Beyond Temperature Peaks: The Growing Persistence and Intensity of Tmin and Tmax Heatwaves in Portugal’s Changing Climate (1980/1981–2022/2023)

1. Introduction

Heatwaves are increasingly frequent and severe, posing significant challenges to societies and ecosystems worldwide [1,2]. Mainland Portugal has experienced a notable rise in extreme heat events [3,4], which can lead to the occurrence of heatwaves [5], having severe repercussions for human health, economic activities, and natural environments. Addressing this problem requires a comprehensive evaluation of some of the most relevant heatwave characteristics, focusing on their temporal and spatial dimensions. This study aims to provide such an evaluation by analysing heatwave trends using the dense gridded ERA5-Land temperature dataset [6], thereby addressing the lack of consistent and comprehensive ground-based weather data in the country, with a focus on key urbanised areas in mainland Portugal (Figure 1). However, the lack of a standardised global definition of heatwaves presents a significant challenge, necessitating region-specific studies to understand and mitigate their impacts effectively.

Prolonged periods of excessive heat, commonly referred to as heatwaves, are among the most detrimental natural hazards. They significantly impact human health, sometimes leading to increased morbidity and mortality rates [7], as well as affecting infrastructure, the economy, and natural ecosystems [8]. The Intergovernmental Panel on Climate Change (IPCC) [9,10] describes a heatwave as "a period of abnormally hot weather, often defined with reference to a relative temperature threshold, lasting from two days to months." The precise definition of a heatwave varies by region due to differing normal temperature ranges [11], but key elements include an extended period (at least two days) of unusually hot weather and temperatures surpassing a relative threshold. The IPCC has concluded that strong evidence exists indicating that hot days—i.e., days when daily temperatures exceed a specific threshold [12]—and heatwaves have become more frequent from the 1950s [13]. Since then, heatwaves have also become more intense and prolonged across most land regions, driven by climate change [14]. According to Perkins-Kirkpatrick and Lewis [14], Perkins et al. [15], Perkins-Kirkpatrick and Gibson [16], many regions have experienced significant increases in the number of heatwave days. Some areas, such as the Mediterranean and parts of Africa, have seen particularly pronounced changes, with heatwave days increasing by approximately 2.5 to 5.5 days per decade. This trend highlights the alarming acceleration of extreme heat events [14,17].

Recent research—e.g. [3,4]—indicates that over the past few decades, heatwaves in mainland Portugal have become more frequent, intense, and longer-lasting. Additionally, Portugal has seen a statistically significant increase in extreme heat events, particularly during spring and summer. The frequency of very hot days (daily maximum temperature above 35°C) and tropical nights (daily minimum temperature above 20°C) has risen across the country [18,19]. Heatwaves are projected to become more frequent, severe, and longer-lasting, especially in northern and inland regions [20]. Even under high mitigation scenarios, studies in Environmental Science & Policy predict that the number of heatwaves in Portugal will more than double compared to historical records [21,22,23]. The impacts of heatwaves extend well beyond just rising temperatures, e.g. notably increasing the risk of wildfires: in recent years, extreme heat combined with prolonged drought conditions has led to severe fires. A study published in Agricultural and Forest Meteorology journal found that 97% of extreme wildfires in Portugal occurred during heatwaves, with 90% of extreme wildfire days coinciding with heatwave days [24]. For example, in July 2022, Portugal experienced its longest and most intense July heatwave since 1941, with the northern region of Pinhão (Douro River) registering a record high temperature of 47°C [25]. This extreme heat contributed to significant wildfires, particularly in central (Leiria) and southern (Algarve) regions, where over 30,000 ha were burned, resulting in substantial injuries and fatalities [26]. Such events have far-reaching socio-economic effects, including extensive property damage, loss of life, and increased strain on healthcare and emergency services [27].

The heatwave extreme phenomenon can be assessed using various metrics, including frequency (number of heatwave days), intensity (mean and maximum daily temperatures during heatwaves), and duration. According to Russo et al. [12], a heatwave can be defined by the cumulative exceedances of a specified temperature threshold over consecutive days. This study utilises similar metrics to analyse heatwave characteristics in Portugal comprehensively. Heatwaves can also be quantified using the Heatwave Magnitude Index (HWMI), which combines intensity and duration into a single metric [28], allowing for comparison between different events. Another approach is the Heat Wave Severity Index (HWSI), which incorporates both temperature and humidity effects [29], providing a more holistic measure of heat stress by accounting for heat and moisture. Additionally, extreme value theory can estimate the return periods and probabilities of heatwaves surpassing certain thresholds [30], enabling an assessment of the likelihood and severity of rare and extreme heatwave events. These metrics offer complementary perspectives on heatwave characteristics, facilitating a thorough investigation of their impacts and changes over time.

