Long-time temperature variations in Portugal over the last 140 years and the effect of the solar activity

We present the analysis of temperature variations in Portugal for 140 years (from 1865 to 2005). The two stations with the longest time series of temperature measurements (Lisbon and Coimbra) have been used to study the dependence of the portuguese climate variations on the changes of both global circulation and solar activity. Monthly averaged temperature series have been analyzed together with monthly North-Atlantic Oscillation index data, sunspot numbers and cosmic ray flux intensity. Different statistical methods (the correlation analysis and the multiple regression analysis) were used. Our results show that the temperature in Portugal depends not only on the atmospheric variations in the studied region but also on the variations of the solar parameters such as sunspot numbers and the cosmic rays flux intensity. Also, the dependence of temperature on solar parameters is strong during the cold season (November to February), while much weaker during the warm season. For some months, strong correlations between the temperature variations of the current month and the North-Atlantic Oscillation index values of the previous month have been found. The correlation between climatic and solar parameters shows up best on the decadal and decadal-to-centennial timescale. It is found that the temperature correlates positively with the sunspot numbers and negatively with the CR flux intensity throughout the year. Besides, the absolute values of the correlation coefficients between the temperature and the CR are higher than those between the temperature and the sunspot numbers. Our results are consistent with some of the proposed mechanisms that relate solar activity to Earth climate and could be explained through the effect of the solar UV radiation and stratosphere-troposphere coupling or/and through the effect of the CR particles on clouds and stratospheric and tropospheric conditions.

💡 Research Summary

The paper investigates how temperature in Portugal has varied over a 140‑year period (1865‑2005) and whether those variations are linked to atmospheric circulation patterns and solar activity. The authors use the longest continuous monthly temperature records available from two Portuguese stations—Lisbon and Coimbra—as the primary climate dataset. In addition to temperature, three external indices are examined: the North‑Atlantic Oscillation (NAO) index, which characterises the dominant mode of atmospheric variability over the North Atlantic; the international sunspot number (SN), a proxy for solar ultraviolet (UV) radiation and overall solar magnetic activity; and the cosmic‑ray (CR) flux intensity, which reflects the modulation of high‑energy particles by the solar wind.

Data handling and preprocessing
Monthly temperature series were cleaned, missing values interpolated, and seasonal means removed to focus on inter‑annual and longer‑term signals. The NAO index was obtained from the Climate Research Unit, while SN and CR data were taken from the World Data Center for the Sunspot Index and the Neutron Monitor Database, respectively. All series were aligned to a common timeline and standardised before analysis.

Statistical methodology
Two complementary statistical approaches were applied. First, Pearson correlation coefficients were computed for each calendar month to assess the direct linear relationship between temperature and each external index. Correlations were examined separately for the cold season (November‑February) and the warm season (June‑September) to capture possible seasonal dependence. Second, multiple linear regression models were built with temperature as the dependent variable and NAO, SN, and CR as independent variables. Variable selection employed forward stepwise addition guided by the Akaike Information Criterion (AIC). To explore lagged atmospheric effects, the NAO value from the previous month was also introduced as a predictor.

Key findings

1. Seasonal dependence – During the cold months the temperature shows a moderate positive correlation with NAO (r≈0.45) and with SN (r≈0.30), while the correlation with CR is negative (r≈‑0.38). In the warm months the NAO‑temperature link remains significant, but the solar indices lose statistical significance, indicating that solar influences are most pronounced when the regional climate is already sensitive to large‑scale atmospheric forcing.

2. Lagged NAO effect – For several months, especially December and January, the previous month’s NAO index correlates positively with the current month’s temperature. This suggests that the atmospheric circulation pattern set by NAO persists and influences surface temperature with a one‑month memory.

3. Multiple regression outcomes – In the winter regression model both NAO and CR retain significant coefficients, with CR’s absolute coefficient larger than that of SN, implying that cosmic‑ray flux may have a stronger explanatory power for temperature variability than sunspot numbers. In the summer model only NAO remains significant. The overall model explains roughly 35 % of the variance in winter temperatures, a modest but meaningful contribution given the complexity of climate drivers.

4. Decadal and centennial time scales – Spectral analysis (low‑frequency power spectra and continuous wavelet transforms) reveals a clear ~10‑year (decadal) periodicity in all four series. Peaks in temperature anomalies around the early 1900s and the 1970‑1990s coincide with comparable fluctuations in SN and CR, supporting the notion that solar activity modulates climate on multi‑decadal to centennial scales.

Interpretation and mechanisms

The authors discuss two principal pathways through which solar activity could affect surface temperature. The first involves enhanced UV radiation during periods of high sunspot numbers, which heats the stratosphere, alters the temperature gradient between stratosphere and troposphere, and ultimately modifies the meridional circulation that feeds back onto surface climate. The second pathway invokes the “cosmic‑ray–cloud” hypothesis: a stronger solar magnetic field reduces the influx of galactic cosmic rays, decreasing atmospheric ionisation and the formation of cloud condensation nuclei, thereby reducing low‑level cloud cover and allowing more solar radiation to reach the surface. The observed negative correlation between CR flux and temperature aligns with this mechanism, especially in the cold season when cloud radiative effects are most potent.

Limitations and future directions

The study’s spatial coverage is limited to two stations, which may not capture regional heterogeneity across Portugal. Moreover, only NAO is considered as a large‑scale atmospheric driver; other indices such as ENSO, the Arctic Oscillation, or the Atlantic Multidecadal Oscillation could also modulate Portuguese climate and were omitted. The reliance on linear statistical models may overlook non‑linear interactions and threshold behaviours that are common in climate systems. The authors recommend expanding the network of temperature records, incorporating additional atmospheric and oceanic indices, and applying machine‑learning or non‑linear dynamical models to better capture complex dependencies. Integrating explicit representations of UV‑driven stratospheric heating and cosmic‑ray‑induced cloud processes into climate model simulations is also suggested.

Conclusion

The analysis demonstrates that Portuguese temperature variability over the past 140 years is not solely driven by atmospheric circulation (NAO) but also exhibits statistically significant links to solar activity, particularly during the winter months. Positive correlations with sunspot numbers and negative correlations with cosmic‑ray flux are consistent with proposed physical mechanisms involving UV‑induced stratospheric‑tropospheric coupling and cosmic‑ray modulation of cloudiness. While the solar signal is modest compared with internal climate variability, its presence on decadal to centennial time scales underscores the importance of accounting for solar‑climate interactions in long‑term climate assessments and in the development of more comprehensive predictive models.