text new page (beta)
English (pdf)
Article in xml format
Article references
Automatic translation
Send this article by e-mail
Cited by SciELO 
Similars in
SciELO 
Permalink
1. Introduction
Planetary climate will continue to change as a consequence of human greenhouse gas emissions (IPCC, 2013). This includes a surge in global temperatures, a redistribution of hydrological patterns and an increment of extreme events (Collins et al., 2013). Just from 1950 until today the mean global temperature has increased by 0.7 ºC, with some regions (such as the high latitudes of the Northern Hemisphere) already experiencing as much as an additional 3 ºC (Hartman et al., 2013). Similarly, over the past two decades we have seen changes in precipitation seasonality (Chou et al., 2013; Feng et al., 2013), in hurricane intensity (Knutson et al., 2013) and on the strength of droughts (Trenberth et al., 2014). Thus, we are living under a new global climatic regime, completely different from the whole of the Holocene and likely in the history of life on Earth (Steffen et al., 2018).
As a result of climate change, several components of the Earth system that provide key contributions to humanity are or will be impacted. From a reduction in food producing systems to rising sea levels, more intense droughts, vector-borne diseases spread, and increasing wildfire frequency to mention just a few (IPCC, 2014). These negative impacts are and will be widespread globally. However, the true magnitude of the impact is dependent on the particularities of each region and the vulnerabilities of each country (Wootten et al., 2017). Thus, there is a pressing need to understand local and regional particularities of climate change.
Mexico is one of the most vulnerable countries to climate change due to its geographical location. Precipitation in the country is highly variable and depends on multiple global factors such as El Niño Southern Oscillation, the North Atlantic Oscillation (Lachinet et al., 2012), the North Atlantic and Pacific hurricane season length and strength (Krishnamurti et al., 1990), and precedent soil moisture and vegetation altering the intensity of the Mexican monsoon (Xiang et al., 2018). As a result, the country’s precipitation pattern is highly variable on the interannual scale (with annual national means varying from 620 to 1025 mm according to my calculation, and similar values estimated by Livneh et al., 2015) leading to recurrent extreme precipitation events across most of the territory; a pattern likely to strengthen in future climate change scenarios (Magaña et al., 1997).
Worrisome climate change has already strengthened climate variability and extremes in Mexico (e.g., Gay et al., 2015). This has resulted in increasing drought and water scarcity, as well as the opposite, namely increased storms, intense rainfall and flooding across the country (Mora et al., 2018). A large increment in potential evapotranspiration has also been observed, as a result of higher temperatures, which lead to increasingly drier conditions due to a reduction of soil moisture and water availability (Livertman and O’Brien, 1991). Both the continuous increment in dryness and extreme droughts have strong negative impacts on grain production (Arce et al., 2020), crop yields (Murray-Tortarolo et al., 2018) and livestock populations (Murray-Tortarolo and Jaramillo, 2019), ultimately impacting the society and causing increase migration to the USA (Feng et al., 2013).
Despite the above mentioned conditions, we currently ignore the exact magnitude and particularities of past climate change in Mexico. In particular, we are missing a country-level spatially explicit analysis, based on observed data. Thus, the main objective of this work is to fill this gap by showing how climate has changed in Mexico over the last 70 years (from 1950) in terms of precipitation, temperature and the full hydrological budget. Specifically, I seek to understand changes at the national level and regionally within the country.
2. Methodology
2.1 Datasets
I used mean monthly temperature, precipitation, potential evapotranspiration (based on the Penman-Monteith formula) and number of rainy days (defined as a day with precipitation values above 0.1mm) for the period 1950-2017, extracted from CRUv4.02 datasets (for a full explanation on how the variables are computed please see Harris et al., 2014). These global databases have estimated monthly values at a 0.5º × 0.5º resolution from 1901-2017. I selected the period 1950-2017 due to inherent limitations of the dataset, as 63% of the weather stations do not present measurements prior to the selected period. Due to the low resolution of the pixels for a national-level analysis, I re-scaled the grid to 0.1º × 0.1º using a bilinear interpolation that accounted from topographic patterns for better presentation of the results; however, nation-level statistics were done on the original half degree resolution.
2.2 Advantages and limitations of employed data
I selected the CRUv4.02 database based on several important advantages for this particular analysis; however, it also has some important limitations that need to be considered when interpreting and using my results. In the next paragraph I detail both.