Similar to many other regions, Portugal faces challenges due to its relatively sparse weather station network, leading to significant data gaps and inconsistencies in record series [31,32]. These limitations hinder comprehensive climate studies that rely solely on weather station data. To address these challenges, researchers increasingly use reanalysis datasets such as ERA5-Land [6,33], which offer continuous and homogenised data for a range of variables, including daily minimum and maximum temperatures (T_min and T_max), at various timescales for global land areas, covering mainland Portugal [34]. The high spatial resolution of ERA5-Land enables detailed analyses of temperature extremes at both local and regional levels, offering a more reliable data source compared to individual weather stations [6,35,36]. Recent studies evaluating ERA5 and ERA5-Land reanalysis air temperature in Portugal—e.g. [4,34]—validate the effectiveness of ERA5-Land data in accurately reproducing average and extreme temperatures across the country. Although ERA5-Land data extends back to 1950, researchers predominantly utilise the period from 1981 or 1980 onwards due to enhanced data quality, consistency, and accessibility [6]. These findings demonstrate that ERA5-Land accurately captures the spatial variability of temperature, closely aligns with observed data, and establishes its reliability for climate research, highlighting the dataset’s suitability for studying heatwave trends and supporting robust long-term analyses spanning decades. The high spatial resolution of the ERA5-Land reanalysis (0.1° × 0.1°, ca. 9 km in Portugal) facilitates the assessment of climate change impacts [36], particularly regarding increasing temperature extremes and the occurrence and intensity of heatwaves in mainland Portugal. This dataset provides a more robust source of information compared to ground-based records [34,37]. The use of these datasets in climate studies is essential for formulating effective adaptation strategies and policies to address climate variability. Such research can guide targeted climate adaptation and mitigation efforts, ultimately protecting human health, infrastructure, and ecosystems from the escalating threat of extreme heat.

Given the aforementioned heatwave trends and their impacts, it is crucial to undertake detailed localised studies to understand the specific characteristics and evolution of this phenomenon in different regions. Thus, the primary objective of this study is to evaluate the temporal changes in various heatwave characteristics, including the number of heatwave days, mean heatwave temperature, and heatwave intensity, at selected reanalysis grid-points across mainland Portugal over the past few decades, focusing on a period from 1980 to 2023. By employing widely used statistical methods in climate research, such as Sen’s slope and the Sequential Mann-Kendall test, on the ERA5-Land temperature dataset, this research provides valuable insights into the evolving trends of extreme temperatures in mainland Portugal’s changing climate, revealing an increasing occurrence and severity of heatwaves, particularly in coastal and southern regions. These findings support the development of effective climate adaptation and mitigation strategies, which are crucial for addressing the growing challenges posed by more frequent and intense heatwave events.

2. Study Area and Data Preparation

Mainland Portugal (Figure 1a), situated on the southwestern periphery of Europe, covers approximately 89,000 km² and is bordered by Spain to the north and east, and the North Atlantic Ocean to the west and south [38]. The country features a diverse topography, ranging from mountainous regions in the north and central areas to rolling plains in the south. Its extensive coastline, stretching over 1,700 km [39], also plays a significant role in shaping local climate conditions. This geographic diversity, combined with the influence of the North Atlantic Ocean, results in distinct climatic features across the country, making it an ideal region for studying spatial and temporal patterns in temperature extremes.

The climate of mainland Portugal is predominantly Mediterranean, characterised by mild, wet winters and hot, dry summers, though significant variation exists between coastal and inland areas. Coastal regions tend to experience more moderate temperatures due to the maritime influence, whereas interior areas, particularly those at higher altitudes, face more extreme temperature conditions, with hotter summers and colder winters [5,40]. In some southern inland regions, summer temperatures frequently exceed 40°C, while winter temperatures in mountainous areas (e.g. Serra da Estrela) can drop below freezing [39].

The main dataset used in this study is continuous series of minimum (T_min) and maximum (T_max) daily temperatures from the openly accessible, regularly updated ERA5-Land reanalysis [6], covering 1,012 grid-points evenly distributed, as shown in Figure 1a. This also addresses the limitations of Portugal’s relatively sparse ground-based weather station network [31,32,41], providing high-resolution, continuous, and homogenised temperature data, thereby enabling a more comprehensive analysis of temperature extremes at both local and regional levels [6,34]. The ERA5-Land T_min and T_max series cover 43 hydrological years, from 1 October 1980 to 30 September 2023. For calculation purposes (explained in the next section), an additional 15 days were included before and after the reference period. Although the dataset is extensive (with a length of [43×365+30]×1,012) this study focuses on analysing heatwave trends at 15 selected grid-points (out of 1,012) to represent the country’s diverse geography and climate, including both coastal and inland urban centres. The actual names of these locations are shown in Figure 1b, while their corresponding codes adopted in the study are presented in Figure 1c. The inclusion of diverse climate locations facilitates a comprehensive examination of temperature extremes in both maritime-influenced regions and those subject to more continental climate patterns.

Using ERA5-Land reanalysis data for 43 hydrological years (1980/1981 to 2022/2023), temperature trends across Portugal were analysed. This includes the averages of minimum, T_min, and maximum, T_max, daily temperatures (Figure 1b and c), as well as of the annual extremes in the 43 years of these variables—AMIN for the annual daily minimum (Figure 2a and b) and AMAX for the annual daily maximum (Figure 3a and b). This description provides a complete view of the thermal conditions in Portugal over the past four decades, forming a basis for the study of heatwaves in the 15 regions of interest.

During the 43-year hydrological period, T_min across mainland Portugal had an overall mean of 10.3°C, with significant spatial variation. Northern high-altitude regions, such as Bragança (BGCA), exhibited lower temperatures compared to southern lowland areas, such as Faro (FARO), where temperatures were generally milder. The lowest mean of T_min was 5.6°C, while the highest reached 15.6°C, as shown in Figure 1b. In contrast, T_max for the same period revealed an overall mean value of 20.0°C, with the hottest locations predominantly concentrated in the Alentejo region, including Beja (BEJA) and Évora (EVRA). The mean of T_max ranged from a low of 14.3°C to a high of 23.4°C, as illustrated in Figure 1c.

