The Philippines is a Southeast Asian country comprising more than 7000 islands located in the northwestern Pacific between 4∘40′ and 21∘10′ N and 116∘40′ and 126∘34′ E (Fig. 1). The country experiences an average annual rainfall of about 2000 mm, influenced by two monsoon seasons, the northeast monsoon from November to April and the southwest monsoon from May to October (Cinco et al., 2014). About 35 % of the annual rainfall is related to TC activity, while its contribution rises to about 50 % for Luzon and decreases to 4 % for the southern island of Mindanao (Cinco et al., 2016). Part of the rainfall amount in the Philippines is of orographic nature due to north–south-oriented mountain ranges of more than 1000 m spanning the largest islands of Luzon and Mindanao (Villafuerte et al., 2014). The majority of the steadily growing population in the Philippines (101 million as of the 2017 census) lives in densely populated, low-elevation areas close to the coastlines (Cinco et al., 2014, 2016; Philippine Statistics Authority, 2017).

In total, 186 daily precipitation samples were collected from 11 March 2014 to 27 October 2015 using a Palmex collector (Gröning et al., 2012) at the Marine Science Institute of the University of the Philippines Diliman located in Quezon City, which is a part of Metropolitan Manila. The rain station was installed on the rooftop of the Marine Science Institute (14∘39′02.5′′ N, 121∘04′08.6′′ E), which is centrally situated in the campus and surrounded by trees and various green spaces. The rooftop location proved ideal for rainwater collection, as it allowed for unobstructed access to rainwater without any potential sources of contamination. Samples were collected daily at 10:00 and transferred without headspace to 30 mL HDPE (high-density polyethylene) bottles for storage prior to analysis. Samples were sent to the Earth Observatory of Singapore, Nanyang Technological University, Singapore, and were analysed for stable isotopes using a Picarro L1230-i laser spectroscopy instrument. We followed the procedures described by Van Geldern and Barth (2012) for post-run corrections and calibration. Three in-house water standards used for calibration include KONA (0.02 ‰ of δ18O; 0.25 ‰ of δ2H), TIBET (-19.11 ‰ of δ18O; -143.60 ‰ of δ2H), and ELGA (-4.25 ‰ of δ18O; -27.16 ‰ of δ2H). They are calibrated against the international reference water standards VSMOW2 (Vienna Standard Mean Ocean Water 2) and SLAP2 (Standard Light Antarctic Precipitation 2). Long-term analysis of our QA/QC (quality assurance/quality control) standards yields a precision of 0.04 ‰ for δ18O and 0.2 ‰ for δ2H. We used δ18O and δ2H to calculate deuterium excess, which is defined as d-excess =δ2H - 8×δ18O and is commonly regarded to reflect evaporation conditions of moisture source regions.

The International Best Track Archive for Climate Stewardship (IBTrACS) dataset contains global TC best-track data and is a joint effort of various regional meteorological institutions and centres that are part of the World Meteorological Organization (WMO). The data are publicly available and comprise information on a storm's eye/centre with its coordinates, wind speed, and pressure, etc., with a temporal resolution of 6 h (Knapp et al., 2010a, b; Rios Gaona et al., 2018). Apart from visualization of cyclone paths, we used the dataset to calculate the spatial distance between the coordinates of a storm's eye and our sampling site.

We used version 5 of the IMERG Final daily product (Integrated Multi-satellitE Retrievals for GPM, Global Precipitation Measurement), a remotely sensed precipitation dataset from satellites, to highlight cyclonic tracks and precipitation patterns of several TCs passing by Metropolitan Manila and to identify which rainfall events were not affected by cyclonic activity but instead were associated with local or other regional convection activities. Such a dataset is beneficial, as it provides quasi-global grid-based rainfall estimates for land and the oceans (Poméon et al., 2017). The Integrated Multi Satellite Retrievals for GPM (IMERG) dataset from the Global Precipitation Measurement (GPM) programme with a fine 0.1∘ grid size (Huffman et al., 2017, 2019) has been available since March 2014 and provides precipitation data in different temporal resolutions, such as half-hourly or daily. Such satellite rainfall data have been previously utilized to show TC tracks and related rainfall intensities (Rios Gaona et al., 2018; Villarini et al., 2011).

Daily rainfall, mean daily relative humidity, and mean daily temperature data were obtained from the Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA), which maintains a rainfall monitoring station about 2.7 km away from our sampling site. The data are freely available for the period 2013 to 2017 and can be accessed on the Philippines Freedom of Information website (https://www.foi.gov.ph/, last access: 18 October 2019).

