The presentation and discussion of the results are structured according to the three main goals set out in the introduction of this article. First, we analyze the fog and low clouds (FLCs) spatiotemporal distribution, characterizing its frequency of presence, seasonal and daily cycles, interannual variability and trends. Second, we discuss our results regarding the influence of ENSO on FLCs interannual variations. Finally, we consider how our results on recent spatiotemporal FLCs changes relate to the past and present spatial distribution of Tillandsia landbeckii.
Fog and low cloud spatiotemporal distribution
Seasonal and diurnal FLCs presence cycles
To assess FLCs seasonal and diurnal cycles, we derived average FLC presences from the Geostationary Operational Environmental Satellite (GOES) images for February and September. To study seasonal differences, all five-observation daytimes are considered. As expected for this latitude (~ 20°S), February (summer) presents the minimum Sc-clouds presence (e.g., ~ 15% over the ocean areas and ~ 4% over coastline areas), while September (early spring) is characterized by the maximum Sc-clouds occurrence (e.g., ~ 75% over the ocean areas and ~ 55% over coastline areas). Here, FLCs spatial distribution is highly determined by local topography considering that the prominent Coastal Cordillera (with elevations > 1300 m a. s. l.) acts as a natural barrier for fog penetration. In February, in the presence of FLCs, the base of the thermal inversion layer is higher, and the scarce FLCs summer subsistence is only occasionally able to reach farther inland (Fig. 2a). Conversely, September’s dataset shows a high FLCs presence in areas between the coast and 1000 m a. s. l., with values gradually decreasing for higher elevations (Fig. 2b). In September, the base of the thermal inversion layer is at ~ 1000 m a. s. l. (del Río et al. 2021); thus, a high percentage of the advective low stratus clouds is forced by the coastal ridges to stay at 1000 m a. s. l. or below. At the same time, the presence of fog corridors is associated with areas with lower reliefs through which the fog access inland, consistently with the observations of Farias et al. (2005). For September, Fig. 2b shows an apparent decrease of FLCs presence from West to East. Within the first ~ 15 km onshore, an area considered representative of the advective fog belt, we found an average FLCs presence of 50–20%, characterized by high spatial variability.
Figure 3 shows the FLCs presence anomalies reported in Fig. 2 for the five-observation daytimes in February and September. These anomalies resulted from the different FLCs average between the selected hour and the five-observation daytimes. In September, we observe higher FLCs presence over the ocean and the first ~ 15 km onshore at night and early morning, when surface temperatures over the ocean do not significantly contrast with the ones at land, generating calm wind conditions (Lobos-Roco et al. 2021). Within these conditions facilitating the air condensation and limiting the inland transport, the FLC deck stays steady above the ocean while it causes higher radiative cooling. During the night and early morning, the elevated MBL moisture, the ocean–land low thermal contrast and the bounded MBL depth increase FLCs formation over the ocean and onshore areas, until reaching the Coastal Cordillera that obstacles MBL influx to inland areas. At 00:39, 04:39 and 07:39, we observed a higher FLCs presence until a 1000 m a. s. l. altitude. Then, FLCs gradually disappear in the morning as solar radiation increases, dissipating the cloud presence during the daytime. The FLCs morning dissipation over the land is related to two processes. First, the increased land temperature (solar radiation) provoke a stratification in the interphase ocean–land that inhibits the MBL well-mixed (Lobos et al. 2018). Second, the diurnal circulation observed over the Atacama Desert returns the airflow described by Rutllant et al. (2003), increasing the subsidence and dry air advects and therefore contributing to dissipate the FLCs. From noon to the evening, southerly winds face those areas where the coastline has a southwest orientation and a high elevation. These conditions provoke the rise of air masses, which cool and condense forming orographic fog (Cereceda et al. 2002). This phenomenon can be observed in the September 19:39 image of Fig. 3. Areas with a regular presence of orographic fog have been associated with the location of unique fog oases (e.g., Alto Patache) (Muñoz-Schick et al. 2001).