Regarding the annual minima (AMIN), the overall mean for T_min (Figure 2a) was -1.6°C, with colder temperatures observed in the north. The lowest mean AMIN T_min was estimated at -8.1°C, while the highest was 6.5°C. For AMIN T_max (Figure 2b), the mean was 22.1°C, with the lowest value of 18.3°C and the highest of 25.3°C. In contrast, the annual maxima (AMAX) series followed a distinct pattern. The overall mean for AMAX T_min (Figure 3a) was 22.1°C, with values ranging from a low of 18.3°C to a high of 25.3°C. From Figure 3b, AMAX T_max showed a more pronounced variation, with a mean of 36.7°C. The highest mean values of AMAX T_max were found in the eastern Alentejo region, while lower ones were observed in some coastal areas, such as Lisbon (LISB) and Lagos (LGOS). The AMAX T_max ranged from 27.3°C to a maximum of 41.0°C.

As illustrated in the preceding figures, temperature extremes in mainland Portugal exhibit significant spatial diversity, influenced by a complex interplay of geographical factors. The variation between the cooler northern highlands and the warmer southern lowlands, coupled with the contrast between temperate coastal areas and more extreme inland regions, underscores the importance of altitude, latitude and proximity to the North Atlantic Ocean in shaping local climatic conditions. Characterising this spatial heterogeneity is crucial for understanding the regional distribution of extreme events, such as heatwaves, and other temperature trends.

3. Methods

This study aimed to evaluate temporal changes in heatwave characteristics at the 15 selected reanalysis grid-points across mainland Portugal. These grid-points, representing key urban areas in different climatic contexts, were analysed over the 43-year hydrological reference period, from 1980/1981 to 2022/2023.

3.1. Heat Wave Magnitude Index

For this study, heatwave events were identified and quantified using the Heatwave Magnitude Index (HWMI) developed by Russo et al. [12]. This index defines a heatwave as an episode lasting a minimum of three consecutive days, during which the daily temperature exceeds a calculated threshold. The threshold was established based on the reference period, encompassing 43 hydrological years, with an additional 15-day buffer on either side, i.e., from 16 September 1980 to 15 October 2023. The threshold is derived as the 90th percentile of daily temperatures, calculated in this research for both T_min and T_max, using a 31-day moving window centred on each day of the hydrological year (1 October to 30 September). For any particular day d, the threshold is determined by calculating the 90th percentile of a dataset $A_d$, which is mathematically expressed as:

$$A_d=\bigcup _{y\in Y}{T}_{y,d-15},\dots ,T_{y,d},\dots ,T_{y,d+15}$$

(1)

where ⋃ represents the union of sets, $T_{y,i}$ denotes the temperature recorded on day i of year y, and Y encompasses all years within the reference period. The annual number of heatwave days is computed by summing the total days that meet the heatwave criteria within a given year:

$$N_{HW}=\sum _{d=1}^{365}I\left(d\right)$$

(2)

where $N_{HW}$ is the number of heatwave days in each year, and $I\left(d\right)$ is an indicator function defined as:

$$I\left(d\right)=\left\{\begin{array}{cc}1\hfill & \mathrm{if}\mathrm{day}d\mathrm{is}\mathrm{part}\mathrm{of}\mathrm{a}\mathrm{heatwave}\hfill \\ 0\hfill & \mathrm{otherwise}\hfill \end{array}\right.$$

(3)

The annual intensity of heatwaves is defined as the highest daily temperature observed during heatwave conditions in a given year:

$$H{W}_{intensity}=max(T_{d1},T_{d2},\dots ,T_{dn})$$

(4)

where $T_{di}$ represents the peak temperature on the i-th day classified as part of a heatwave, and n is the total number of heatwave days in the year. This adapted methodology, based on the work of Russo et al. [12], provides a robust approach for quantifying and comparing heatwaves across different years and locations. By analysing both the occurrence, or persistence, and peak temperatures of heatwave events, the characteristics of extreme temperature episodes can be comprehensively assessed. This approach enables the identification of heatwave events and the calculation of a series of 43 values (series length) for the following three key variables:

• Annual number of days experiencing heatwaves, denoted as ANDH
• Annual average temperature of days experiencing heatwaves, denoted as AATW
• Annual intensity of heatwave days, denoted as AIHD

Although the HWMI was initially developed for maximum temperatures (T_max), its application in this study has been extended to include minimum temperatures (T_min), as mentioned at the beginning of this subsection. This extension is justified by the growing recognition of the crucial role that nighttime temperatures play in heatwave impacts and their potential for a more pronounced increase compared to daytime temperatures [42]. By analysing both T_max and T_min, a more comprehensive assessment of heatwave events is provided, capturing not only daytime extremes but also the persistence of high temperatures throughout the diurnal cycle. This approach aligns with established recommendations that advocate for the consideration of both daytime and nighttime temperatures to achieve a fuller understanding of heatwave dynamics and impacts [43], a perspective that continues to be supported by current research [42].