A total of 186 daily precipitation samples were collected during the 19 months of the study period spanning from 11 March 2014 to 27 October 2015 in Metropolitan Manila. Their stable isotope compositions show large seasonal isotopic variability; δ18O ranges from 4 ‰ to -13.84 ‰, and δ2H ranges from 16.84 ‰ to -99.1 ‰ (Fig. 2). The highest δ18O value of 4 ‰ was observed on 9 April 2014 during the annual dry period, whereas the lowest δ18O value of -13.84 ‰ was observed on 16 September 2014 in association with TC activity. The mean δ18O value of precipitation at the study site is -5.29 ‰ for non-TC rain systems, while TCs, as large regional convective systems, have the potential to cause a change in δ values of up to almost 9 ‰ relative to the mean. The average δ18O value of the nine TCs that tracked within <500 km from the sampling site is -10.24 ‰ (SD of 2.11), a factor of 2 larger than the mean from non-TC precipitation (average is -5.29 ‰, SD of 2.64).

An inter-annual variation of stable isotopes in precipitation is observed in the time series of Metropolitan Manila, where the generally humid summer months are characterized by heavy rainfall and exhibit lower isotope values compared to the rest of the year (Fig. 2). The precipitation isotopes are characterized by slightly higher values during winter and spring, when temperatures and relative humidity are lower with less frequent rainfall. Especially early 2015 shows drier conditions with sporadic rainfall and relative humidity levels of about 60 % to 70 %. This is also reflected in the precipitation collected on 1 March 2015 with δ18O of 0.01 ‰ and δ2H of 9.8 ‰, respectively. Although d-excess shows relatively high temporal variability, ranging from -15.18 ‰ to 24.31 ‰, it largely clusters in a small range between 5 ‰ and 15 ‰.

Based on the daily isotope measurements of rainfall events between 2014 and 2015, we determined the LMWL (local meteoric water line) for the study site to be δ2H = 7.2674×δ18O + 5.4103 (Fig. 3), indicating that slope and intercept of the LMWL are lower due to the influence of tropical precipitation compared to the GMWL (global meteoric water line) with δ2H = 8×δ18O + 10 (Craig, 1961).

In order to assess meteorological controls on the isotopic composition of daily precipitation at Metropolitan Manila, we investigated the correlation between δ18O, daily precipitation amount, daily mean temperature, and daily mean relative humidity. Additionally, δ18O is compared to d-excess (n=187) (Fig. 4). We found that δ18O is weakly correlated to d-excess (R2=0.2187), precipitation amount (R2=0.1087), and relative humidity (R2=0.1323). No association is observed between δ18O and temperature (R2=0.0338).

In order to get further insights into the seasonal variations, we also calculated the average values for each month in the time series for every isotopic and climatic parameter, while rainfall is reported as monthly totals (Table 2). δ18O is relatively low during the summer months, for instance with -7.29 ‰ in September 2014 compared to the months of winter and spring with -0.53 ‰ in April 2014 or -0.66 ‰ in February 2015. Similarly, the monthly rainfall total is less in winter and spring with 19.2 mm in March 2014 and 29.2 mm in January 2015 compared to the summer months such as July and August 2014 with 455.4 and 420.7 mm, respectively. As mentioned before regarding the daily measurements, we also observe on the monthly scale conditions which are more humid in the summer. We investigated the relationship between the isotopic composition of precipitation (δ18O) and meteorological parameters (total monthly rainfall, average relative humidity, and temperature) on a monthly scale. δ18O and δ2H are strongly correlated (Pearson correlation coefficient) with r=0.96 (n=18, p value =<0.0001, and 99 % confidence level), whereas the relationship between δ18O and d-excess yields an r value of -0.64 (n=18, p value =0.003). A clear negative correlation was determined between δ18O and precipitation with r=-0.67 (n=18, p value =0.002) and between δ18O and relative humidity with r=-0.85 (n=18, p value =<0.0001). δ18O and temperature are not correlated with r=0.04 (n=18, p value =0.87).

A relationship between isotopic value and the distance of the TC towards the sampling site was found. The TCs' distance of up to 500 km to sampling site and the precipitation isotope value are correlated with r=0.55 (n=16, p value =<0.05, and 99 % confidence level). This relationship weakens with an increase in the distance from the sampling site: a distance of 500 to 1000 km yields an r value of 0.2 (n=19, p value =0.41); the distance of 1000 to 1500 km yields an r value of 0.18 (n=24, p value =0.40); and a 1500 to 2000 km distance results in an r value of 0.1 (n=21, p value =0.69).

