Air temperature has a significant impact on health, environment, agriculture, society, and economy. Variability, extremes, and trends have been extensively investigated, based on daily, monthly, or annual data in Portugal [1
]. However, studies based on hourly data are not currently available. In effect, hourly data provides a much more accurate picture of the thermal conditions at a given location or region, also contributing to improving our understanding of climatic patterns and regimes. This kind of information is very useful for assessing climatic conditions and suitability of a given region for growing specific crops or forest species. This is particularly relevant for mainland Portugal, where temperature is spatially heterogeneous, showing strong spatial gradients. In addition, there is a strong seasonality and large inter-annual variability in precipitation and temperature across the territory [3
Mainland Portugal is located in the transitional region between the sub-tropical anticyclone and the sub-polar low-pressure systems [2
], therefore its climate is influenced by the alternation of anticyclones and depression systems. However, locally the climate is also greatly influenced by the complex orography, latitude, and proximity of the Atlantic Ocean [5
]. In fact, mainland Portugal presents large orographic heterogeneities, being more mountainous in the northern region (where maximum terrain elevation reaches 1993 m a.s.l.) than in its southern part (Figure 1
a). These heterogeneities play a key role on the precipitation distribution [4
] and on temperature pattern (Figure 1
). With a predominantly mild Mediterranean climate [2
], Portuguese climates feature typical warm and dry summers and cool and rainy autumn–winter periods, but with large inter-annual variability and seasonal irregularity. The annual mean temperature varies between about 9 °C, in the north and central mountains, to approximately 18 °C in the southern coast and inner Alentejo (Figure 1
The territory is administratively organized in NUTS (Nomenclature of Territorial Units for Statistics) level 2 (Figure 1
a) and NUTS level 3 divisions (Figure 1
b). These units will be used herein to summarize information on the prevailing thermal conditions throughout the country.
It is widely accepted that the two dominant global climatic drivers of plant distribution are temperature and water availability [5
]. The plant growth and development is dependent upon the temperature surrounding the plant and each species has a specific temperature range, commonly characterized by a thermal optimum, minimum, and maximum thermal limits [6
]. For fruit species, a certain heat accumulation is essential for adequate growth, physiological development and to obtain ripe fruits [7
]. Many woody plant species growing in temperate climates also require a certain amount of exposure to cold temperatures to break winter dormancy and prepare to springtime warming [9
]. Several studies have combined heat and chill indices for studying thermal requirements of different fruit species [10
]. Santos, Costa, and Fraga [7
] evaluated the thermal growing conditions of fruit species in mainland Portugal using two bioclimatic indices: growing degree hours and chilling portions. These two indices were estimated based on gridded daily mean temperatures.
With its large range of climates, Portugal produces various types of fruits and crops. Agricultural soils cover more than 35% of the total area of Portugal [17
]. Grapevines and olive trees are the fruit crops that occupy the largest area, of 178,770 ha and 361,483 ha, respectively [18
]. In terms of fruit production, viticulture is responsible for the biggest production (802,278 t), followed by fresh fruits (542,203 t) and citrus fruits (402,924 t), according to the Portuguese statistics office [18
]. In Portugal, some fruit species are located in very narrow areas. This is the case, for example, of carob and citrus fruits, which are mainly located in the Algarve region, while chestnut trees are located in the northeast of Portugal. Other species, such as olive trees and grapevines are distributed throughout several regions of the country, with a wide range of cultivated varieties and under very different thermal conditions. Olive tree (Olea europaea
L.) is an important crop in Mediterranean countries, being drought-tolerant, as it may survive and produce fruits with low water availability [19
]. Typically, grapevine (Vitis vinifera
L.) is also a key crop in Mediterranean countries, comprising many of the world renowned winemaking regions located in areas with Mediterranean-type climates [21
The objectives of the present study are twofold: (1) to provide a very high-resolution mapping (~1 km resolution) of the current thermal conditions over mainland Portugal on the hourly timescale, thus improving former classifications based on the daily timescale; (2) to evaluate the thermal growing conditions of main fruit species in Portugal, with special emphasis on the grapevines and olive trees.
2. Materials and Methods
In this study, an innovative hourly observational database is used to calculate the number of hours within different temperature intervals, i.e., to assess the empirical distribution of hourly temperatures. Furthermore, the number of hours per different temperature intervals of eight fruit species are analyzed separately, giving more emphasis to grapevines and olive trees, owing to their key role in the Portuguese economy.
