compared to the rest of the countries in Latin America. However, this group of small Caribbean countries must be analyzed with caution, as despite the potential to return to the mean, there are islands that exhibit a marked and high positive trend if not approached with prudence. Particularly, when considering the results regarding the return to the mean in average, maximum, and minimum temperatures, another interesting relationship can be observed regarding the pattern presented by Uruguay, Paraguay, and Bolivia. These countries share characteristics of mean reversion and do not exhibit a positive trend in temperature behavior. This prompts reflection on the demographic composition, considering that they represent countries in the southern cone and have not had a directly significant impact on temperature behavior despite being surrounded by larger countries.
On the other hand, another pattern that can be analyzed is the result that only seven countries exhibit mean reversion in recorded minimum temperatures, which aligns with the analysis of small Caribbean countries. In this context, the main challenge is the positive trend in their temperatures. Curiously, Guatemala can be classified as one of the countries that will be most affected by the increase in temperatures. Despite being a small country, it can be classified with large countries, and therefore, it experiences greater volatility in temperature changes and trends. For example, this is similar to Brazil and Mexico. This inference suggests that the impact of climate change present in southern Mexico is affecting the North American region, particularly dominated by Guatemala.
Finally, the results reported in this work may be helpful as a public policy recommendation, since in Latin American countries the problem of climate change, in addition to the social costs it implies, is predicted to have an effect on the order of U.S. $100 billion by the year 2050 due to decreases in agricultural yields, the disappearance of glaciers, floods, droughts, and other events caused by global warming (Vergara et al. 2013). It is important to keep in mind that climate change affects many aspects of life in Latin America, including the economy due to the reduction in terms of productivity and production growth, agriculture from the perspective of social and environmental vulnerability from cycles of droughts, and floods that result in a decrease in agricultural yields mainly affecting food security. Another fundamental aspect in the analysis of the effects of climate change is health, since increases in temperature may increase the incidence of vector-borne diseases such as dengue fever and malaria.
As detailed by the United Nations Educational, Scientific and Cultural Organization (2016), regional articulation in climate matters is increasingly relevant. On the one hand, collaboration between states that are located in the same region (and that share resources and feel the immediate effects of natural disturbances in a similar way) is a historical necessity, the importance of which grows with the acceleration of climate problems.
As a result, the design of international climate policy means that the success of the measures established at the global level depends on whether and how each state will comply with the commitments it has made. In this scenario, Latin American regional collaboration is particularly important as these states not only share a geographic region but are also developing countries with similar demands and challenges and they are strengthened when they act in concert—both during negotiations and in policy implementation.
Acknowledgments.
Luis A. Gil-Alana gratefully acknowledges financial support from the Grant PID2020-113691RB-I00 funded by MCIN/AEI/10.13039/501100011033. An internal project from the Universidad Francisco de Vitoria is also acknowledged. JEL classification: B22; C01; C22; Q51; Q54. Comments from the editor and three anonymous reviewers are gratefully acknowledged.
Data availability statement.
Data are available upon request. The program used for the computation is based on Fortran, though new codes in R will be also available from the authors upon request.
APPENDIX A
The Test Statistic
APPENDIX B
Results Based on Log Transformations
Tables B1–B4 show the estimated log-transformed values for the mean, maximum, and minimum temperatures.
Estimates of the differencing parameter d under the assumption of autocorrelated (Bloomfield) errors. Series: mean values. The table indicates the estimates of d and the 95% confidence intervals (in parentheses) for the two scenarios of (i) with a constant (in column 2) and (ii) with a constant and a linear time trend (in column 3). In bold is the selected specification for each series in relation to the deterministic terms.
Estimated values of the differencing parameter d along with the intercept and the time trend, under the assumption of autocorrelated errors. Series: mean values. The values in column 2 are the estimated d (and 95% confidence bands) on the selected models. In columns 3 and 4, the values are the estimated intercept and time trend coefficients. In parentheses are the corresponding t values.
Estimated values of the differencing parameter d along with the intercept and the time trend, under the assumption of autocorrelated errors. Series: maximum values. The values in column 2 are the estimated d (and 95% confidence bands) on the selected models. In columns 3 and 4, the values are the estimated intercept and time trend coefficients. In parentheses are the corresponding t values.