Overall, precipitation isotopes associated with TCs mark the lower range of δ18O values during the study period. Especially during the 2014 season, precipitation with low isotope values mostly occurred throughout the passage of TCs. For instance, Rammasun led to the lowest δ value (Fig. 5, point a, -13.84 ‰) of the whole study period, while other TCs such as Fung-Wong (Fig. 5, point c, -12.16 ‰), Kalmaegi (Fig. 5, point b, -11.39 ‰), or Hagupit (Fig. 5, point d, -9.88 ‰) caused other negative excursions in isotopic values. The 2015 season is characterized by on average a slightly higher isotopic enrichment during the summer months with heavy rainfall. Nonetheless, a similar noticeable isotope signal is visible with low δ18O values clustered along the lower end of the sample range, for example, caused by Linfa (Fig. 5, point f, -8.5 ‰) or Koppu (Fig. 5, point i, -8.7 ‰). The other TCs that occurred during the study period and were investigated by us were Mekkhala (Fig. 5, point e, -10.77 ‰), Twelve (Fig. 5, point g, -7.7 ‰), and Mujigae (Fig. 5, point h, -7.5 ‰). However, relatively negative isotope samples (Fig. 5) also originated from non-TC rainfall systems. Those events are discussed below.

Out of the nine TCs that occurred within a 500 km radius from the sampling site, Rammasun and Kalmaegi left clearly observable, distinct isotopic signatures during their approach and dissipation, which we will therefore present in more detail in the next paragraphs. Typhoon Hagupit (Fig. 5, point d) similarly led to a clear isotopic evolution pattern during its time of occurrence in the Philippines and is shown in the Supplement (Sect. S1).

Typhoon Rammasun's rainfall intensity based on the IMERG precipitation data together with its track from IBTrACS is shown in Fig. 6a. Typhoon Rammasun stands out in our study period, as it moved straight towards the National Capital Region of the Philippines, resulting in a direct hit. Rammasun, locally named Glenda, made landfall in the Bicol Region of southern Luzon on 15 July, with wind speeds of about 160 km/h. On 16 July, it passed south of Metropolitan Manila, 50 km from our sampling site, with maximum winds of 130 km/h, gradually losing strength over land. As Rammasun approached on 15 July, the precipitation exhibited a relatively high δ18O value of -4 ‰, while rainfall was weak (Fig. 7a). On 16 July, the δ18O value shifted to -13.84 ‰, while the typhoon's track was the closest to our sampling site, and rainfall amount was high. As Rammasun moved away, precipitation isotopes became more positive, and the rainfall amount decreased. The characteristic isotopic evolution with time related to Rammasun's distance and rainfall intensities can be seen in Fig. 8a, where the different radii indicate the distance to the sampling site, and the strong isotopic depletion observed on 16 July is also evident. As Rammasun with its storm centre tracked towards the northwest and away from Metropolitan Manila, our precipitation samples were relatively isotopically enriched for the following 2 d, namely -9.12 ‰ on 17 July and -6,26 ‰ on 18 July.

Typhoon Kalmaegi, locally named Luis, was the first typhoon to make landfall in the Philippines, 2 months after Rammasun. Kalmaegi reached typhoon intensity on 13 September, making landfall the following day in northern Luzon, with maximum wind speeds of about 120 km/h. Kalmaegi tracked relatively far away from the sampling site (about 350 km), but the accumulated rainfall it produced was centred south of the track, placing it considerably closer to the National Capital Region (Fig. 6b). Despite the distance of the eye from the sampling site, a characteristic isotopic pattern was visible, with the most negative δ18O value of -11.39 ‰ on 15 September, coincident with the highest rainfall amount (Fig. 7b). The following day, δ18O values returned to higher values with the increase in distance from the eye. This is also seen in a spatial representation in Fig. 8b, visualizing the track of Kalmaegi and the respective δ18O values. Kalmaegi was first approaching the sampling site on 14 September and passed away on 15 and 16 September. The lowest δ18O value was observed on 15 September and is indicated in the figure in dark blue.

