1. Introduction
Management of the effect of fires on the supply of water to regional communities requires knowledge of their effect on forest structure, soils, and hydrogeology and on the local and regional water balance. In January of 2017, wildfires burned a few more than 550,000 hectares of mostly forested land in Central Chile [
1]. The most intense and damaging fires occurred in the Maule Valley, east of the coastal city of Constitución, and have become known as the Las Máquinas fires. While they were historically severe, it is expected that climate change will further increase the frequency and intensity of wildfires in Central Chile [
2,
3]. These fires occurred principally in the coastal range of Central Chile. The Coastal mountains, while not particularly high, are a steep and rugged landscape that is forested and dotted with small, rural communities. The forests are a combination of commercial plantations of
Pinus radiata and
Eucalyptus spp. and the fragmented and vulnerable Roble-Hualo native forest which is named after the deciduous
Nothofagus species that dominate the overstorey. The plantations provide employment for rural Chileans while the native forest has high conservation value. The water that flows to the Maule River is a regionally important supply of drinking water and irrigation and is important to the many rural communities in the ranges.
The effect of fire on water balance depends upon the combination of many factors, including the rainfall regime before and after the fire and the effects of the fire on vegetation and the soil micro- and macro-structure. Central Chile, where the fires occurred, has a Mediterranean climate with an annual dry season extending from November to March or April. The 8 months that preceded the fires were unusually dry, in particular between July and October. The lack of the winter rains and a very hot start in January created the risk of severe fire conditions [
4]. The climatic conditions that increased the fire risk also created a dry catchment with historically low moisture storage [
5]. The fires further changed the local controls on the water balance by causing a rapid change in leaf area and, therefore, transpiration and interception [
6,
7,
8]. The combined effect of this period of drying, followed by a rapid reduction of cover may have caused complex and hard to predict changes in catchment storage and hydrology. Therefore, understanding these hydrological changes after fires is a key factor in water resources management [
9].
One measure of the hydrological behavior within a catchment is the Mean Transit Time (MTT) of water. MTT is the average time of a water molecule traveling through a catchment, from when it enters the system to when it exits at any point in the stream as discharge [
10]. Therefore, MTT integrates all the hydrologic processes in a single measure of catchment behavior. Tritium activity (TU, where 1 TU represents a
3H/
1H ratio of 10
−18) in water has been used to estimate the MTT of water in forested catchments worldwide (e.g., [
11,
12]). This has improved the understanding of how, and from which sources, the hydrological system is recharged (e.g., [
13]). To date, tritium has not been used to investigate changes in MTT after a wildfire or in forested catchments in Mediterranean climate areas, such as drought-prone Central Chile. As noted above, fire destroys or damages the canopy, can change soil water repellency, and can open or close soil macropores. Combined, these changes may result in profound hydrologic change and affect the rate of groundwater recharge, streamflow, baseflow, and soil moisture storage [
14]. Moreover, extremely dry conditions lead to decreasing stored groundwater and soil water [
5] and interact with the fire in the affected catchment. How these hydrological processes interact to reach a new state depends on each catchment, but site monitoring, water age estimation, residence time, and hydrological models are all helpful to support observations and planning.
As well as affecting the vegetation, and therefore transpiration and throughfall, wildfire can also alter important soil properties that can have implications for the water cycle [
15]. Commonly reported changes after a fire include an increase in peak flows and storm flows and a severe reduction in baseflow, baseflow recession, and low flow [
16,
17]. These changes in streamflow generation processes have been attributed to the formation of a hydrophobic layer [
18] or soil sealing [
19]. There have been a large number of investigations of the effect of fire on streamflow dynamics (e.g., [
20,
21]) that have applied a range of methods including cluster and regression analysis [
22], runoff generation at hillslope scale [
23], chemical analysis for groundwater interaction [
24], geochemical and end-member mixing analysis [
25], isotope mass balance methods [
26], catchment runoff modelling (e.g., [
27]), and paired catchment experiments [
28,
29].