Estimated values of the differencing parameter d along with the intercept and the time trend, under the assumption of autocorrelated errors. Series: minimum values. The values in column 2 are the estimated d (and 95% confidence bands) on the selected models. In columns 3 and 4, the values are the estimated intercept and time trend coefficients. In parentheses are the corresponding t values.
REFERENCES
Abbritti, M., L. A. Gil-Alana, Y. Lovcha, and A. Moreno, 2016: Term structure persistence. J. Financ. Econ., 14, 331–352, https://doi.org/10.1093/jjfinec/nbv003.
Abbritti, M., H. Carcel, L. Gil-Alana, and A. Moreno, 2023: Term premium in a fractionally cointegrated yield curve. J. Banking Finance, 149, 106777, https://doi.org/10.1016/j.jbankfin.2023.106777.
Adedoyin, F., I. Ozturk, I. Abubakar, T. Kumeka, O. Folarin, and F. V. Bekun, 2020: Structural breaks in CO2 emissions: Are they caused by climate change protests or other factors? J. Environ. Manage., 266, 110628, https://doi.org/10.1016/j.jenvman.2020.110628.
Alley, R. B., and Coauthors, 2003: Abrupt climate change. Science, 299, 2005–2010, https://doi.org/10.1126/science.1081056.
Baillie, R. T., 1996: Long memory processes and fractional integration in econometrics. J. Econ., 73, 5–59, https://doi.org/10.1016/0304-4076(95)01732-1.
Belbute, J. M., and A. M. Pereira, 2015: An alternative reference scenario for global CO2 emissions from fuel consumption: An ARFIMA approach. Econ. Lett., 136, 108–111, https://doi.org/10.1016/j.econlet.2015.09.001.
Bloomfield, P., 1973: An exponential model for the spectrum of a scalar time series. Biometrika, 60, 217–226, https://doi.org/10.1093/biomet/60.2.217.
Bloomfield, P., 1992: Trends in global temperature. Climatic Change, 21, 1–16, https://doi.org/10.1007/BF00143250.
Bloomfield, P., and D. Nychka, 1992: Climate spectra and detecting climate change. Climatic Change, 21, 275–287, https://doi.org/10.1007/BF00139727.
Bunde, A., J. Ludescher, C. L. E. Franzke, and U. Büntgen, 2014: How significant is West Antarctic warming? Nat. Geosci., 7, 246–247, https://doi.org/10.1038/ngeo2126.
Caballero, R., S. Jewson, and A. Brix, 2002: Long memory in surface air temperature: Detection, modeling, and application to weather derivative valuation. Climate Res., 21, 127–140, https://doi.org/10.3354/cr021127.
Claudio-Quiroga, G., and L. A. Gil-Alana, 2022: CO2 emissions persistence: Evidence using fractional integration. Energy Strategy Rev., 43, 100924, https://doi.org/10.1016/j.esr.2022.100924.
Coggin, T. D., 2012: Using econometric methods to test for trends in the HadCRUT3 global and hemispheric data. Int. J. Climatol., 32, 315–320, https://doi.org/10.1002/joc.2271.
Crato, N., and P. Rothman, 1994: Fractional integration analysis of long-run behavior for US macroeconomic time series. Econ. Lett., 45, 287–291, https://doi.org/10.1016/0165-1765(94)90025-6.
Dagsvik, J. K., M. Fortuna, and S. H. Moen, 2020: How does temperature vary over time?: Evidence on the stationary and fractal nature of temperature fluctuations. J. Roy. Stat. Soc., 183A, 883–908, https://doi.org/10.1111/rssa.12557.
Dickey, D. A., and W. A. Fuller, 1979: Distribution of the estimators for autoregressive time series with a unit root. J. Amer. Stat. Assoc., 74, 427–431, https://doi.org/10.2307/2286348.
Diebold, F. X., and G. D. Rudebusch, 1989: Long memory and persistence in aggregate output. J. Monetary Econ., 24, 189–209, https://doi.org/10.1016/0304-3932(89)90003-2.