As stable water isotopes fractionate during the physical process of evaporation and condensation, they serve as effective tracers in the hydrological cycle (Dansgaard, 1964; He et al., 2018; Risi et al., 2008; Tremoy et al., 2014). Here, we have demonstrated that stable water isotopes can possibly be used to identify TC activity in the Southeast Asian region by excursions in δ18O, providing evidence and supporting the hypothesis that TCs may leave a clear isotopic signal in the Philippines. The strong isotopic depletion is due to high condensation efficiencies in cyclonic convective rainbands, leading to extensive fractionation. This is particularly pronounced in intense, large-scale TCs (Lawrence, 1998; Lawrence and Gedzelman, 1996). In the previous section, we presented our findings of precipitation isotope ratios associated with typhoon activities affecting Metropolitan Manila during the study period of March 2014 to October 2015. Based on our time series, we therefore argue that for the Philippines, the lowest measured isotope value likely indicates the occurrence of a TC, such as is the case for Typhoon Rammasun (Fig. 5). Similarly, other anomalously low δ18O values at our site are caused by TCs making landfall or passing by.

Individual TCs (Rammasun and Kalmaegi) were characterized by consistent isotopic excursions to very negative δ18O values in a range of up to -9 ‰ compared to the mean isotopic value of -5.29 ‰ (Figs. 7 and 8). A TC approaching the sampling site had relatively higher isotope values than at its later stages when it was closest to the site in Metropolitan Manila. When at its closest, strong rainfall together with increased fractionation depleted precipitation isotopes, leading to a distinct drop in isotope value. Such a strong negative isotopic shift in precipitation has been previously observed in other regions (Fudeyasu et al., 2008; Lawrence and Gedzelman, 1996; Munksgaard et al., 2015; Xu et al., 2019). As the TC moved away and rainfall intensities weakened, δ18O in precipitation became again more positive, likely due to evaporative effects (Munksgaard et al., 2015; Xu et al., 2019).

As the strongest TC in terms of wind speeds, damage costs, and fatalities, Typhoon Rammasun reduced δ18O values most during our study period, to -13.84 ‰. Similarly, Typhoon Kalmaegi led to extensive damage and caused a significantly negative excursion in precipitation of δ18O values to -11.39 ‰, suggesting that the lowest isotope values might indicate the occurrence of the strongest TC at that time at our site in the Philippines. We note that our isotopic measurements are similar to observations elsewhere. For example, the range of δ18O values caused by Typhoon Shanshan affecting the subtropical Ishigaki Island was -6 ‰ to -13 ‰ (Fudeyasu et al., 2008); Tropical Cyclone Ita led to a range of -4.8 ‰ to -20.2 ‰ in northeastern Australia (Munksgaard et al., 2015); several TCs which made landfall in Texas resulted in isotope values from -3.9 ‰ to -14.3 ‰ (Lawrence and Gedzelman, 1996); and hurricanes that affected Puerto Rico and southern Texas were found to deplete δ18O values up to -18 ‰ (Lawrence, 1998). The lowest value resulting from Typhoon Phailin on the Andaman Islands was reported to be -5.5 ‰, and Cyclone Lehar depleted the precipitation sample to -17.1 ‰ (Chakraborty et al., 2016). For TCs within a distance of up to 500 km from the sampling site at the University of the Philippines Diliman in Metropolitan Manila, we measured an isotopic range of -7.7 ‰ (Typhoon Koppu) to -13.84 ‰ (Typhoon Rammasun). Despite the overall comparability to our measurements, differences exist. The lowest values observed in some studies are considerably more negative than at our site (Lawrence, 1998; Munksgaard et al., 2015). However, we attribute these differences to a variety of features, such as the specific climatic condition at each site and differences in temperature, humidity, and altitude or latitude, which are likely contributing factors to the observed isotopic variation by altering isotopic fractionation. Further, rainout history, the location of typhoon tracks, topography, and respective strength of each TC as well as its distance to the sampling site most likely have a significant influence as well (Fudeyasu et al., 2008; Good et al., 2014; Munksgaard et al., 2015; Xu et al., 2019).