3.2. Trend Analysis

Sen’s slope estimator and the Mann-Kendall test [44,45] were applied to the three aforementioned variables (i.e., ANDH, AATW, and AIHD) to assess temporal trends. The significance level, $\alpha$, was set at 0.05. It is important to note that for some years within the 43-year hydrological period (max length of the series was 43), no heatwave days were identified. Consequently, the values of the other two variables for those years were recorded as NA (Not Available). The subroutines implemented in R [46] were capable of handling these cases during trend calculations.

3.3. Sequential Mann-Kendall Test

The Sequential Mann-Kendall test, also known as the Mann-Kendall-Sneyers (MKS) test [47,48], was applied to the same three variables to obtain the evolution of trends throughout the studied period, providing a more comprehensive analysis than the single ratio offered by Sen’s slope. Only the progressive trend of the Sequential Mann-Kendall test was considered, as breakpoint analysis was not part of this study. The Sequential Mann-Kendall test statistic is calculated as follows:

$$u\left(t\right)={\displaystyle \frac{t-E\left(t\right)}{\sqrt{var\left(t\right)}}}$$

(5)

where t is the cumulative number of times a data point in the time series is greater than the previous data points, $E\left(t\right)$ is the expected value of t, and $var\left(t\right)$ is the variance of t.

This analysis provides insight into the temporal evolution of trends in heatwave characteristics across the study area, allowing for a more nuanced understanding of climate change impacts on extreme temperature events in mainland Portugal.

4. Results

This section presents the outcomes of an analysis conducted on heatwave characteristics across selected 15 grid-points in mainland Portugal, encompassing the 43-year hydrological period from October 1980 to September 2023. The investigation focuses on trends in three key annual parameters: the number of heatwave days (ANDH), the average temperature during heatwaves (AATW), and the intensity of heatwave events (AIHD). Statistical evaluations were performed utilising Sen’s slope estimator and the Mann-Kendall-Sneyers (MKS) test to address the significance of these trends.

The Heatwave Magnitude Index (HWMI) was applied to identify heatwave days for both minimum (T_min) and maximum (T_max) temperatures. The thresholds ($A_d$) used for heatwave identification, as defined in Equation 1, are illustrated in Figure 4a for T_min and Figure 4b for T_max. These thresholds were subsequently smoothed using Locally Weighted Scatterplot Smoothing [49] (LOWESS), with a smoothing factor of $f=0.4$. LOWESS smoothing was applied to reduce noise and capture the underlying seasonal patterns whilst preserving local variations. These thresholds were established using a 31-day moving window based on a reference period spanning 43 hydrological years, from 16 September 1980 to 15 October 2023, including the additional 15-day buffer on either side of the period.

Analysis of the threshold patterns reveals distinct seasonal variations for both T_min and T_max. T_min thresholds show a smooth seasonal variation, with summer peaks generally between 15°C and 20°C, and winter lows typically between 5°C and 10°C. In contrast, T_max thresholds exhibit a more pronounced seasonal amplitude, with summer peaks generally between 25°C and 35°C, and winter lows typically between 15°C and 20°C. Notably, geographical variations are evident, with coastal areas often displaying higher T_min thresholds, especially in winter, likely due to maritime influences. Conversely, coastal locations tend to have lower T_max thresholds compared to inland areas.

These threshold patterns provide a nuanced basis for identifying heatwaves across different regions of mainland Portugal, accounting for both geographical and seasonal variations in temperature norms. A comprehensive quantitative assessment of heatwave trends for T_min and T_max is presented in subsequent tables and figures, encompassing diverse climatic locations across mainland Portugal. This analysis enables the identification of spatial variations and long-term changes in heatwave characteristics, providing a comprehensive overview of heatwave patterns across the studied geographical areas, which represent both coastal and inland environments.

4.1. Trends in T_min Heatwave Characteristics

The annual number of heatwave days (ANDH) for T_min shows a consistent upward trend across all grid-points, with increases ranging from 0.8 days per decade in Nazaré (NZRE) to 4.2 days per decade in Faro (FARO), as reported in Table 1. The most substantial rise is observed in Faro (FARO), indicating a notable intensification of heatwave occurrences in this southern region. Statistically significant trends (p < 0.05) are identified in Portalegre (PTLG), Beja (BEJA), Lagos (LGOS), and Faro (FARO), suggesting that these areas are experiencing a robust increase in heatwave frequency. The average ANDH for T_min varies across locations, with Nazaré (NZRE) recording the lowest average at 15.5 days and Lisbon (LISB) the highest at 18.0 days, highlighting spatial variability in heatwave occurrences.

In terms of the annual average temperature during heatwave days (AATW) for T_min, most grid-points exhibit slight warming trends; however, Aveiro (AVRO) is an exception with a negligible cooling trend of -0.02°C per decade. Beja (BEJA) shows the most pronounced warming trend at 0.72°C per decade, although none of these trends reach statistical significance. The average AATW values span from 12.05°C in Bragança (BGCA) to 18.63°C in Faro (FARO), reflecting diverse climatic conditions across the study area.

The annual intensity of heatwave days (AIHD) for T_min also indicates positive trends across all grid-points, regardless of the $\mathit{p}-value$, ranging from 0.02°C per decade in Nazaré (NZRE) to 0.90°C per decade in Beja (BEJA). Beja (BEJA) and Faro (FARO) demonstrate the highest increases at 0.90°C and 0.77°C per decade, respectively, with statistically significant trends observed in these locations. The average AIHD ranges from 17.24°C in Bragança (BGCA) to 22.82°C in Faro (FARO), further underscoring the spatial variability of heatwave intensity.