In February, the maximum FLCs presence occurs during the night and especially at dawn (07:39) when lower air temperatures prevail. At noon, inland FLCs presence is almost null due to the desert heating. At the same time, very low FLCs are observed in the offshore area close to the coast. According to Rutllant et al. (2003), the FLCs absence during the noon-afternoon is caused by the enhanced air subsidence over the coast due to the convection created by the heating of the western slopes of the Andes.
Fog and low clouds interannual variability
The interannual variability of the FLCs presence is described under spatial and temporal perspectives for the sampling areas (offshore and inland, see Fig. 1c) throughout the study period (1995–2017). First, we analyzed the relationships among the FLCs presence variations of different sampling areas. Second, we observed the interannual variability, based on the FLCs presence monthly averages and expressed in standard deviation (SD). Table 2 reports the percentages of FLCs presence’s hourly and daily means for February and September, together with their interannual variability. February presents a similarly high FLCs presence variability of the same order as the mean itself for the different analyzed areas and times because of the low FLCs presence during summer. However, the lowest interannual variations are detected at 65 km offshore for all the observation daytimes, while the highest interannual variations are found in nearshore areas. During September, the FLCs presence interannual variability is significantly lower than the one observed in February. Hourly and daily means of this variability present similar patterns characterized by lower variations offshore and higher variations inland.
FLCs presence interannual variations of offshore and onshore areas are significantly correlated to each other in February. At the same time, the correlation between the FLCs presence 65 km offshore and in the other sampling areas decreases toward the coast. For example, there is a correlation of r = + 0.85 and r = + 0.62, with a 99% confidence level (p < 0.001) between observations taken 65 km offshore and 5 km offshore and 25 km inland, respectively (Fig. 4). Here, the higher altitude of the thermal inversion layer (TIL) during February generates a scarce FLCs presence and the influence of the relief over FLCs spatial distribution is lower. In Fig. 4, February’s results show that the highest values for FLCs presence occur in the years 1998, 2003, 2005, 2010, and 2014–2016, while the lowest ones are observed in the years 1999–2002, 2008, 2011–2012 and 2017.
For September, no significant correlations emerged between FLCs presence in onshore areas (especially those located 15 km and 25 km inland) and offshore ones. Moreover, minimum and maximum FLCs frequency years do not correspond for onshore and offshore areas (see Fig. 4). September also shows a large interannual variability, but a clear FLCs presence trend is observed at the same time. This trend moved from a period of lower (1995–2000) to one of maximum (2007–2012) FLCs presence, and finally decreased (2013–2017) (Fig. 4).
FLCs interannual variations observed for different daytimes in February (Fig. 5) show slight differences compared to the FLCs daily mean (Fig. 4). Results referring to all the observed daytimes show that the most significant FLCs average variability is found in areas close to the coast. In the ocean (20 and 5 km offshore areas), the highest FLCs interannual variability is detected at 04:39 and the lowest at 19:39. Over the coastline and 5 km onshore, the highest FLCs interannual variability occurs at noon and the lowest is again at sunset. This pattern recurs for 15 and 25 km inland areas. The years in which maximum and minimum FLCs presence occurred are characterized by a high agreement among their FLCs daily observations. As expected, the FLCs interannual variability observed for the five selected daytimes in September shows similar patterns compared to the correspondent daily mean, with a lower relative variation over the ocean and near shore areas, and higher relative variation in the inland ones (see Fig. 5). Despite the FLCs variations characterizing the sampled daytimes, the variability was higher at noon and sunset for all the sample areas, particularly from 5 km offshore toward the inland. Conversely, lower FLCs variability was observed at dawn and night.
Fog and low clouds trends
The FLCs satellite identification for September and February within the study period (1995–2017) provides us evidence to determine FLCs presence tendencies. Based on the cluster areas (shown in Fig. 7) and using daily averages, it is possible to observe differing trends between observations corresponding to the ocean surface, nearshore areas (both onshore and offshore), and inland ones.