Probably the most important benefit is the regularity of the grid. The database uses a regular 0.5º grid, which allows for quick computation of statistics and trends, and simplifies plotting (for a total of 2442 grid points for the whole territory). In the particular case of Mexico, the climate information for each grid-point is based on an average of 7.6 weather stations, which translates in a total of 104 to 348 stations used across the whole territory, depending on the year. But, importantly, one station can provide information for several grid points, as mean climate is interpolated from all available datapoints, thus providing a large extent of information. The dataset also benefits from providing a full spatiotemporal cover that has been shown to correctly replicate coarse-resolution observations (Livneh et al., 2015). However, it has some important limitations. The output data itself is not homogenized; thus, linear temporal trends should be analyzed carefully. In addition, a 0.5º grid is too coarse to capture the full extent of local micro-climates derived from the largely heterogeneous Mexican topography. As a result, spatially explicit trends should be interpreted carefully and, if used for local-scale studies, they should be regarded as a general guideline and not as observations. Despite the previous, decadal patterns and national-level trends can be considered to be reliable (Harris et al., 2013). In the future, a comparison with other observation-based datasets could provide further insights on how climate changed in Mexico, particularly at the regional scale within the country.
2.3 Data analysis
National level statistics (mean, deviation and trends) were calculated at two different temporal scales, seasonal and annual, for the whole time period. I split the year into the usual four seasons (spring, summer, autumn and winter), but also into two precipitation seasons (with an equal number of six months): wet (June-November, which accounted for 79-87% of the total annual rainfall based on my calculation) and dry (December-May). For the whole time series (based on annual means) I calculated the net change in the variables, as the slope of a 10-year running mean and from decadal averages. The monthly national scale means for precipitation minus potential evapotranspiration were calculated for each decade, but only the first and last are presented to simplify the results. I also calculated the standard error for each month (n = 10 based on each year of the decade) and perform a repeated measure analysis of variance (RMANOVA) using both decades. Similarly, I calculated the average number of rainy days per season for the same decades.
For the spatial analysis I calculated the gridded mean (0.5º × 0.5º grid) and linear trend (I presented the gridded slope to show general spatial patterns, and the significance of the trends as stipple) for temperature, precipitation and precipitation minus potential evapotranspiration for the full time period. I further re-calculated the later at a seasonal scale. Regional-level statistics (mean and trends for all variables) were calculated for three regions within the country which were determined a posteriori, based on distinctive spatial patterns: the surrounding area to the Gulf of Baja California, the Central Highlands and the Yucatan Peninsula.
3. Results
3.1 Changes in mean national-scale temperature and precipitation
Mean climate in Mexico has changed over the last 70 years. In particular, the mean national-scale temperature has increased by +0.71 ºC from 1951 to 2017 (10-year running mean, p < 0.001) or by +0.96 ºC if calculated using decadal means (p < 0.001) (Fig. 1, left panel). Interestingly, there was a -0.36 ºC decline during the period 1951-1980 (p = 0.01), also seen globally and likely driven by the global increment in atmospheric aerosols (Tegen et al., 2000); thus, the warming trend over the last 40 years was even stronger, gaining almost 2 ºC (+1.93 ºC, p < 0.001). As a result of the warming, the five hottest years in this record occured in the last 10 years (2017, 2016, 2015, 2012 and 2009, in that order) all of which had an exceeding temperature of at least +0.9 ºC compared to the mean of the whole period or an additional +1.3 ºC if compared to the 1951-1960 average. However, the change was not homogeneous within the year, with spring temperatures (March-May) increasing the most with +0.93 ºC and summer-autumn (July to November) the least with +0.57 ºC. Thus, representing a higher temperature increment in the dry season (December-May: +0.82 ºC) than during the rainfall season (July-November: +0.58 ºC) (Table I).
Fig. 1 Temporal evolution of mean annual temperature (left) and precipitation (right) across the whole of Mexico from 1951-2017. Dots represent the annual mean and the red line a 10-year running mean. The red arrow indicates the net decadal change from the 1950s to the 2010s in both variables.