Di Luca, A., A. J. Pitman, and R. de Elía, 2020: Decomposing temperature extremes errors in CMIP5 and CMIP6 models. Geophys. Res. Lett., 47, e2020GL088031, https://doi.org/10.1029/2020GL088031.
Elliott, G., T. J. Rothenberg, and J. H. Stock, 1996: Efficient tests for an autoregressive unit root. Econometrica, 64, 813–836, https://doi.org/10.2307/2171846.
Ericsson, N. R., M. H. I. Dore, and H. Butt, 2022: Detecting and quantifying structural breaks in climate. Econometrics, 10, 33, https://doi.org/10.3390/econometrics10040033.
FAO, 2023: Oficina Regional de la FAO para América Latina y el Caribe: Corredor Seco Centroamericano. https://www.fao.org/americas/regional-initiatives/central-american-dry-corridor/es.
Franzke, C., 2010: Long-range dependence and climate noise characteristics of Antarctic temperature data. J. Climate, 23, 6074–6081, https://doi.org/10.1175/2010JCLI3654.1.
Franzke, C., 2012: Nonlinear trends, long-range dependence, and climate noise properties of surface temperature. J. Climate, 25, 4172–4183, https://doi.org/10.1175/JCLI-D-11-00293.1.
Franzke, C. L. E., and Coauthors, 2020: The structure of climate variability across scales. Rev. Geophys., 58, e2019RG000657, https://doi.org/10.1029/2019RG000657.
Gil-Alana, L. A., 2003: An application of fractional integration to a long temperature series. Int. J. Climatol., 23, 1699–1710, https://doi.org/10.1002/joc.968.
Gil-Alana, L. A., 2004: The use of the Bloomfield model as an approximation to ARMA processes in the context of fractional integration. Math. Comput. Modell., 39, 429–436, https://doi.org/10.1016/S0895-7177(04)90515-8.
Gil-Alana, L. A., 2005: Statistical modeling of the temperatures in the Northern Hemisphere using fractional integration techniques. J. Climate, 18, 5357–5369, https://doi.org/10.1175/JCLI3543.1.
Gil-Alana, L. A., 2008a: Time trend estimation with breaks in temperature time series. Climatic Change, 89, 325–337, https://doi.org/10.1007/s10584-008-9407-z.
Gil-Alana, L. A., 2008b: Warming break trends and fractional integration in the northern, southern and global temperature anomaly series. J. Atmos. Ocean. Technol., 25, 570–578, https://doi.org/10.1175/2007JTECHA1025.1.
Gil-Alana, L. A., 2012: Long memory, seasonality and time trends in the average monthly temperatures in Alaska. Theor. Appl. Climatol., 108, 385–396, https://doi.org/10.1007/s00704-011-0539-0.
Gil-Alana, L. A., 2015: Linear and segmented trends in sea surface temperature data. J. Appl. Stat., 42, 1531–1546, https://doi.org/10.1080/02664763.2014.1001328.
Gil-Alana, L. A., 2018: Maximum and minimum temperatures in the United States: Time trends and persistence. Atmos. Sci. Lett., 19, e810, https://doi.org/10.1002/asl.810.
Gil-Alaña, L. A., and P. M. Robinson, 1997: Testing of unit roots and other nonstationary hypotheses in macroeconomic time series. J. Econ., 80, 241–268, https://doi.org/10.1016/S0304-4076(97)00038-9.
Gil-Alana, L. A., and L. Sauci, 2019a: Temperatures across Europe: Evidence of time trends. Climatic Change, 157, 355–364, https://doi.org/10.1007/s10584-019-02568-6.
Gil-Alana, L. A., and L. Sauci, 2019b: US temperatures: Time trends and persistence. Int. J. Climatol., 39, 5091–5103, https://doi.org/10.1002/joc.6128.
Gil-Alana, L. A., M. A. Martin-Valmayor, and C. Hube-Antoine, 2023: An analysis of temperature anomalies in Chile using fractional integration. Stochastic Environ. Res. Risk Assess., 37, 2713–2724, https://doi.org/10.1007/s00477-023-02414-3.