While there have been many studies of the effects of fire on forest hydrology, to date, none of these have related the concentration of tritium to pre- and post-fire streamflow. McDonnell et al. [
30] discussed the limitations of standard hydrography and hydrometric for understanding sub-soil hydrological processes. McDonnell and Beven [
31] argued for the use of tracers in hydrology to complement hydrometrics and allow the more complete investigation of changes to hydrological processes caused by disturbances such as wildfire. Thus, this paper aims to (i) determine if we can detect changes of catchment groundwater water transit that may be ascribed to the impact of the fires, and (ii) unveil changes in hydrologic variables (i.e., runoff and baseflow). Within the area that was burnt in January 2017 in Central Chile [
1], there are three experimental catchments where high and low baseflow tritium concentration of stream water has been monitored since 2009. All three catchments were completely burned by this high severity fire. This study reports tritium isotopic composition of water in the stream (high and low baseflow) and rain in these burnt catchments before and two years after the fire and uses these data to test the hypothesis that the MTT of water in these catchments is not affected by wildfire.
4. Discussion and Conclusions
In this paper, we used a combination of streamflow, rainfall, and tritium data from three catchments to test whether the Mean Transit Time of water through these catchments was unaffected by the ‘Las Máquinas’ wildfire of late January 2017. This was proposed despite previously reported changes in the peak and low flows in at least one of the catchments [
32]. This expectation was based on two pieces of evidence. Firstly, the pore flow pathways were not affected by the fire (e.g., [
60]) and, secondly, the peak and low flows reflect the celerity of water travel while the MTT is determined more by the bulk velocity of water movement [
31]. The data support this hypothesis and indicate that the effect of the fire on streamflow magnitude varied between the three catchments, but that MTT was fairly constant or even increased slightly. The poor relationship between measures of flow and the MTT of water for the Quivolgo catchments and the observation that MTT was not greatly changed after the fire suggests that, for these catchments, fire has affected storage and celerity but has not affected the velocity of water in these catchments. This change in the balance between celerity and velocity likely results from changes in infiltration, storage, and drainage in these catchments.
The megafire burned a significant amount of litter and humic soil within the
N. glauca forest in the Quivolgo area [
61]. These findings together with the fact that rainfall is not generating high peak flows in Q2 suggest that deep soil infiltration may have been enhanced within the first year after the fire. This enhancement may be due to soil cracking during the drought [
62], particularly after the removal of the leaf cover, and/or to the rapid development of vacant old root channels as preferred infiltration pathways after the fire. (e.g., [
63,
64]). Additionally, according to [
20], evapotranspiration from Q2 accounts for a small proportion of rainfall and, because streamflow has not increased, there is a high probability of this unaccounted water being stored deeper in the soil where it is mixing with stored water but not yet affecting the average water age.
The peak flow in Q2 after the fire was much lower than Q1 and Q3, and much lower than before the fires. In [
20] it was shown that after evapotranspiration of native forest in Q2 was accounted for, the reduced streamflow still left a net unaccounted water balance of 915 mm in 2017 and 421 mm in 2018. This suggests that the system has changed to one in which flow has decreased, even though the rainfall increased. It is an unusual behavior but is not without precedent (e.g., [
64,
65]). In [
20], it was hypothesized that before the fire, infiltration had been blocked by a combination of filled root channels and very dry soil, whereas after the fire these potential preferred pathways opened, increasing the infiltration to the deep fractured rock system.
Runoff and water yield seem not to have an evident association with MTTs. In Q1, where runoff and water yield were not as high as Q2 before the fire, indeed virtually zero in 2015 and 2016, MTTs were shorter than the native forest. MTT in Q3 was similar to Q2 both before and after the fires (
Table 6) and more than 50% greater than Q1. Stormflow response was greatest in Q3, before the fires, and was 1.5 times that of Q2. After the fires, when all catchments were flowing, the Q1 response to the storm was 300 times that of Q2, and Q3 was 155 times that of Q2. It may be that catchment volume and size is playing an important role in water age and streamflow. Q2 is the biggest catchment, followed by Q3, which is one explanation for why transit times are shortest in Q1 (the smallest catchment) and longest in Q2. However, it is out of our study scope, and in the future, we plan to explore catchment geomorphology effects on MTT.
After the fires, MTT in Q1 increased. However, the native forest catchment, Q2, MTT seems unaffected, but runoff and water yield were reduced. In Q2, the lowest monthly streamflow in the 2009–2016 period was 9.6 mm in 2014. However, the lower monthly streamflow post-fires were 12.0 mm in February 2017, indeed stream flow seems less variable through the seasons. Precipitation occurs mainly in winter, when the main species (
N. glauca) drops all its leaves, therefore, more water gets into the soil, there is lower evapotranspiration and, while we would expect more water to be stored in the soil [
66], often this is also the period of highest streamflow.