4.2. Trends in T_max Heatwave Characteristics

The results for T_max heatwave characteristics are outlined in Table 2. Overall, the trends are more pronounced and widespread compared to those for T_min. The ANDH for T_max shows positive trends across all grid-points, with increases ranging from 2.2 days per decade in Nazaré (NZRE) to 4.4 days per decade in Lagos (LGOS). Lagos (LGOS) exhibits the highest increase with an additional 4.4 heatwave days per decade, followed closely by Lisbon (LISB) and Faro (FARO), both showing increases of 4.2 days per decade. Statistically significant trends are found in 11 out of the 15 grid-points, indicating a robust increase in T_max heatwave occurrences across mainland Portugal. The average ANDH for T_max ranges from 17.9 days in Lagos (LGOS) to 22.5 days in Santarém (STRM).

The AATW trends for T_max present a mixed picture; while most grid-points show warming trends, five locations—Braga (BRGA), Porto (PRTO), Nazaré (NZRE), Santarém (STRM), and Lisbon (LISB)—exhibit slight cooling trends. Beja (BEJA) and Évora (EVRA) stand out with significant warming trends of 0.96°C and 0.81°C per decade respectively; however, only Évora’s (EVRA) trend is statistical significant ($\mathit{p}-value<0.05$). The average AATW values range from 23.97°C in Guarda (GRDA) to 30.12°C in Beja (BEJA).

The AIHD for T_max also shows positive trends across all grid-points, ranging from 0.23°C per decade in Nazaré (NZRE) to 1.23°C per decade in Évora (EVRA). The highest increases are noted in Évora (EVRA, 1.23°C/decade), Beja (BEJA, 1.19°C/decade), and Faro (FARO, 1.18°C/decade). Statistically significant trends are identified in eight out of fifteen grid-points, indicating widespread intensification of T_max heatwaves.

These findings indicate a general trend towards more persistent, hotter, and intense heatwaves across mainland Portugal over the study period. While all three heatwave characteristics show increasing trends, the most striking and potentially impactful change is observed in the persistence of heatwave occurrences. The annual number of heatwave days exhibits the most consistent and statistically significant increases across the studied 15 grid-points, particularly for T_max. This trend towards more persistent heatwaves is especially pronounced in southern and coastal regions, with locations such as Lagos (LGOS), Faro (FARO), and Lisbon (LISB) experiencing some of the highest increases in heatwave days per decade. Presumably, the prolonged exposure to elevated temperatures by persistent heatwaves may pose greater risks to human health, ecosystems, and infrastructure than shorter, more intense heat events. Notably, while increases in average temperature during heatwaves and intensity are observed, they are generally less pronounced and less consistently significant across locations compared to the increases in persistence. The evolution of these trends is explored in the next subsection using the MKS test, which provides additional insights.

4.3. Temporal Evolution of Heatwave Characteristics

The Mann-Kendall-Sneyers (MKS) test was applied to analyse the temporal evolution of the three key heatwave characteristics (ANDH, AATW, and AIHD) for both T_min and T_max across the 15 grid-points, with results presented in Figure 5, Figure 6, and Figure 7. As mentioned, only the progressive statistic $u\left(t\right)$ was represented in these figures. This decision was taken to focus on the forward evolution of trends and to simplify the visual interpretation of the results, as the progressive statistic offers sufficient information to identify significant trend changes over time. The MKS test yields standardised, dimensionless statistics indicating trend strength and direction over time, with positive values suggesting upward trends and negative values indicating downward trends. When these plots are interpreted, focus should be placed on areas where lines exceed significance thresholds (usually $\pm 1.96$ for $95\%$ confidence). Since the MKS calculations require a minimum number of data points [47,48], the initial years are excluded, leading to plots that start a few years after the beginning of the time series. In this application, the year 1980/1981 was excluded from the plots, and attention was given to the consistently upward trends observed from the 1990s onward.

Figure 5 illustrates the temporal evolution of annual number of heatwave days (ANDH) for (a) T_min and (b) T_max. For T_min (Figure 5a), a gradual increase in ANDH is observed across most grid-points, with the trend becoming more pronounced in recent years. The southern locations, particularly FARO and LGOS, exhibit the most significant and consistent upward trends throughout the study period. These locations show $u\left(t\right)$ values exceeding the significance threshold earlier and more consistently than other areas, indicating a robust warming signal. For instance, FARO’s $u\left(t\right)$ values cross the significance threshold around 2005/2006 and remain consistently above it, suggesting a sustained increase in T_min heatwave days. FARO shows a sharp increase in ANDH from approximately 12 days in the early 1980s to over 37 days by 2020/2021. In contrast, some northern and central locations, such as BGCA and CMBR, show more variable patterns with less pronounced trends, often not crossing the significance threshold. These locations exhibit fluctuations between 10 and 25 annual heatwave days of T_min throughout much of the study period, with an upward trend only becoming apparent after 2019/2020. The ANDH trends for BGCA and CMBR culminate in notable peaks, reaching as high as 55 and 35 heatwave days per year, respectively. The $u\left(t\right)$ values for BGCA, correspondingly, fluctuate around zero for the majority of the reference period, only showing a consistent positive trend in the last few years, aligning with the observed increase in heatwave days. This variability suggests regional differences in the evolution of T_min heatwave frequency.

