16 T.A. Cinco et al. / Atmospheric Research 145–146 (2014) 12–26 extreme events is from the period 1951–2010 with the normal value based on the period 1961–1990. Percentiles were computed using all non-missing days where the 1st percentile is the 4th lowest value, and the 99th percentile is the 4th highest value. The statistical significance of the trends was calculated using the Kendall-tau test with trends noted as significant at the 95% confidence level. A statistically significant positive trend for the precipitation indices indicates an increase in either frequency or intensity based on values above the 99th percentile. The hot day index calculates the frequency of days above the 99th percentile based on the mean values for 1961–1990 whereas the cold nights index represents the frequency of daytime temperatures below the 1st percentile for the same period. For the temperature indices a significant positive trend represents an increase in the frequency of hot days and cold nights. If, as the wider regional level data suggests, there is a warming trend we would expect to see significant increases for the hot days and significant decreases (negative values) for the cold nights. Neither the temperature nor the rainfall indices are deseasonalised and although, for temperature at least, the effect of this is minimal in a tropical country such as the Philippines this fact should be considered when reviewing the results, especially those for rainfall. 4. Results and discussion 4.1. Temperature trends Fig. 2 shows temperature anomalies versus the normal value for the period 1961–1990. Mean temperatures in the Philippines begin to rise consistently from 1978 and become consistently positively anomalous in 1983 which approximately matches global trends in the second half of the 20th century (Alexander et al., 2006; Frich et al., 2002). The overall trend of temperature anomalies increases throughout the latter half of the observed period ranging from a minimum of + 0.1 °C (1993) to a maximum of + 1.0 °C (1998). 1998 thus represents the largest positive anomaly of the entire observed period which also corresponds with the declining phase of one of the most significant El Niño events with the central and equatorial Pacific. The Philippines and Indonesia in particular, suffered from widespread drought and forest fires relating to lower than average rainfall during 1997– 1998 (Moya and Malayang, 2004). Despite this overall upward trend, there remains variability within the latter half of the last century and the first decade of this. In particular, 1992 to 1995 shows a decreasing trend also detectable in global data for the same period (HadCRUT3 and GISSTemp) and which could be associated with the 1991 eruption of Mount Pinatubo in central Luzon, Philippines. This reduction of global temperatures has been linked to the ejection of sulphur dioxide into the stratosphere from Mount Pinatubo, creating stratospheric aerosols which reflected solar radiation and produced a negative radiative forcing of − 3.7 W m−2 (IPCC, 2013b). Fig. 3 shows the mean maximum temperatures during the period 1951–2010 based on the normal values from 1961 to 1990. As with the mean values discussed earlier, 1998 sees the largest anomaly reaching a maximum of + 0.9 °C versus the normal value. Over the entire period, the data shows that the anomalies are more variable, moving between positive and negative from 1951 to 2010 and the shallow line of best fit illustrates a slower rate of warming than the overall mean values. Unlike the Tmean and Tmin values (discussed below), negative anomalies continue to be observed after 1978. Fig. 4 shows the minimum temperature anomalies for the period 1951–2010 against the baseline period of 1961–1990 and demonstrates an overall increasing trend with values becoming positive in 1977 and ‘peaking’ in 1998 with a +1.0 °C anomaly. From 1996 to 2010, the end of the observed period, positive anomalies are consistently greater than 0.5 °C with 2005 and 2006 marking the peak warm years of the first decade of the 21st century. We have observed that Tmean and Tmin anomalies for the period 1951–2010 show a consistently positive trend with Tmax anomalies showing slightly less consistency but a warming trend nonetheless. The larger increase found in the minimum temperatures over the observed period suggests that overall nights (often the minimum daily temperature) are becoming warmer in the Philippines, demonstrating reduced variability and increased convergence between diurnal (normally the Tmax) and nocturnal temperatures. However, it should be noted that this could possibly be attributed to the ‘noise’ associated with urbanisation and the heat island effect discussed in Section 3.2. Tmean, Tmax and Tmin all show 1998 to be the year with the highest positive anomaly which also coincides a period of severe drought in the Philippines, often associated with a strong El Niño event. When referring to El Niño, we use the Niño 3.4 index and definition of ENSO variability offered by Trenberth (1997) in which the average sea surface temperature (SST) anomalies in the 5°N–5°S, 120°W–170°W region of the Central Pacific are used as an indicator, with positive (negative) values of ± 0.5 °C for six consecutive overlapping phases (3 month periods) beginning in the period June–August suggesting an El Niño (La Niña) phase (Trenberth, 1997). Using this definition and according to National Oceanic and Atmospheric Administration's (NOAA) (NOAA, n.d.) historic records for the 1951–2010 period under study here, the following years are considered as El Niño (La Niña) years: 1951, 1953, 1963, Table 1 Description of the extreme weather event indices used when analysing the time series data following Manton et al. (2001). Index Description Hot days index Cold nights index Extreme precipitation intensity Extreme precipitation frequency Frequency of days with maximum temperature above the 1961–1990 mean 99th percentile Frequency of days with minimum temperature below the 1961–1990 mean 1st percentile. Mean intensity of events greater than or equal to the 99th percentile each year. Mean frequency of events greater than or equal to the 99th percentile each year.