In September, by applying linear regression, FLCs daily presence average changes over the ocean (cluster 1) showing a weak, yet statistically significant, positive trend (1%/annual; p < 0.001). The same trend is also present in cluster 2 (0.9%/annual; p < 0.001), which mainly covers the nearshore areas, both offshore and onshore. This trend reveals that the average FLCs presence in clusters 1 and 2 increased by ~ 20% between 1995 and 2017. In the onshore and inland areas (clusters 3 and 4), the trend showed again to be weak and positive (0.8%/annual; p < 0.002). However, cluster 4 is not expected to be statistically significant due to the continuous very low FLCs presence in this area.
Figure 6 shows the FLCs presence trend based on monthly averages for the study period, using the Theil–Sen method for a pixel-by-pixel scale of analysis. Within this figure, differences are visible between ocean areas, nearshore areas and inland ones. Figure 6 also shows that between the coastline and ~ 1000 m a. s. l., a weak-positive trend prevails, whereas above this altitude (1000 m a. s. l.) some limited areas present a weak-negative trend (e.g., over the Oyarbide Tillandsia field). These results, consistent with those described by Schulz et al. (2011a), highlight the interplay between FLCs presence trends, the thermal inversion layer altitude and the topography of the study area. These findings are particularly relevant when considering that the areas characterized by a decreasing FLCs trend, corresponding to the location of Tillandsia fields, might be affected by a decreasing presence of fog, provider of water and nutrients inputs. Figure 6 also reveals that the areas with a weak and negative trend are not found along the entire 1000 m contour line, but rather in those areas where the altitudes near 1000 m a. s. l. are closer to the coast and with an S-SW orientation, opposite to the predominant wind direction (Muñoz and Garreaud, 2005). The latter highlights the influence of these factors on fog occurrence and, consequently, the presence and geographical distribution of Tillandsia fields (Pinto et al. 2006).
In February, a higher FLCs fragmentation indicates low FLCs presence and less regular cloud deck formation. This results in an almost null FLCs presence trend over the study period, with no statistical significance in any studied areas (ocean, nearshore and inland ones).
The null FLCs presence trends detected during February within the different cluster areas are consistent with those observed by Muñoz et al. (2016) at the Iquique Airport, where FLCs summer presence results equivalent to its average over the last 30 years. A similar condition occurs during winter–early spring when we detected a weak-positive trend over the ocean and in nearshore areas, consistently with the Airport observations. Moreover, in agreement with our results, Seethala et al. (2015) showed an increase of the low clouds annual average trend in the southeast Pacific (SEP) for 1984–2009 based on global satellite data. Despite the accordant results obtained by the latter study, these should be interpreted with caution, as the research mainly addressed low clouds over the Pacific Ocean on a synoptic scale, not necessarily capturing the coastal-land influences on these clouds. Conversely, the studies of Schulz et al. (2011a) and Quintana and Berrios (2007), also based on the Iquique Airport observations, show negative trends in cloud cover. However, these researchers applied different methodological approaches, based on observations made at specific daytimes, and using different criteria for what we have defined as “low cloud” in this article. On the other hand, they analyzed a more extended time series compared to our study, including the decade of the 1970s when a significant increase in low cloud presence occurred.
We have not found prior results to compare with the FLCs cover weak-negative trend that we observed above ~ 1.000 m a. s. l. in the Coastal Cordillera during September. In this month, while FLCs presence increases in the ocean and nearshore areas, it decreases in higher reliefs. In the area next to the Oyarbide Tillandsia fields, del Río et al. (2021) estimated a mean altitude of the thermal inversion layer of 1060 m a. s. l. with a 1000 m a. s. l. mode and ± 150 m of variability for September 2016 and 2017. Such results, derived by the analysis of satellite images, suggest a decreased FLCs presence for these areas. However, local factors such as topography, wind speed and wind direction can influence fog distribution in ways that are still not sufficiently understood. Furthermore, these factors could counteract a decline in the thermal inversion layer. The dieback in fog-dependent vegetation observed by Schulz et al. (2011b) is a possible evidence of this negative trend since less fog coverage leads to less water input and a consequent decreasing presence of these species.