Table I Seasonal (based on four thermal and two rainfall seasons) changes in mean annual temperature (ºC) and precipitation (mm and percentage of the seasonal or annual mean) across Mexico, for the period 1951-2017.
| Season | Temperature change (ºC) | Precipitation change (mm and %) |
| Spring (MAM) | 0.93 | 2.29 (3.18%) |
| Summer (JJA) | 0.58 | 18.37 (5.06%) |
| Autumn (SON) | 0.57 | 13.51 (5.42%) |
| Winter (DJF) | 0.69 | -0.34 (-0.7%) |
| Dry season (DJFMAM) | 0.81 | 1.95 (1.66%) |
| Wet Season (JJASON) | 0.575 | 31.88 (5.20%) |
| Annual | 0.71 | 34.41 (4.51%) |
Contrary to temperature, precipitation did not show a clear annual trend over the last 70 years. On average, there was a +34 mm increment in mean annual precipitation, but it was not significant (p = 0.08) (Fig. 1, right panel). The five wettest years on record were distributed across the whole time series (2013, 1958, 1984, 1981, 1973), just like the five driest years (1953, 1994, 1987, 1977, 1982), which are potentially related to variation in El Niño Southern Oscillation as suggested by previous studies (Magaña et al., 2003). Nonetheless, there were statistically significant changes in the seasonal precipitation. In particular, the wet season precipitation (June-November) had an increment of +32 mm (p = 0.02), which accounts for 94% of the whole annual trend. In contrast, dry season precipitation (December-May) showed no changed (+1.95 mm) with even a small decline in winter (December-February) precipitation (Table I).
3.2 Seasonal changes in the water balance
My results show a clear seasonal change over the last 70 years: the wet season has become wetter and the dry season drier (Fig. 2). In that sense, the July-November (wet season) water balance shows an excess of +54 mm of available water (not evaporated) for soil moisture recharge and runoff, when comparing 2008-2017 against 1951-1960. In contrast, December-March precipitation (dry season) had an average decrement of -39mm.
Fig. 2 Monthly precipitation minus potential evapotranspiration as the mean of two decades: the start of the period (1951-1960; black dotted line) and the end (2008-2017; red dotted line). Solid lines indicate one standard error from the mean (n = 10).
Two key seasonal aspects are also evident: the consistent monthly differences across decades and the unchanged number of wet days in each season. In this sense, the increment/decrement in wet season/dry season water balance is consistent across all months (except for October-November). Thus, we see an increase in excess available water throughout the wet season (statistically significant for July, August and September) and a decrease across all the dry months (statistically significant throughout the whole season). In contrast, the number of wet days in each season did not change (from 46.5 ± 2.2 to 48.3 ± 2.3 rainy days in the wet season from 1951-1960 to 2008-2017, and from 14.4 ± 0.6 to 14.4 ± 0.6 in the dry season). This means that the seasonal changes in water availability, particularly in the wet season, are driven by changes in the amount and not the seasonality of precipitation.
3.3 Spatial distribution of trends
Spatially, the climate trends were largely heterogeneous across the country, but more so for precipitation than for temperature. In the case of temperature, there was a consistent increase all over the country over this seven decades, that ranged from +0.01 to +0.029 ºC per year (or an increment of +0.1 to +2.7 ºC across the whole time series), with three marked regional patterns. The region surrounding the Gulf of Baja California had the highest increment, gaining as much as +2.7 ºC over the last 67 years. The Yucatan Peninsula also displayed a thermal increment, although much smaller (+0.7 ºC). Finally, the mid and mid-west regions had only a minor increment of +0.1 ºC (Fig. 3).
Fig. 3 Mean (top) and linear trend (bottom) for temperature (left) and precipitation (right) for the period 1951-2017. Stippling indicates regions with statistically significant trends (p < 0.05).
Precipitation shows both increments and decrements across the nation, which consistently lead to the zero-sum national trend. In that sense, the mountainous regions in the east, middle and particularly the west, display a decrease in precipitation between -40 and -70 mm yr-1. The pattern is particularly strong across the region surrounding the Gulf of Baja California, where the decline represented as much as 20% less annual rainfall. In contrast, precipitation increased in the central highlands (+32 mm yr-1) and the Yucatan Peninsula (+53 mm yr-1); nevertheless, the percentage increment in the first region represented an order of magnitude higher (+18% and 0.2%, respectively) (Fig. 3).
The spatial integration of precipitation minus the potential evapotranspiration reveals the arrangement of changes in water availability across the country. In this sense, the pattern is remarkably similar to the changes in precipitation and is also spatially consistent across seasons. Thus, there has been a general decline in the hydrological balance across the country that extends throughout the year, and an increment in the Central Highlands and the Yucatan Peninsula. In addition, there is a positive trend in the wet season water balance in the south (centered around the state of Guerrero), which is also seen in the total precipitation. Importantly, a key region with strong negative trends is the one surrounding the Gulf of Baja California (Fig. 4).