After fires, in Q1 and Q3, water ages increased which may be due to (1) more new water infiltrating deeper in the regolith and not passing through the soil matrix to the stream [
67], or (2) new water entering the system pushing old water within the macropores to streamflow [
68,
69]. Thus, two different mechanisms could produce a similar result in the stream. Peak flows increase following a fire are often observed and usually ascribed to increased hydrophobicity and lower ET; in our case, this occurred in Q1 and Q3 but not Q2. For instance, Q3 BFI in 2015 and 2016 was 0.69 whereas in 2017 it was 0.51 which means half of the flow was surface runoff. Particularly in Q1, in the first-year post-fire surface runoff was almost 70% of the total flow, while in the second-year post-fire that surface flow dropped to only 34% of streamflow. However, in the first year after the fire, the Q1 catchment started to flow in May–June 2017 in a year with above the average rainfall (1400 mm), and some soil disturbing phenomena such as hydrophobicity, soil sealing, hyper-dry conditions, could also explain the high runoff. Nonetheless, the second-year post-fire BFI increased suggesting new water displacing older water to the stream.
An earlier study at these sites in 2015 [
70] found transit times between 9 and 15 years (Q1 and Q2). In Q1 we have not found a significant difference in MTTs over time, however, this catchment went dry for a period, and we might expect older water in the stream. However, there is still little difference between MTTs before and after fires.
With almost 20 ha of native forest, Q3 seems to behave similarly to Q2, as we can see in the similar MTTs in summer 2014 and no statistically significant differences between these two catchments. This might be due to the similarities in the native forest composition and cover in Q3 and Q2, particularly in the wet zones near the streams. Moreover, in the spring of 2014 within all catchments, MTTs were at their lowest in the study period. In this pre-fire sampling, 1250 mm of rain had fallen in that year, the second wettest after the 2017 rainfall and it is possible that such an amount of water has mixed with older water decreasing mean water age.
When comparing source and outlet samples, the mature native forest (Q2) catchment had the greatest differences in MTTs, while Q1 and Q3 had no significant differences. In Q2, water sampled at the outlet had almost double the age of samples from the stream source in all LPMs results, despite the samples taken no more than 600 m apart. In 2018, the source sample MTT is about 8.5–11 years old, but the sample at the outlet is about 22–15 years old (
Table 6). In Q1, we could not identify a difference between the age of source and outlet samples. A mixing process might be occurring in the soil, where new water is being mixed with the old due to an increase in infiltration rates. However, there could also be a bias of the methods with a probability of underestimating the water age due to limitations of the LPMs [
71].
Cartwright and Morgenstern [
39] found that the MTT differences between catchments can be related to the pattern of evapotranspiration. However, in this study, it was not possible to measure evapotranspiration after the fire within the pine catchments and was possible only in the native forest [
20]. After fires, both Q1 and Q3 have increasing runoff (
Table 3), and water is getting older probably because after the initial increase in surface runoff component (BFI reducing) new water is entering the soil matrix displacing older water as seepage to the stream. In Q2, water flowing out at the source is around half the age of the stream water at the outlet, which could reflect longer flow paths in this catchment, a more complex groundwater flow entering along the stream at different points, or simply more subsurface flow entering the stream. Different water ages in samples from source and outlet introduce uncertainties in the calculations of MTTs [
72].
Wildfires are a well-known disturbance that affects the hydrologic cycle. With an expectation of more fires in the future [
73], how these fires affect the hydrology of a site will be of increasing concern, especially in drier zones such as the Mediterranean coastal zone of Chile. Age tracers, such as tritium, can help us understand how water moves within the different flow pathways within a catchment [
74]. After 10 years of drought and with a mega-fire in 2017, MTTs have not materially changed. Sampling both the source and at the outlet, showed essentially no difference in MTT in Q1, a little difference in Q3, and a significant difference in Q2. Within the following years from this study, a sampling schedule to continue to investigate both the long-term drought and the effect of wildfire in these catchments will be maintained.
Finally, to improve our knowledge of our sites and in conjunction with hydrometrics and tritium sampling, a geophysical sampling schedule (e.g., electrical resistivity or seismic waves) and boreholes should be performed in order to check water table fluctuations and their relationship with water age and other hydrologic variables. Moreover, a sampling schedule for other isotope sampling (i.e., deuterium and oxygen-18) and a continuum tritium sampling in water and rain schedule (days or monthly) to increase MTT parameterization within sites should be performed in the future.