For T_max (Figure 5b), the increase in ANDH is more evident and widespread across all grid-points. A notable acceleration in the upward trend is observed from the early 2000s onwards for most locations. The year 2002/2003, known for its severe heatwave in some parts of the Iberian Peninsula [50], marks a clear turning point in many locations, after which the ANDH values consistently exceed the long-term average. The coastal locations, including LGOS, FARO, and LISB, demonstrate the most substantial increases, with their $u\left(t\right)$ values crossing the significance threshold earlier and remaining consistently above it. For example, LGOS shows a sharp increase in $u\left(t\right)$ values from around 2000/2001, crossing the significance threshold around 2005/2006 and continuing to rise steeply thereafter. This suggests a more pronounced and sustained increase in T_max heatwave days in these coastal locations compared to inland grid-points. By 2020/2021, these coastal locations (LGOS, FARO, and LISB) frequently experience over 45 annual heatwave days of T_max, compared to 15-20 days per year in the 1980s.

Interestingly, even locations that showed less pronounced trends for T_min, such as BGCA and CMBR, exhibit significant upward trends for T_max. This discrepancy between T_min and T_max trends suggests that daytime temperatures are increasing more rapidly than nighttime temperatures in terms of heatwave frequency.

Figure 6 depicts the temporal evolution of annual average temperature during heatwaves (AATW) for (a) T_min and (b) T_max. For T_min (Figure 6a), the trends are generally less pronounced and more variable across grid-points compared to ANDH. Some locations, such as BEJA and EVRA, show a gradual warming trend, with $u\left(t\right)$ values approaching or occasionally exceeding the significance threshold in recent years—these locations with AATW T_min values increasing by approximately 2.5°C over the 43-year hydrological reference period. For instance, BEJA’s $u\left(t\right)$ values show a consistent upward trend from the mid-1990s, crossing the significance threshold briefly in the early 2010s. This suggests a gradual but significant increase in nighttime temperatures during warm periods in these southern inland locations. Other locations, such as AVRO and NZRE, exhibit more stable or slightly declining patterns, with $u\left(t\right)$ values rarely crossing the significance thresholds. AVRO, for example, shows $u\left(t\right)$ values fluctuating around zero throughout the study period, indicating little change in T_min AATW. This stability in coastal locations could be attributed to the moderating influence of the North Atlantic Ocean on nighttime temperatures [40].

For T_max (Figure 6b), the AATW trends are more distinct, with a majority of grid-points showing an upward trend, particularly from the late 1990s onwards. The inland southern locations, notably BEJA and EVRA, demonstrate the most significant warming trends, with their $u\left(t\right)$ values consistently exceeding the significance threshold in recent years. In these locations, AATW values for T_max have increased by approximately 4°C over the reference period, with particularly high values recorded in recent years such as 2017/2018 and 2022/2023. EVRA, for instance, shows a sharp increase in $u\left(t\right)$ values from the early 2000s, crossing the significance threshold around 2010/2011 and continuing to rise thereafter. This indicates a substantial and sustained increase in daytime temperatures during warm periods in these areas. Coastal locations generally show less pronounced or more variable AATW trends for T_max, with $u\left(t\right)$ values fluctuating around the significance thresholds. LISB, for example, exhibits periods of both increasing and decreasing trends, with $u\left(t\right)$ values crossing the significance threshold in both positive and negative directions at different times. This variability could be related to complex interactions between land and sea temperatures in coastal areas.

Figure 7 presents the temporal evolution of annual intensity of heatwave events (AIHD) for (a) T_min and (b) T_max. As with the previous figures, interpretation should focus on where $u\left(t\right)$ values exceed the $\pm 1.96$ significance thresholds. For T_min (Figure 7a), the trends in AIHD are generally positive but show considerable variability across grid-points. Southern locations, particularly BEJA and FARO, exhibit the most consistent and significant upward trends, with their $u\left(t\right)$ values crossing and remaining above the significance threshold in recent years. In these locations, AIHD values for T_min have increased by approximately 3.5°C over the study period. BEJA, for instance, shows a steady increase in $u\left(t\right)$ values from the mid-1990s, crossing the significance threshold around 2010/2011 and continuing to rise. This indicates a significant intensification of nighttime heatwaves in these southern regions. Northern and central locations show more variable patterns, with $u\left(t\right)$ values often fluctuating around the significance thresholds. BGCA, for example, exhibits periods of both increasing and decreasing trends, with $u\left(t\right)$ values rarely exceeding the significance thresholds. This variability suggests that the intensity of nighttime heatwaves in these regions is less consistently increasing compared to southern areas.

The AIHD trends for T_max (Figure 7b) are more pronounced and widespread compared to T_min. A clear acceleration in the upward trend is observed from the early 2000s for most grid-points, with $u\left(t\right)$ values consistently exceeding the significance thresholds. The inland southern locations, such as EVRA and BEJA, show the most substantial increases in AIHD, with their $u\left(t\right)$ values crossing the significance threshold earlier and remaining consistently above it. By the end of the study period, these locations frequently record AIHD values 5°C higher than in the 1980s. EVRA, for example, demonstrates a sharp rise in $u\left(t\right)$ values from the late 1990s, crossing the significance threshold around 2005/2006 and continuing to increase steeply thereafter. This indicates a dramatic intensification of daytime heatwaves in these inland southern regions. Coastal areas also demonstrate significant upward trends, albeit with some variability in the timing and consistency of threshold exceedance. FARO and LGOS, for instance, show strong upward trends in $u\left(t\right)$ values, crossing the significance threshold in the early 2000s and remaining above it. However, the rate of increase is generally less steep compared to inland southern locations, possibly due to the moderating influence of the ocean on extreme temperatures [40].