4. Discussion
My results show a change in climate across Mexico over the last 70 years, consistent with global patterns (see Harris et al., 2013). Firstly, the net country-scale temperature increment (+0.71 ºC from 1950 to 2017) coincides with a global increment of land temperature (+0.67 ºC) for the same time period, with a higher gain since the 1980s (IPCC, 2013). Interestingly, the interannual variation is similar across global and national scales; e.g., extremely hot years, such as 2017, coincide in magnitude with +1.3 ºC above the mean (Thompson, 2017). Thus, it is possible that the surge in temperature across the nation is the result of global climate change and not the natural variability of Mexican climate, as also supported by local studies (e.g., Martínez-Austria et al., 2016). However, a formal attribution is needed to support such statement. Hence, it is very likely that temperature will keep rising in the future, at a rate determined by the global pathway (Magaña et al., 2000).
However, important regional differences occur within the country in the thermal trends. In that sense, the Northwest warmed much faster than the rest of the country. At least two potential mechanisms may be responsible for such regional pattern. Firstly, the extremely fast warming of the northern Pacific Ocean, which has led to increasing temperatures along western North America (Barnett et al., 2005; Karmalkar and Bradley, 2017) and impacted extreme heatwave frequency in northern Mexico (Martínez-Austria et al., 2016). Secondly, a link to precipitation trends: a reduction in available water can increase atmospheric temperature through a reduction in latent heat fluxes. Globally, a reduction in precipitation has been shown to increase the consecutive number of hot days (Mueller et al., 2012), which could be the case of the high temperature increment in the areas where precipitation decreased (e.g., Sonora). In contrast, the additional rain and positive water balance in the central highlands means lower sensible heat fluxes and a smaller temperature gain. Thus, it is very likely that future regional temperature trends will be closely linked to changes in available water and oceanic heat fluxes.
Precipitation did not change at a national scale, in line with global patterns (Harris et al., 2013); however, there was an increment in the seasonality of water availability. In that sense, the dry season got drier and the wet season wetter throughout the country, similar to what has been observed globally for recent decades (Chou et al., 2013) and particularly over the tropics (Feng et al., 2015). The analysis shows a stable number of wet days across the years, which means that the increment in water availability during the wet season is the result of higher precipitation rates (based on the data employed, mean national-level wet season rainfall rate changed significantly from 12.05 to 12.27 mm day-1 or a 2% change) and must be a consequence of more intense rainfall over the same number of days. One potential explanation is an increase in intense precipitation pulses, such as hurricanes or tropical storms that induced precipitation or a stronger monsoon season. On the opposite hand, the deficit in the dry season is likely the result of increased vapor pressure deficit as a consequence of a higher increase in temperature during the season. Interestingly, these results are consistent with a recent global analysis by Mora et al. (2018), who identified an increase in drought and storm events across Mexico, both presently and in the future. This also helps explain the contrasting trend of excess and lack of available water as a result of marked seasonal changes, but further research is needed to attribute its causes.
Finally, regional precipitation patterns present an interesting feature: a general reduction following a U-shape across the limits of the country and an increase in the central plains and the Yucatan Peninsula, leading to mirrored spatial changes in water availability. One potential explanation across the northern and central West Pacific region is the observed reduction in westerlies velocity, which has led to lower winter precipitation rates across western US and potentially Mexico, and may explain the decreasing dry-season water balance (Luce et al., 2013); but it can also be related to seasonal changes in the North American Monsoon (Cook and Seager, 2003; Demaria et al., 2019). In the Yucatan Peninsula, local studies have also shown this increase in rainfall, likely a result of changes in meridional trade winds, which have been shown to shift from shallow (low precipitation) to deep (high precipitation) convection (Díaz-Esteban and Raga, 2019), with a strong increase in seasonality probably modulated by ENSO (Díaz-Esteban and Raga, 2018). For the central highlands, precipitation has been shown to be controlled by the Mexican monsoon (Douglas et al., 1993), likely strengthening as a consequence of higher temperatures. Finally, though spatially explicit trends should be interpreted carefully, the patterns of changes in water availability are almost mimicked by the corresponding trends in river discharge over a similar time period (Milliam et al., 2008), which improves the reliability of the regional hydrological patterns presented.
5. Conclusion
My analysis shows a clear change in Mexican climate over the last 70 years. In particular, the country is getting warmer and the seasonal water balance more extreme. These results are in line with global patterns and current observations, showing the nation to be facing stronger droughts and more intense rainfall. The patterns are very likely the consequence of early climate change, although a formal attribution is needed; thus, it seems likely - and worrisome - that the national-scale climate will become even more seasonally extreme in the future, at a magnitude defined by the global pathway.