The resulting trends must be interpreted in a decadal time frame during which La Niña cold conditions have characterized the Pacific basin considering that both the Interdecadal Pacific Oscillation (IPO) and the Pacific Decadal Oscillation (PDO) indexes show a cooling climate since the late 1990s (Sohn et al. 2013; Witiw and LaDochy, 2015). La Niña pattern entails oceanic-atmospheric conditions characterized by an intensification of the SEPA, which is related to higher air subsidence in the subtropical region, a strengthening of the thermal inversion layer, a shallower MBL, and cooler temperatures (Garreaud et al. 2008; Muñoz et al. 2016). Various authors have observed physical variations caused by La Niña. For example, Falvey and Garreaud (2009) and Vuille et al. (2015) documented a cooling of surface temperatures along the northern Chilean coast during the last decades. This phenomenon is associated with a strong vertical stratification of air temperatures over central and northern Chile and implies an intensified thermal inversion layer, determining increased temperatures above and cooler ones below it. These conditions are propitious for FLCs formation and therefore consistent with the tendency we observed.
Additionally, we must consider the influence of global warming on these trends. Seethala et al. (2015) suggest that the observed SEP FLCs formation trends may be related to natural variations and not necessarily to systematic responses to global warming. Contrarily, Vuille et al. (2015) suggest that the time and magnitude similarities between temperature changes and trends in the Andes and the west coast of South America are unlikely to occur due to natural cycles only and are better correlated with an average rise in global warming. Although various climate models’ projections for the coastal Atacama do not correctly detect the recent MBL cooling (Falvey and Garreaud 2009), nor properly characterize its physics and dynamics (Seethala et al. 2015), climate models agree that the temperature over the thermal inversion will continue to rise (Vuille et al. 2015) and that the SEPA will continue to strengthen (Falvey and Garreaud., 2009). Higher temperatures above the thermal inversion layer, together with an enhanced SEPA, would imply a strengthening of the inversion layer itself, leading to increased stability between the atmosphere layers. Consequently, favorable conditions for the FLCs formation might persist or increase. This assumption could only be confirmed by analyzing longer time-series datasets.
ENSO influences on FLCs variability
In this section, we describe and discuss ENSO influence on FLCs presence and the MBL growth in summer and winter, two seasons corresponding to the lowest and the highest FLCs occurrence, respectively.
ENSO influences on FLCs presence
To analyze the SST influence on the FLCs formation, we assessed the relationship between ONI (an indicator of ENSO phases) and the estimated FLCs presence within the cluster areas (indicated in Fig. 7). February FLCs interannual variability is closely related to ENSO 3.4 SST anomalies, especially in cluster areas 1 and 2, which correspond to the ocean and onshore areas close to the coast (Fig. 7a). The results related to these areas show significant correlations between FLCs presence and ONI (Pearson’s: r = + 0.66 and r = + 0.68, respectively, with a 99% confidence level; p < 0.001; Kendall’s rank: τ = + 0.36; p < 0.02 and τ = + 0.49; p < 0.002, respectively). The results concerning cluster areas 3 and 4 do not show the same significant correlations, which is to be expected due to a very-low-to-null FLCs presence. We observed positive correlations between FLCs presence and ONI with the ENSO warm phase (El Niño), which is related to higher FLCs presence over the ocean and coastal areas. Conversely, under an ENSO cool phase (La Niña), we detected a lower FLCs presence. The maximum FLCs presence occurred in ENSO warm phases characterized by different intensities, for example, the ones reported in the years 1998 (strong), 2003 and 2005 (weak) and 2010 (moderate). On the other hand, minimum FLCs presence occurred during ENSO (−) years which also present varying intensities, for example, the ones of the years 1999–2000, 2008 and 2011 (Moderate), 2001 and 2012 (weak), and even of some neutral years, such as 2002 and 2017.