González-Pérez, M. A., Ed., 2022: Regenerative and Sustainable Futures for Latin America and the Caribbean: Collective Action for a Region with a Better Tomorrow. Emerald Publishing, 296 pp., https://doi.org/10.1108/9781801178648.
Granger, C. W. J., 1980: Long memory relationships and the aggregation of dynamic models. J. Econ., 14, 227–238, https://doi.org/10.1016/0304-4076(80)90092-5.
Granger, C. W. J., 1981: Some properties of time series data and their use in econometric model specification. J. Econ., 16, 121–130, https://doi.org/10.1016/0304-4076(81)90079-8.
Granger, C. W. J., and R. Joyeux, 1980: An introduction to long memory time series models and fractional differencing. J. Time Ser. Anal., 1, 15–29, https://doi.org/10.1111/j.1467-9892.1980.tb00297.x.
Hare, S. R., and N. J. Mantua, 2000: Empirical evidence for North Pacific regime shifts in 1977 and 1989. Prog. Oceanogr., 47, 103–145, https://doi.org/10.1016/S0079-6611(00)00033-1.
Harris, I., T. J. Osborn, P. Jones, and D. Lister, 2020: Version 4 of the CRU TS monthly high-resolution gridded multivariate climate dataset. Sci. Data, 7, 109, https://doi.org/10.1038/s41597-020-0453-3.
Hosking, J. R. M., 1981: Fractional differencing. Biometrika, 68, 165–176, https://doi.org/10.1093/biomet/68.1.165.
Ivanov, M. A., and S. N. Evtimov, 2010: 1963: The break point of the Northern Hemisphere temperature trend during the twentieth century. Int. J. Climatol., 30, 1738–1746, https://doi.org/10.1002/joc.2002.
Kane, I. L., and D. Usman, 2013: Long memory analysis of daily average temperature time series. Comput. Eng. Intell. Syst., 4, 52–57.
Koscielny-Bunde, E., A. Bunde, S. Havlin, H. E. Roman, Y. Goldreich, and H. J. Schellnhuber, 1998: Indication of a universal persistence law governing atmospheric variability. Phys. Rev. Lett., 81, 729–732, https://doi.org/10.1103/PhysRevLett.81.729.
Li, X., Y. F. Sang, B. Sivakumar, and L. A. Gil-Alana, 2021: Detection of type of trends in surface air temperature in China. J. Hydrol., 596, 126061, https://doi.org/10.1016/j.jhydrol.2021.126061.
Ludescher, J., A. Bunde, C. L. E. Franzke, and H. J. Schellnhuber, 2016: Long-term persistence enhances uncertainty about anthropogenic warming of Antarctica. Climate Dyn., 46, 263–271, https://doi.org/10.1007/s00382-015-2582-5.
Mangat, M. K., and E. Reschenhofer, 2020: Frequency-domain evidence for climate change. Econometrics, 8, 28, https://doi.org/10.3390/econometrics8030028.
Maraun, D., H. W. Rust, and J. Timmer, 2004: Tempting long memory—On the interpretation of DFA results. Nonlinear Processes Geophys., 11, 495–503, https://doi.org/10.5194/npg-11-495-2004.
Masson-Delmotte, V., and Coauthors, 2019: Calentamiento global de 1,5°C. IPCC, 32 pp., https://www.ipcc.ch/site/assets/uploads/sites/2/2019/09/IPCC-Special-Report-1.5-SPM_es.pdf.
Paul, R. K., P. S. Birthal, and A. Khokhar, 2014: Structural breaks in mean temperature over agroclimatic zones in India. Sci. World J., 2014, 434325, https://doi.org/10.1155/2014/434325.
Percival, D. B., J. E. Overland, and H. O. Mofjeld, 2001: Interpretation of North Pacific variability as a short- and long-memory process. J. Climate, 14, 4545–4559, https://doi.org/10.1175/1520-0442(2001)014<4545:IONPVA>2.0.CO;2.
