TRENDS OF CHANGE IN PRECIPITATION AND IN DROUGHT SUSCEPTIBILITY AS ASSESSED BY THE STANDARDIZED PRECIPITATION INDEX (SPI) IN NORTHEAST PORTUGAL TENDÊNCIAS DE MUDANÇA NA PRECIPITAÇÃO E NA SUSCEPTIBILIDADE À SECA AVALIADA PELO ÍNDICE DE PRECIPITAÇÃO NORMALIZADA (SPI) NO NORDESTE DE PORTUGAL

Much of NE Portugal is an area of arid zones with moderate to severe susceptibility to desertification and drought. A single and coupled trend analysis of precipitation and SPI drought index was performed in six weather stations in NE Portugal (representative of the regional climate domains), over a seventy-year time series (1931-2000). Precipitation over the study period decreased at most stations, with a more pronounced trend in the wetter areas, whereas in the semiarid areas an increase in precipitation was found. As well as the frequency of months with severe and extreme drought, the frequency of dry and wet months increased when the SPI timescale increased (1 to 6 months) and when a more recent computation period is considered (1931-1960 to 1971-2000). The results highlight a tendency for precipitation extremes to occur in NE Portugal. The general development of trends in precipitation and drought reported in this study confirm a progressive severity in the susceptibility to desertification affecting NE Portugal.


Introduction
Changes in global climate are being reported and future climate scenarios keep alerting to relevant changes in mean air temperature and rainfall amounts and distribution patterns (IPCC, 2014). Uncertainties persist in what regards projections of future climate precipitation amounts but such projections consistently point out future increase in precipitation variability.
As so, actual precipitation extremes are expected to increase in frequency in the future and future extremes are expected to increase in magnitude as compared to the actual ones. As a consequence, droughts and flood events are expected to occur more frequently and with higher magnitude than in present time, prospecting potentially damaging impacts of climate change.
Variability of weather conditions increased in recent years, and if this trend persists in the near future, higher frequency and severity of extreme events are expected, namely an extension of drought periods. Apart from the global scale evidence, for mainland Portugal the scenario of widened precipitation extremes also applies (SIAM2, 2006;Costa et al., 2012;APA, 2015;EEA, 2021;Fraga et al., 2017;Fraga et al., 2021).
Understanding drought phenomena, their monitoring, identification and prediction, help mitigating drought effects around the world, and particularly in arid environments and in the Mediterranean drylands.
As a complex phenomenon, there is not a universal concept to define drought. In general terms, drought is understood as a natural phenomenon associated with water scarcity due to natural precipitation variability, to which has to be added an insufficient supply by water provision humanmade structures (Wilhite, 1992). Each drought event should be defined according not only to the climatic characteristics of the area but also to the impacts that it causes (Wilhite, 1992;Wilhite et al., 2007). In this sense, droughts are commonly classified in four types, according to its duration and related impacts: meteorological, hydrological, agricultural and socioeconomic drought (Wilhite and Glantz, 1985;Wilhite, 1992).
Many drought indexes have been developed worldwide, for a particular or large-scale climatic region, using different data sources. Most drought indexes are based on meteorological parameters, e.g. precipitation, temperature, evapotranspiration, while others include also soil moisture. Studies using remote sources, single or combined with ground data sources, visibly progressed in the last decade (Nicolai-Shaw et al., 2017;Wei et al., 2021). On the thematic of drought and desertification, review works describing and comparing the existing drought indicators have also been published in the last decade (Cherlet et al., 2018;Fernandes e Heinemann, 2009;Svoboda and Fuchs, 2017;Zargar et al., 2011).  Hayes et al., 2011, p. 488).
The SPI was developed by McKee et al. (1993). It requires only precipitation as input data, and it was designed to detect precipitation deficit at multiple timescales (1 to 48-month), an advantage that recommended its wide application. In fact, once SPI is a normalized or standardized index, it has wide applicability and may allow comparison between distinct climate regions around the world (Svoboda and Fuchs, 2017). As pointed out by several studies, SPI's representation of actual drought conditions is more consistent for short-medium timescales than for the larger ones (Vicente-Serrano e López- Moreno, 2005;Tirivarombo et al., 2018). SPI was used in several contexts within the drought and desertification thematic, approaches covering drought forecasting (Mishra and Desai, 2005;Cancelliere et al., 2007) drought monitoring (Tirivarombo et al., 2018), and drought prediction, in this case using Markov Chains to predict SPI Drought Class transition (Steinemann, 2003;Paulo et al., 2005;Paulo and Pereira, 2007;Avilés et al., 2016). As well, SPI was applied in the hydrological context, for example to investigate reservoir effect on catchment response as represented by the streamflow series (Serra, 2014).