The hourly temperature dataset used was obtained from the Portuguese national weather service’s (Instituto Português do Mar e da Atmosfera, IPMA) network of meteorological stations in mainland Portugal. Of the original 133 stations, those with more than 20% of missing hourly data were rejected beforehand. Thus, 63 stations were selected for the 19-year period 2000–2018 (Figure 1
). The geographical locations of the selected stations are shown in Figure 1
. In general, these stations are representative of the altitudinal variation within the territory (Table S1
and Figure 1
). The station with the highest elevation, Penhas Douradas (1380 m), is located in Serra da Estrela (the highest mountain range in mainland Portugal), while the station with lowest elevation is located in Figueira da Foz (4 m). The southernmost weather station is Sagres (37°01′ N), while the northernmost is Monção (42°06′ N). Regarding the longitude, it varies from 9°48′ W in Cabo Raso to 6°27′ W in Miranda do Douro (Table S1
). A preliminary quality-checking of data was carried out. Box-plots of each station time series were plotted to represent the full empirical distributions of observed data and to detect possible erroneous values (not shown). Among the 63 selected stations, 21 stations have less than 5% of missing hourly data, 22 stations between 5% and 10%, 15 stations between 10% and 15%, and 5 stations between 15% and 20% (Figure S1
). Hourly data were obtained by averaging 10-min instant data at each weather station.
A two-step methodology was followed in order to obtain the final maps of the number of hours within each temperature interval: (1) a geostatistical temperature interpolation in order to capture the spatial variability of the number of hours for different temperature intervals, and (2) a bias correction of the interpolated data. First, the number of hours for each different temperature interval was calculated from the long-term means (baseline of 2000–2018) of the hourly temperatures at each weather station separately. The following temperature intervals were selected: ≥4–<8 °C, ≥8–<12 °C, ≥12–<16 °C, ≥16–<20 °C, ≥20–<24 °C, ≥24–<28 °C, ≥28–<32 °C, and ≥32–<36 °C. The number of hours with temperature <4 °C and >36 °C was excluded from this study, because they occur in only a few days of the year and with a restricted geographic distribution, thus not warranting a statistically robust analysis. In order to obtain the number of hours in different temperature intervals, the long-term means for the baseline 2000–2018 were chosen, as they allow overcoming possible biases in the database, e.g., owing to missing data. Furthermore, the aggregation of all years was the best option for this study purposes, which is focused on spatial variability rather than on temporal variability. However, through a sensitivity analysis shown in Figure S2
, it was possible to attest that the choice of the hourly climatological means instead of the whole raw hourly data had only a negligible impact on the relative distribution of the number of hours (shape of the empirical distributions), showing differences of less than 2% in the relative number of hours for all temperature intervals and for all meteorological stations.
Second, the number of hours in each temperature interval and for all selected sites were subject to ordinary least squares (OLS) linear regressions, with three independent variables: latitude, elevation, and Euclidean distance to the coastline (at 1 km resolution). The aim was to perform a regression analysis (one for each temperature interval) with the number of hours as dependent variable and the geographical variables (latitude, elevation, and distance to coastline) as exploratory variables. From the multiple regression analysis, the regression coefficients were obtained and map algebra tools in GIS were applied to produce the final maps from the gridded variables at a 1 km grid resolution.
The resulting OLS maps presented the inherent error of the regression adjustment, i.e., they explain the influence of the geographic variables on the model [22
]. On the other hand, the residuals of the regression fit at each meteorological station tend to reflect other factors beyond the independent variables, such as soil cover properties, local winds (e.g., breezes), or other mesoscale systems [1
]. These factors may also influence the model, underlying important biases between interpolated and observed values. Hence, to improve the OLS results, bias corrections were carried out based on the number of hours for the different temperature intervals, computed for the network of 63 meteorological stations and over the same baseline period (2000–2018). An inverse distance weighted (IDW) interpolation of the biases (residuals) to the 1 km grid was then added to the OLS of the number of hours, separately for each temperature interval. The IDW interpolation determines each grid point value using a linearly weighted combination of a set of sample points. The weight is a function of the inverse distance, assuming that the variable value being mapped decreases with the distance [23
]. Kriging, spline, and polynomial interpolations were also tested, but the IDW method was eventually chosen, since it provided the lowest mean squared error (not shown). The aforementioned procedure was independently repeated for the number of hours within the different temperature intervals. The percentages of hours by the different temperature intervals in a year (relative frequencies of occurrence) were then calculated and their spatial variability was examined.
The thermal conditions associated with different key agricultural and forest species in Portugal were assessed. A total of eight species were selected: almond tree (Prunus dulcis D. A. Webb
), carob tree (Ceratonia siliqua
L.), chestnut tree (Castanea sativa Mill.
), citrus fruit species (e.g., oranges, lemons, clementine’s and mandarins), cork oak (Quercus suber
L.), holm oak (Quercus ilex
L.), olive tree (Olea europea
L.), and grape vines (Vitis vinifera
L.). Their land coverage was retrieved from the fifth level of Land Use and Occupation Map: COS2007 (Source: Direção-Geral do Território, http://www.dgterritorio.pt/
). Owing to the high relevance of olive trees and vines to the Portuguese economy, they were specifically analysed within their main demarcated regions. This procedure enabled to highlight the prevailing thermal conditions in each region. For this purpose, the 50 Denominations of Origin (DOs) or sub-regions for the grapevines, and the 6 Protected Designations of Origin for olive trees (PDOs) were considered. As for the other species there are no protected designations of origin, it was not possible to apply the same methodology.