Percival, D. B., J. E. Overland, and H. O. Mofjeld, 2004: Modelling North Pacific climate time series. Time Series Analysis and Applications to Geophysical Systems, D. R. Brillinger, E. A. Robinson, and F. P. Schoenberg, Eds., Springer Verlag, 151–167.
Phillips, P. C. B., and P. Perron, 1988: Testing for a unit root in time series regression. Biometrika, 75, 335–346, https://doi.org/10.1093/biomet/75.2.335.
Pitman, E. J. G., 1948: Lecture Notes on Nonparametric Statistical Inference. University of North Carolina, 78 pp.
Robinson, P. M., 1994: Efficient tests of nonstationary hypotheses. J. Amer. Stat. Assoc., 89, 1420–1437, https://doi.org/10.1080/01621459.1994.10476881.
Salnikov, V., and Coauthors, 2023: An assessment of the present trends in temperature and precipitation extremes in Kazakhstan. Climate, 11, 33, https://doi.org/10.3390/CLI11020033.
Stern, D. I., and R. K. Kaufmann, 2000: Detecting a global warming signal in hemispheric temperature series: A structural time series analysis. Climatic Change, 47, 411–438, https://doi.org/10.1023/A:1005672231474.
United Nations Educational, Scientific and Cultural Organization, 2016: Toma de decisiones y cambio climático: Acercando la ciencia y la política en América Latina y el Caribe. UNESCO Programme and Meeting Doc., 97 pp.,https://unesdoc.unesco.org/ark:/48223/pf0000245647.
Vergara, W., A. R. Rios, L. M. Galindo, P. Gutman, P. Isbell, P. H. Suding, and J. S. Prólogo, 2013: El Desafío Climático y de Desarrollo en América Latina y el Caribe: Opciones Para un Desarrollo Resiliente al Clima y Bajo en Carbono. Inter-American Development Bank, 105 pp.
Vicente-Serrano, S. M., S. M. Quiring, M. Peña-Gallardo, S. Yuan, and F. Domínguez-Castro, 2020: A review of environmental droughts: Increased risk under global warming? Earth-Sci. Rev., 201, 102953, https://doi.org/10.1016/J.EARSCIREV.2019.102953.
WMO, 2022: State of the Climate in Latin America and the Caribbean 2021. WMO/TD-1295, 44 pp., https://library.wmo.int/records/item/58014-state-of-the-climate-in-latin-america-and-the-caribbean-2021.
WMO, 2023: State of the Climate in Latin America and the Caribbean 2022. WMO/TD-1322, 39 pp., https://library.wmo.int/doc_num.php?explnum_id=11701.
Woodward, W. A., and H. L. Gray, 1993: Global warming and the problem of testing for trend in time series data. J. Climate, 6, 953–962, https://doi.org/10.1175/1520-0442(1993)006<0953:GWATPO>2.0.CO;2.
Woodward, W. A., and H. L. Gray, 1995: Selecting a model for detecting the presence of a trend. J. Climate, 8, 1929–1937, https://doi.org/10.1175/1520-0442(1995)008<1929:SAMFDT>2.0.CO;2.
World Bank, 2023: Climate change knowledge portal. World Bank Group, accessed 16 March 2023, https://climateknowledgeportal.worldbank.org/.
Yao, J., and Y. Zhao, 2022: Structural Breaks in Carbon Emissions: A Machine Learning Analysis. International Monetary Fund, 47 pp.
Yuan, N., Z. Fu, and S. Liu, 2014: Extracting climate memory using Fractional Integrated Statistical Model: A new perspective on climate prediction. Sci. Rep., 4, 6577, https://doi.org/10.1038/srep06577.
Yuan, N., C. L. E. Franzke, F. Xiong, Z. Fu, and W. Dong, 2022: The impact of long-term memory on the climate response to greenhouse gas emissions. npj Climate Atmos. Sci., 5, 70, https://doi.org/10.1038/s41612-022-00298-8.
Zheng, X., and R. E. Basher, 1999: Structural time series models and trend detection in global and regional temperature series. J. Climate, 12, 2347–2358, https://doi.org/10.1175/1520-0442(1999)012<2347:STSMAT>2.0.CO;2.