Drought characterization by IPMA (Instituto Português
do Mar e da Atmosfera) uses PDSI -Palmer Drought Severity Index, developed by Palmer in 1965, which is based on precipitation and temperature data, although IPMA is also applying SPI in internal research projects, following WMO recommendations. Some studies using SPI and PDSI have been conducted in Southern Portugal, as in the Alentejo (Paulo and Pereira, 2006), Algarve and Baixo Alentejo regions (Pulido-Calvo et al., 2020).
However, no specific study applying SPI is reported for the Northeast (NE) region of Portugal. For example, Paulo et al. (2016) carried out an analysis using SPI in several locations across Portugal, but none was located The aridity index (AI) corresponds to the ratio between annual Precipitation and potential evapotranspiration, with four classes of AI observed in NE territory: Semiarid (0.2 < AI <0.5), dry-subhumid (0.5 < AI < 0.65), wet sub-humid (0.65 < AI < 1) and humid (AI > 1) (PANCD, 2011). These climate classes are used for desertification susceptibility purposes, the semiarid corresponding to severe susceptibility and the dry subhumid to moderate susceptibility. According to Köppen Climate Classification, NE Portugal fits in the Csa and Csb climates, both Mediterranean with dry, hot summer (Csa) in the south, and mild dry summer (Csb) in the north and extreme northeast of the region (Köppen, 1936;IPMA, 2021). Agroconsultores and Coba (1991) apply a regional climatic classification based on mean annual temperature and mean annual precipitation, following Gonçalves's works (Gonçalves, 1985a(Gonçalves, , 1985b, which divided the NE territory, known as Trás-os-Montes, in Terra Fria (Cold Land), Terra Quente (Warm Land) and Terra de Transição (Transition Land). These two classifications fairly match with that based on AI and so they may also be used to roughly identify areas with severe susceptibility to desertification in NE Portugal. In fact, most of the semiarid and dry sub-humid areas fall in Gonçalves Terra Quente climate classes (Royer, 2019).
The analysis was carried out on six selected weather stations considered representative of the climatic domains of the study area, NE Portugal ( fig. 1, Table I).
These weather stations are part of the meteorological monitoring network of SNIRH -Sistema Nacional de Informação de Recursos Hídricos -available o n l i n e a t h t t p s : / / s n i r h . a p a m b i e n t e . p t / i n d e x . php?idMain=2&idItem=1. Double mass analysis (Linsley et al., 1982;Arikan and Kahya, 2019) showed no break in the cumulative plot of paired weather stations monthly precipitation series for the study periods defined, and the correlation between all paired series ranged from 0.995 to 0.999, with a median of 0.999, therefore indicating the homogeneity of the data series analyzed in this study.The precipitation change trends over time were carried out with annual mean values of two reference periods: the decades and the thirtyyear climatological normal. Moreover, an interannual analysis was performed by selecting two key months: the months with the highest and lowest precipitation of the hydrological year (October to September). In both cases, the temporal evolution of the coefficient of variation (standard deviation / average) of the precipitation series was also assessed, together with the influence of terrain variables on precipitation temporal trends derived for each weather station. SPI computation requires precipitation monthly series as input data. For each station, the climatological normal periods with complete data series were identified and SPI was computed for each one of the thirty-year in the NE region. Some approaches assessing the spatial variability of precipitation and drought events in Portugal are reported in literature, commonly using one type of cluster analysis to define homogeneous regions in what regards the studied parameters (Santos, et al., 2010;Moreira et al., 2015).
Much of Trás-os-Montes mountain region, NE Portugal, are drylands enduring soil degradation processes, therefore facing medium to high susceptibility to desertification and drought (Figueiredo et al., 2015b).

The Intermunicipal Plans for Adaptation to Climate
Change of Terra Quente Transmontana and Terra Fria Transmontana (PIAAC_TQT, 2018;PIAAC_TFT, 2018) identified the main climate vulnerabilities to which the NE territory is susceptible, which include excessive precipitation events, heat waves and droughts, while minor expression was assigned to strong winds, cold waves, frosts and fog. Some of these vulnerabilities play also an essential role on wildfire occurrence, a hazard with potential and actual negative impacts on natural resources in this region, which increased frequency has been reported (Cavalli, et al., 2019;Figueiredo et al., 2014). Due to this critical scenario, typical of the Mediterranean mountains, it is important to monitor drought and study changes in climate patterns, supported in long term data series. In this line, the present study aimed at analyzing temporal change trends in precipitation and drought events over 70 years in the last century in the northeast (NE) Portugal. Moreover, an analysis was carried out on the extension of drought periods in the region, using SPI generated by long term precipitation series of representative weather stations.