4. Discussion and Conclusions
This study provides an innovative analysis of the near-surface air temperature in mainland Portugal, based on hourly observations over the recent period of 2000–2018 and from a large weather station network (63 meteorological stations). The whole process is fully implemented in a GIS and the final maps are published on Supplementary Material
, in a widely used file format (shapefile), thus allowing their future application to a wide range of assessments. Although a longer data period would be preferable, typically 30 years, as is recommended by the WMO (World Meteorological Organization), data prior to 2000 have important gaps and inconsistencies that may strongly limit the statistical robustness of the results. In addition, of the 133 meteorological stations with hourly data available, only 63 are used, due to the significant number of data gaps. The present study is a trade-off between the density and temporal consistency of the meteorological station network. A set of hourly temperature maps is obtained using the most recent data available, thus providing accurate information on the current thermal conditions throughout the country.
In the first part of the study, the results allow a detailed geographical characterization of the thermal conditions in mainland Portugal, at a high spatial resolution of 1 km. Such information is of great value to growers and stakeholders in the agrarian sector [30
], as it provides a comparison of the different climatic conditions in each agrarian zone. Additionally, the results provide information concerning the relationship between geographical variables and temperature. Thus, final maps can be automatically updated with new hourly temperature data and easily managed within a GIS environment, thus permitting future improvements resulting from new data inputs.
Overall, the outcomes highlight a strong spatial variability and strong seasonality of temperatures in mainland Portugal. The NH 4–8 °C patterns mainly reflect wintertime conditions, whereas the NH 32–36 °C patterns are a manifestation of summertime conditions. The number of hours in the year with cold conditions (4–12 °C) is higher in the northern-central regions, mainly in mountainous areas. On the other hand, the number of hours in the year with temperate conditions (12–20 °C) is largely influenced by the distance to the coastline, presenting much lower values over inner Portugal, mostly in the northeast. Still, the very warm conditions (number of hours with temperature between 28–36 °C) are essentially restricted to inner-southern Portugal and to the upper Douro valley.
The present study results are also key to understanding climate change impacts on each agrarian region in Portugal. Taking into account previous studies regarding climate change projections for Portugal [7
], it is expected that temperatures should increase. It should be noted that extreme temperatures (>30 °C) are expected to be more frequent under future climates. Hence, the number of hours in each class of these maps are expected to shift under future climatic conditions.
In the second part of the study, some application examples are shown, through the extraction of information for the NUTS level 3 administrative divisions in mainland Portugal. This work may be replicated in the future, in other territorial units up to the local level or other socio-economic sectors. The most striking correspondence between geographical distribution of cool and very warm areas and occurrence of the species is that of chestnut and holm oak (Figure 5
a,d and Figure 6
). Carob and citrus trees do not have chill requirements while chestnut trees require moderate chilling accumulation. In fact, in carob and citrus trees, the low temperatures may cause chilling and freezing injuries on plants and consequently limit crop production and yield [32
]. The chestnut trees located in the northeastern part of the country, are the species with the major number of hours with lower temperatures. In addition to the moderate chill requirement, the chestnut trees present a high sensitivity to summer droughts [34
]. Of note also, the holm oak trees, located mainly in the southeast of the country, have the highest number of hours within the range of 32–36 °C. However, according to Terradas [35
], holm oak is less adapted to extreme drought in comparison with other evergreen Mediterranean tree species, such as carob, olive, or cork oak. The areas of species like: holm oak, olive, almond, and cork oak are those that have more hours with temperatures between 28–36 °C (very warm conditions).
More specifically, vine and the olive trees are analyzed by region to accurately characterize their current growing conditions and to identify regions with similar features. It is important to highlight that there are different varieties/cultivars of grapevines and olive trees that tend to be grown in the most suitable areas of each region. Many of these varieties are either native or have been empirically selected in each region over several decades or even centuries [7
]. As such, they already tend to be grown in the most favourable pedoclimatic conditions so as to be economically viable and sustainable, thus warranting some correspondence between their spatial distribution and bioclimatic suitability. Santos, Costa, and Fraga [7
] selected eight main fruit species in Portugal, to analyze the potential shifts in their thermal conditions, using two bioclimatic indices (growing degree hours and chilling portions) and based on the daily grid dataset from E-OBS [36
]. For the future, increases of heat accumulation and decreases of chilling accumulation are projected over most of Portugal. Inner southern Portugal is expected to undergo the most detrimental climatic changes for temperate fruit and nut trees, since decreases of heat accumulation and significant reductions of winter chill are expected in this region [7
]. The results for the current period, presented herein, are in line with this preceding study, though the present study allows a more detailed and accurate characterization of the thermal conditions in mainland Portugal by using hourly data.
This knowledge is of great importance for different socio-economic sectors, such as agriculture and forestry, e.g., by assisting bioclimatic zoning and suitability assessments. An accurate characterization of the bioclimatic conditions of a given territory is an asset that may potentiate a better planning and management. The resulting knowledge may help decision support systems provide more reliable and accurate guidelines to stakeholders, decision-makers, or farmers. Furthermore, the application to the fruit species undertaken in the present study is a mere illustration of a possible usage of the obtained information. It may be applied to a much wider range of areas of research fields and practical applications such as territorial planning, biodiversity, hydrology, agriculture and forestry systems, and civil protection.