We used IMERG satellite precipitation data to assess why other very low isotopic excursions occurred on various days (Fig. 5). IMERG data with their fine spatiotemporal resolution allow for the identification of convective rainfall areas and the passage of TCs and other rain systems (Fig. 6). Our analysis shows that precipitation events with anomalously low isotope signals unassociated with TCs are largely related to local, strong convective rainfall events or large-scale and slow-moving rain areas passing over the National Capital Region. Therefore, the degree of convection is responsible for the other observed low δ18O outliers that are not related to cyclone rainfall, as strong convection and long stratiform rainfall leads to intense fractionation (He et al., 2018; Risi et al., 2008; Tremoy et al., 2014). Contrarily, we speculate that the more positive isotope values clustering along the higher end of the sample spectrum around 0 ‰ are associated with local, short convective rainfall events and light-intensity rain as confirmed with IMERG satellite precipitation data. Additionally, the PAGASA rain gauge data indicate that rainfall amounts are very low during days with such very enriched isotope samples, such as 0.3 mm/d for the highest recorded sample of 4 ‰ on 9 April 2014. Interestingly, TCs at our site were found to be related to low isotope values together with high rainfall amounts (Fig. 5), while the majority of other low isotopic values unassociated with TCs were characterized by on average lesser rainfall amounts. This possibly indicates that TCs in the Philippines, besides using for instance modern-day satellite or radar data, can be detected using these two parameters, i.e. strong isotopic depletion coupled with high rainfall amounts.

The aforementioned local convective precipitation events have the potential to induce a signal of very negative δ18O, which is not related to TC activities. We therefore label such a signal as a “false non-TC signal”, as it is induced by non-TC rainfall. This results in the fact that TCs occurring during our study period do not entirely cluster along the lowest range of isotope values as seen in Fig. 5. Nevertheless, Typhoon Rammasun caused a clear drop in δ18O and stands out in the dataset. This might be the case because Rammasun's track and heavy rainfall come in closest proximity (50 km) to the sampling site. Other TCs occurring within the 500 km radius did not lead to such a clear negative isotopic signature, likely because these typhoons did not pass the sampling site at all or because heavy rainfall occurred elsewhere within the TC rainfall system (see Sect. S2 in the Supplement for their tracks and accumulated rainfall areas). Some of these TCs have intense rainfall areas over other parts of the Philippines and are characterized by a variable track, likely influenced by land interactions. Land interaction reduces TC strength and can lead to rainout due to orographic effects induced by the north–south-oriented mountain ranges (Park et al., 2017; Xie and Zhang, 2012; Xu et al., 2019). Especially Typhoon Koppu rained out before making landfall and abruptly changed its track, instead of passing by Metropolitan Manila. Similarly, Typhoon Mekkhala's intense rainfall occurred along the eastern coasts before it started to dissipate. Evidently, due to these factors the isotope values associated with those TCs were not as negative as during Rammasun. Therefore, a TC, which is relatively far away from the sampling site, produces an isotope signal that is not as clear and as negative, thus averaging out between the other low values from rain systems unassociated with TC.

δ18O, δ2H, and the second parameter of d-excess all show seasonal variabilities and are influenced by several climatic factors, including precipitation amount, temperature, and relative humidity. The scale of their influence varies depending on daily or monthly values. The results indicate that δ18O on daily levels is not influenced by temperature, relative humidity, or precipitation amount (Fig. 2) as drivers of isotopic variability. Instead, we speculate that other processes, such as large-scale convection and processes at the moisture source region, might influence stable isotopes of precipitation at our study site (Conroy et al., 2016; He et al., 2018; Kurita, 2013). Interestingly, δ18O is not affected by precipitation amount on short timescales (Fig. 4), which has also been previously confirmed in other tropical regions, suggesting that the tropical amount effect is not reflected on daily timescales (Belgaman et al., 2016; Dansgaard, 1964; He et al., 2018; Kurita et al., 2009; Marryanna et al., 2017; Permana et al., 2016). However, comparing monthly δ18O to δ2H and d-excess and to monthly average precipitation, relative humidity, and temperature, the results are clearly different (Table 2). These monthly observations show close relationships with each other; especially δ18O and precipitation amount are linked (see Sect. 3.1). The close relationship between these two parameters can be attributed to the tropical amount effect (Aggarwal et al., 2012; Bowen, 2008; Conroy et al., 2016). The relatively close relationship with r=-0.67 between monthly δ18O and monthly total precipitation might be likely due to the influence of regional convective activities on the isotopic composition of precipitation (Bony et al., 2008; He et al., 2018; Moerman et al., 2013; Risi et al., 2008).

Our observations provide details on the spatial distance from the collection site towards TCs' centres, as our findings indicate that the distance from the storm's centre to the sampling site impacts the isotopic value (see Sect. 3.1). This suggests that a TC more than 500 km away from the sampling site has no influence on precipitation isotopes (Munksgaard et al., 2015). Thus, the closer the TC is to the sampling site, the more negative the isotope signal and the larger the δ change are. This relationship might provide information on storm structure and intensity, as the intensity increases with the proximity of the TC to the sampling location. We thus confirm that the isotope value at our location is a function of the closest approach of the storm's centre to the sampling site (Lawrence and Gedzelman, 1996).