Methodology
The natural susceptibility of the study area, NE Portugal, increased by past and present human activities, has been already reported in literature. The Soil Map of Northeast of Portugal indicates that 60% of the territory (1.309 million ha) corresponds to degraded soils (Agroconsultores and Coba, 1991). Major soils that cover this surface are Leptosols (72%), with incipient profile development (Figueiredo, 2013). They are shallow, acid, present high stoniness and low organic matter content, limiting soil water storage, therefore conditioning land use options.
For each one of these decades, the three selected SPI timescales were correlated by linear regression. distribution in NE Portugal ( fig. 3). For the selected stations, mean annual precipitation is positively correlated with elevation (r² = 0.63) and latitude (r² = 0.72). The correlation between those variables is often reported.
For the NE region, Figueiredo et al. (1990) obtained determination coefficients (r²) of 0.513 between annual precipitation and elevation, and of 0.430 between annual precipitation and latitude. Besides, mapping exercises of the regional climates distribution was very much supported on the altitudinal gradients of temperature and precipitation observed throughout this region (Gonçalves, 1985a;Agroconsultores and Coba, 1991). The effect of longitude on precipitation reflects continentality, a factor that may explain mean annual precipitation at Malhadas as compared with that at Vinhais (respectively below and above the expected from the altitudinal gradient).
Malhadas is the most eastern inland station, under the continental influence of the Iberian Meseta, whereas Vinhais is the most western one, still receiving the Atlantic influence (Gonçalves et al., 2016).
To explore the precipitation temporal pattern of evolution in each station throughout the whole study period, annual precipitation averages for three periods of climatological normal (1931-1960, 1951-1980, 1971-2000) and their overlapped decades : 1951-1960 and 1971-1980 are presented (fig. 4).

Change trends in precipitation
Mean annual precipitation in the selected weather stations for the whole study period , which shows the strong altitudinal influence in precipitation     for each station and its respective elevation.  para cada estação e sua respectiva altitude.

Fig. 3 -Precipitação média anual do período de estudo
The NE Portugal presents sharp climatic contrasts, well represented in fig. 3, and fig. 4   . The standard deviation is higher in the overlapped decades, as well as the coefficients of variation, once they represent just a 10 years series, while the climatological normal periods represents a 30 years period. Considering the coefficients of variation (CV) just for the three climatological normal periods, Carviçais, Alfândega da Fé and Malhadas presented a decrease in the CV from the latest to the most recent thirty-year period, while for the remaining stations the CV rised over time.
The months with the highest and the lowest precipitation of each hydrological year (October from September) were selected to analyze the temporal change trends that occurred along the study period in the wettest and the driest seasons, respectively. As expected in a Mediterranean domain, the wettest month was January (more frequently) or December, and the driest was August. The evolution of the mean monthly precipitation    According to Costa et al. (2012), NE Portugal may be an exception in the national territory, since the future projections of regional climate models reveal an overall drying trend across Portugal, while a precipitation increase is projected over NE Portugal, concentrated in winter. Moreover, as also projected, an increase is expected in the numbers of extreme precipitation events and its contribution to the seasonal totals in winter and spring. As such contributions compensate the decrease in the drier seasons mean precipitation, in the future regional climate higher than actual annual precipitations are therefore expected (Costa et al., 2012).