These analyses reveal a consistent increase in the persistence of heatwave events for both T_min and T_max, with more pronounced trends for T_max from 1980/1981 to 2022/2023. For T_min, ANDH shows upward trends, particularly in southern locations, with Faro (FARO) experiencing an increase of 4.2 days per decade. The persistence of T_max heatwaves demonstrates even greater increases, with 11 out of 15 grid-points exhibiting statistically significant rises in ANDH. Lagos (LGOS) shows the highest increase at 4.4 days per decade, followed closely by Lisbon (LISB) and Faro (FARO) at 4.2 days per decade. Therefore, it appears that coastal locations experience the most pronounced increases for both temperature variables. AIHD further supports this trend, with significant increases observed in southern regions. In contrast, AATW exhibits less change compared to ANDH and AIHD, indicating that the primary shift is occurring in the occurrence and intensity of heatwaves rather than their average temperature. These findings indicate a new pattern in heatwave occurrence and duration across Portugal, particularly in coastal and southern locations, signifying growing persistence of heatwave conditions. This increased persistence, particularly from the early 2000s onwards, suggests a potential climate regime shift in mainland Portugal.

5. Discussion

The analysis of heatwave characteristics across mainland Portugal from 1980/1981 to 2022/2023 reveals significant trends in the persistence and intensity of heatwave events, particularly for maximum temperatures (T_max). These findings align with broader European and global trends of increasing heatwave frequency and severity [14].

5.1. Persistence of Heatwave Days

The consistent upward trend in the annual number of heatwave days (ANDH) for both T_min and T_max across all studied grid-points clearly indicates more days under heatwave conditions. This trend is particularly pronounced in coastal and southern Portugal, with Lisbon, Lagos, and Faro experiencing the most substantial increases for T_max. These findings are consistent with observations across the Mediterranean region, where climate change is amplifying heatwave occurrences [51]. Interestingly, some grid-points show unexpected trends. For instance, Bragança (BGCA), located in the northern interior, exhibits a statistically significant increase in ANDH T_max (3.0 days/decade, $\mathit{p}-value=0.024$), despite its inland location. This suggests that the intensification of heatwaves is not limited to coastal areas and may be influenced by local topography [52]. The spatial variability in heatwave persistence, with coastal areas showing more significant increases, suggests a potential influence of land-sea temperature contrasts on heatwave dynamics. This pattern may be related to changes in atmospheric circulation patterns, such as the strengthening of the Azores High, which has been linked to increased heatwave frequency in southwestern Europe [53].

5.2. Intensity of Heatwave Events

The increasing trends in the annual intensity of heatwave days (AIHD), particularly for T_max, indicate that heatwaves are becoming hotter, although randomly. This intensification is especially evident in southern inland locations such as Évora (EVRA) and Beja (BEJA). The combination of increased frequency and intensity poses significant challenges for various sectors, including agriculture, public health, and energy demand [54]. Notably, some grid-points with less pronounced ANDH trends show significant increases in AIHD. For example, Guarda (GRDA), despite its moderate increase in ANDH T_max (3.6 days/decade), shows a significant rise in AIHDT _max (1.03°C/decade, $\mathit{p}-value=0.037$. This highlights the importance of considering both frequency and intensity metrics when assessing heatwave trends [55].

5.3. Implications of Nighttime Temperature Trends

While T_max trends are more pronounced, the increasing trends in T_min heatwave characteristics, particularly in southern regions—exemplified by Faro (FARO)—are equally concerning. Elevated nighttime temperatures reduce the opportunity for nocturnal relief during heatwave events, potentially exacerbating health risks, especially for vulnerable populations [56]. Interestingly, some northern and central coastal locations such as Porto (PRTO) and Aveiro (AVRO) show contrasting trends. While both have similar, non-significant increases in ANDH T_min, AVRO shows a significant increase in ANDH T_max (3.6 days/decade, $\mathit{p}-value=0.013$), while PRTO’s increase is not statistically significant (2.9 days/decade, $\mathit{p}-value=0.054$). This difference suggests that local factors, such as urban heat island effects or specific coastal dynamics, may be influencing heatwave trends [57].

5.4. Statistical Significance and Climate Signals

The interpretation of statistical significance in climate trend analysis requires careful consideration. While traditional approaches might lead to discarding results that do not meet the arbitrary $\mathit{p}-value<0.05$ threshold, recent literature suggests a more nuanced approach [58]. In the context of this study, even trends that do not meet conventional significance thresholds may represent meaningful climate signals, especially when considered alongside statistically significant results and broader regional patterns. This approach aligns with that of Portela et al. [59], who mapped rainfall trends regardless of statistical significance, whilst highlighting locations meeting the selected significance level threshold. For instance, while Nazaré (NZRE) shows the lowest increases in both ANDH T_min and ANDH T_max, these trends still align with the overall pattern of increasing heatwave frequency across Portugal. The lack of statistical significance in these trends should not necessarily lead to their dismissal, but rather prompt further investigation into local factors that might be moderating heatwave intensification in this coastal location.