Figure 9 displays all the precipitation samples associated with TC presence and activities within a 2000 km radius from Metropolitan Manila and further highlights the relationship between distance and isotopic depletion, additionally providing a spatial indication of a TC's quadrants and its tracks relative to the location of the sampling site. The strongest depletion occurs within the 500 km radius. However, two relatively negative outliers are located within a 1000 to 1500 km radius in the northwest quadrant (see points a and b in Fig. 9). These two samples were taken during the passage of Tropical Storm Kujira on 22 and 23 June 2015 (Fig. 5), which was more than 1000 km away from Metropolitan Manila travelling east along the coast of Vietnam as seen with IBTrACS data. We investigated these two samples with IMERG satellite precipitation data and identified them as a part of a mesoscale system, with strong convective cells delivering intense rainfall, leading to distinct isotopic depletion and inducing a false non-TC signal of a very negative δ18O value, which is not related to TC activity.

IMERG satellite precipitation data also reveal that the highest rainfall intensities occur at the left side of the TC track for all the TCs within the 500 km radius, except for Hagupit and Mekkhala, which are more complex cases (Fig. 6a and b and Supplement Sects. S1 and S2). This is in contrast to the results from Villarini et al. (2011), who found that the largest rainfall accumulation appeared on the right side of the hurricane tracks. They also noted that large rainfall amounts occurred far away from the storm's track, which we can confirm and quantify with our observations. The largest rainfall totals vary in a range of 50 to 150 km away from the storm's centre depending on the TC. For Kalmaegi the intense rainfall areas are up to 150 km away from the storm's centre. These areas with the highest rainfall totals should most likely coincide with the most negative isotope value, indicating that the strongest depletion occurs in the outer cyclonic rainbands. This is consistent with previous findings (Gedzelman et al., 2003; Lawrence and Gedzelman, 1996; Munksgaard et al., 2015). However, Fudeyasu et al. (2008) observed the highest isotope values in the inner eye wall, i.e. in close proximity to the storm's centre. We could not investigate this further, as no TC passed by our site in a distance of about 20 km, which is the size of a typical typhoon's eye (Weatherford and Gray, 1988).

Isotope proxies from paleoarchives such as tree rings and speleothems have been utilized to reconstruct past cyclone activities (Frappier, 2013; Frappier et al., 2007; Miller et al., 2006; Nott et al., 2007). For instance, stalagmites yielded a record of weekly temporal resolution with negative isotopic excursions related to TC activity (Frappier et al., 2007). Such a high temporal resolution from stalagmites makes our in situ measurements very comparable, highlighting the potential to use both in conjunction. Similarly, high-resolution tree-ring isotope analysis identified the occurrence of Hurricane Ivan and Hurricane Frances in 2004, which both resulted in the lowest observed precipitation isotope values for that year (Li et al., 2011). Nevertheless, it is important to consider possible limitations at the study site that arise in paleotempestology, such as sea level change or the disruption of sedimentological records through floods or tsunamis. These need to be evaluated when comparing precipitation isotopes related to TCs with other proxy records such as speleothems and coastal deposits and when choosing the study area (Oliva et al., 2017). However, the aforementioned paleotempestology studies suffer from uncertainty regarding parameters such as TC intensity and distance to the storm's centre affecting the isotope signal. Our study provides further information on these parameters, as we hypothesize that immediate proximity of a TC results in very low δ18O values. Therefore, we might aid with a better interpretation of paleoarchives. Moreover, these studies are limited in number and only focus on a few regions affected by TCs, such as Central America and the southeastern USA (Frappier et al., 2007; Miller et al., 2006). However, more paleotempestology studies investigating paleoarchives related to typhoon footprints covering different regions and countries would provide a better understanding of past TC activity, ultimately resulting in better and more accurate climate reconstructions. TC projections related to climate change could also be improved, which is especially relevant for decision makers dealing with TC-related impacts and damages. Our in situ isotope measurements provide baseline data input in an understudied tropical region, providing isotopic data of TC occurrence and quantifying the isotopic depletion associated with TC activity. Further, our 19-month dataset suggests that the lowest measured isotope value at the Philippines study site is associated with TC activity, resulting in the distinct negative isotopic shift in the time series (Fig. 5). As rainout history, topography, distance of the track, or rainfall unassociated with TCs can induce a weak or false non-TC signal, it is important to choose stalagmites or trees as archives based on their location, ideally covering a spatial gradient, thus capturing a TC in its full size.