SPI change trends
As mentioned before, SPI identifies the wet and dry months of a long-term monthly precipitation series, following the classification presented in Table II. Furthermore, the SPI for the two overlapped periods were confronted by linear regression considering the same SPI timescale and the respective climatological normal. Regression functions obtained relate 120 paired sets of monthly SPI data in each overlapped decade, the ones issued from the earlier 30-year period (x) and the other issued from the more recent one (y). For most of the selected stations, the slopes of the regression line and the determination coefficients obtained in each case (Table III) tend to increase as the timescale increases, for the same overlapped decade. In other words, the SPI data distribution of each series is changed from the earlier to the more recent overlapped decade, and the regression parameters change so as to fit the more frequent extremes in more recent periods.
Focusing the analysis on the dry months, three stations were selected as example ( fig. 8), where the earlier aggregated SPI classes are now presented as defined in Table  II. As drought severity depends on duration, intensity and geographical extent of a specific event (Wilhite and Glantz, 1985), at longer timescales the SPI responds more slowly to short term precipitation changes because the precipitation of the following month has less impact in the index. Hence, drought conditions assessed by SPI6 tend to persist with higher severity (McKee et al., 1993), summing up a larger number of months as compared to shorter timescales. In this study, this can be observed in both overlapped decades for Vinhais and Alfândega da Fé. Actually, Vinhais stands as a good example to show the SPI trend towards the extremes from earlier to more recent times, as in the more recent precipitation series more months are classified in the dry classes. It is important to remind that Vinhais falls in the wet sub-humid climatic domain. For the semiarid Alfândega da Fé, the same pattern was found, however, more pronounced in the first overlapped decade (1951)(1952)(1953)(1954)(1955)(1956)(1957)(1958)(1959)(1960) since for the overlapped decade 1971-1980 fewer months are included in SPI dry classes. This may be explained by the increase in rainfall amounts recorded in Alfândega da Fé, as discussed in the previous subsection (PIAAC_TQT, 2018). However, the opposite pattern is observed for Macedo de Cavaleiros, where the more recent overlapped period includes much more severe drought episodes than the earlier overlapped period ( fig. 8).
In this context, Kovacs et al. (2014) stress that increases in rainfall amounts alone do not necessarily lead to less severe hydrological drought conditions and higher soil moisture because of the rise in actual evapotranspiration. In fact, the extremely dry periods for Macedo de Cavaleiros station are more frequent in the more recent overlapped decade .

Precipitation variability and drought in NE Portugal
A drought episode reflects a below-average rainfall expected for a region. Thus, to illustrate the relation between the precipitation amounts and SPI values  This is not only because of precipitation changes through time, as part of its inherent variability, but also because more data are available as data series grow in extent.
Differences in drought variability patterns suggest splitting Continental Portugal in two sub-regions (Martins, et al., 2012;Moreira et al., 2015;Santos, et al., 2010). Analyzing the spatial drought variability based on SPI and applying principal component analysis,   concluded that the maximum number of consecutive dry days for the Trás-os-Montes regions tends to increase, in addition to the maximum number of consecutive wet days (Santos et al., 2019).
Along the present study, change trends occurring in NE Portugal, concerning precipitation and drought, were identified and discussed, using a point approach based on weather stations representative of the different climatic domains prevailing in the region. This is understood as a contribution to address prospected challenges in regional water resources management in a changing environment and to consider in regional plans of adaptation to climate change as PIAAC_TQT (2018)

Conclusion
A single and coupled trend analysis of precipitation and SPI drought index was performed in six weather stations in NE Portugal, along a seventy-year time series. This included three climatological normal periods (1931-1960, 1951-1980 and 1971-2000), with two overlapped decades (1951-1960 and 1971-1980). The selected weather stations are representative of the climatic domains found in the region, ranging from semiarid to humid.
Precipitation over the last century study period tends to decrease in most weather stations. This trend is more pronounced in stations located in wetter areas. On the contrary, one station located in the semiarid shows an increase in mean annual precipitation and for the months with related highest and lowest precipitation (January and August, respectively), which is in agreement with future climate projections for the region (PIAAC_TQT, 2018;Costa et al., 2012). The decrease in mean rainfall amounts observed in stations located in wet subhumid and dry subhumid areas matches with reported changes in the aridity index in the same areas.
SPI was computed for some stations in the NE Portugal and allowed to assess drought frequency and severity across the study area. SPI class frequency analysis in the overlapped periods enabled to detect temporal changes in the proportion of dry, normal and humid classes. As SPI timescale increases (1 to 6 months) and a more recent climatological normal is taken as input data (1931-1960 to 1971-2000), the frequencies of dry and wet months in the overlapped periods increased, at the expenses of the frequency of normal months. Frequencies of months with severe and extreme drought during the overlapped periods increased as well, more visibly for the SPI 3-month timescale and for stations located in dry sub-humid and semiarid climatic domains.
Globally, results highlight a tendency towards the occurrence of precipitation extremes in NE Portugal. Both extremes conditions, wet and dry, are damaging in frequency.
These statistically driven findings agree with currently reported climate change projections, yet with a larger uncertainty when comparing precipitation with temperature.
Climate change projections are likely to have impacts on water resources availability, therefore, pointing out the need for tuned management of natural water bodies or reservoirs (Andrade et al., 2011). Water scarcity (or excess) is a critical limitation to crop productivity and the agricultural product in Portugal, affecting crops as vines and olives, among others cultivated in NE Portugal (Yang et al., 2020). Moreover, changes in precipitation temporal variability and seasonal distribution also have consequences to wildfire hazard (Costa et al., 2012).
The general change trends in precipitation and drought reported in the study confirm a progressive severity in the susceptibility to desertification actually affecting NE Portugal.