5.5. Spatial Patterns of Temperature Trends Across Mainland Portugal

The analysis of annual temperature and extreme values (AMIN and AMAX) trends across 1,012 grid-points in mainland Portugal from 1980/1981 to 2022/2023 reveals a complex spatial pattern of warming, corroborating and extending findings from the heatwave characteristics analysis. Examining the temperature trends offers a nuanced understanding of climate change impacts across different regions, highlighting the spatial variability of warming patterns. The use of high-resolution gridded data allows for a detailed assessment of local and regional variations.

Figure 8 illustrates a widespread warming trend in the annual means of both temperatures, with T_max generally exhibiting more pronounced increases. This figure is particularly significant as it provides a clear visual representation of the spatial distribution of warming trends across mainland Portugal. T_min trends range from 0.07°C to 0.29°C per decade, whilst T_max trends are slightly higher, typically ranging from 0.07°C to 0.35°C per decade. The spatial pattern revealed in Figure 8 shows a clear west-east gradient in warming trends, with western regions experiencing more unnoticeable warming compared to eastern ones. For T_max trends from north to south, northern coastal areas exhibit more moderate warming of around 0.13°C per decade, while southern regions show trends exceeding 0.30°C per decade. Notably, inland regions, e.g. Bragança (BGCA), show some of the highest warming rates, with the annual mean of T_min increasing by 0.25°C per decade and that of T_max by 0.27°C per decade. Overall, the results show pronounced warming in interior regions suggesting a potential amplification of continental climate effects under global warming. Coastal areas, particularly along the western coast, exhibit a more moderate warming trend. However, even these coastal regions are experiencing significant warming, with most grid points showing positive trends for both T_min and T_max.

Trends in annual mean temperatures also reveal interesting local variations that warrant further investigation. For instance, some mountainous regions in central Portugal, such as Serra da Estrela near Guarda (GRDA) and Castelo Branco (CABO), show slightly lower warming rates compared to surrounding lowlands, possibly due to altitude effects on temperature trends. Additionally, these trends align with the increasing frequency and intensity of heatwaves observed in the previous analysis, particularly for T_max. The more rapid warming of T_max compared to T_min suggests an increase in daily temperature ranges.

The trends in the annual minimum (AMIN) temperatures, depicted in Figure 9, exhibit considerable variability across Portugal with a somewhat west-east gradient (from cooling on the coast to warming inland). AMIN T_min trends range from approximately -0.20°C to 0.50°C per decade (Figure 9a), whilst AMIN T_max trends are positive, ranging from about 0.05°C to 0.60°C per decade (Figure 9b). Interestingly, some coastal locations show moderate warming in AMIN T_min but lower or even slight cooling trends in AMIN T_max. For instance, Nazaré (NZRE), near the western coast, show positive trends for AMIN T_min but negative trends for AMIN T_max. Differently, eastern inland locations such as Bragança (BGCA) demonstrate warming trends in AMIN for both T_min (0.34°C per decade) and T_max (0.34°C per decade). These patterns suggest complex local factors influencing minimum temperatures, related to land-sea interactions and topographical features.

Figure 10, illustrating trends in the annual maximum (AMAX) temperatures, reveals consistently positive trends across Portugal, particularly for AMAX T_max. The trends in AMAX T_min range from slightly negative, localised values to approximately 0.55°C per decade, whereas AMAX T_max trends are more pronounced, varying from about -0.14°C (also highly localised) to 0.61°C per decade. Notably, the strongest warming trends in AMAX T_max occur in inland and southern regions. For instance, grid points in the northeast display robust increases exceeding 0.40°C per decade in T_max AMAX. These substantial increases align with the observed intensification of heatwaves, particularly in inland Portugal.

The spatial patterns revealed by these analyses underscore the complexity of identifying climate change impacts on temperature trends in Portugal. Regional variability in heatwave characteristics highlights the need for tailored adaptation strategies, as different areas experience varying temperature changes. The strong warming trends in both AMIN and AMAX indicate a shift towards a warmer climate regime, emphasising the importance of comprehensive monitoring to inform effective adaptation and mitigation efforts.

6. Conclusions

The increasing persistence and intensity of heatwaves in Portugal, as revealed by this study, underscore the urgent need to identify climate change effects to develop robust climate adaptation strategies. These findings can inform policy decisions related to urban planning, public health preparedness, and agricultural practices. For instance, the pronounced trends in coastal areas highlight the need for specific adaptation measures in these densely populated regions, such as improved urban green spaces and building designs that mitigate heat stress [60]. In conclusion, this study provides robust evidence of increasing heatwave persistence and intensity across mainland Portugal, with particularly pronounced trends in inland southern regions, which are already experiencing water scarcity. The spatial variability in these trends, including very specific and local occurrences, underscores the complex interplay of geographical, meteorological, and anthropogenic factors in shaping local extreme temperature patterns. The mention of water scarcity is pertinent as heatwaves and water availability are intrinsically linked; prolonged high temperatures can exacerbate water stress through increased evapotranspiration and higher demand, potentially creating compound events with more severe socio-economic and environmental impacts than temperature extremes alone. This interconnection highlights the need for integrated climate adaptation strategies that address both heat and fresh water-related challenges in vulnerable regions. These findings emphasise the importance of targeted adaptation strategies and further research to mitigate the impacts of increasingly frequent and severe heatwaves in a changing climate, particularly in areas prone to water scarcity, thereby enhancing resilience to compound climate extremes and their cascading effects on different sectors.