Despite the high correlation between ENSO and FLCs presence during February, there are few years when the same correlation is not significant, although this result varies according to different cluster areas. For example, cluster 1 shows a very high FLCs presence during 2014, not seen in other years with a neutral ENSO phase. However, FLCs presence significantly decreases in cluster 2, showing values more in line with FLCs frequencies characterizing a neutral ENSO phase. In 2016, despite a very strong ENSO warm phase, FLCs presence was not as significant as during other ENSO warm years.
We performed a linear regression for all the analyzed years to quantify the relationship between the ENSO 3.4 zone and the coastal Atacama FLCs presence in clusters 1 and 2 for February (see Fig. 8a). This regression indicates that ENSO anomalies explain ~ 50% of the interannual FLCs presence variance during summer (p < 0.001). Results related to cluster 1 present a determination coefficient (R2) of 0.36, but if the year 2014 is not considered, this coefficient increases to R2 = 0.51. For cluster 2, this coefficient amounts to R2 = 0.49 (included the year 2014). Such a significant correlation suggests the existence of a potential causal relation between ENSO phases and coastal Atacama FLCs presence during February (summer), both offshore and onshore.
During September, the interannual FLCs presence variability does not significantly correlate with ENSO SST anomalies in zone 3.4. Correlation and regression analyses led to null or statistically insignificant results. The observed FLCs presence tendency (see Fig. 4) is characterized by a first period of lower FLCs presence, which then increases until reaching a peak between 2007 and 2011 approximately, and finally decreases until 2017. Observing this tendency, we derived two conclusions: (1) the trend is nonlinear; (2) at the decadal scale, the Pacific Ocean presents long-term SST oscillations which cycles can last from 20 to 30 years.
One of the indices that estimate the SST decadal variability is the Interdecadal Pacific Oscillation (IPO) (Power et al. 1999). In this sense, the observed FLCs trend presents very similar patterns to IPO during the studied period (Meehl et al. 2021).
During recent decades, a La Niña tendency has predominated (Falvey and Garreaud 2009; Schultz et al. 2011a; Muñoz et al. 2016) with a peak around 2010–2011 coinciding with the maximum FLCs presence of the study period. Currently, this condition is beginning to shift toward an El Niño cycle (Meehl et al. 2021). Within this context, and intending to extract the decadal climatic signal from September FLCs interannual variability, we normalized both variables (the FLCs presence estimated with GOES and the IPO) based on the unfiltered monthly time series (Henley et al. 2015). Additionally, we subtracted the low-frequency IPO signal from the FLCs presence normalized record. Finally, we performed statistical analyses of the high-frequency variability of the FLCs record for the different cluster areas. The results of the correlation analyses show a strong and significant inverse relationship between FLCs and the ENSO SST anomalies in ENSO zone 3.4 (see Fig. 7b). ONI shows a significant and strong inverse correlation (r = −0.69; r = −0.70; r = −0.78 and r = −0.84, for cluster 1, 2, 3 and 4, respectively) with a 99% confidence level. Despite these statistically significant results, the ones related to cluster 4 should be cautiously interpreted due to its high interannual variability (of ~ 50%) caused by the very-low-to-null FLCs presence characterizing these areas.
Figure 7b shows that the ENSO cool phase is related to higher FLCs presence over the ocean and coastal areas. Conversely, under an ENSO warm phase, we observed a lower FLCs presence, revealing an opposite response to ENSO phase anomalies between summer and winter–early spring. Despite the resulted high correlations, the relationship between the interannual FLCs variations and ENSO phases present observable magnitude differences between areas of oceanic predominance (clusters 1 and 2) and inland ones (clusters 3 and 4). This correlation is stronger in nearshore and inland areas. In the first 2 clusters (Fig. 7), the maximum FLCs presence occurred in ENSO cool phases characterized by different intensities, as in the years 1995–1996 (weak-neutral), 1998–1999 (moderate), 2000 (weak) and 2010 (strong), and even in the neutral years 2005 and 2016–2017. In clusters 3 and 4, the maximum FLCs presence occurred during these same years, with the addition of high FLCs events also in 2007–2008 (moderate-to-neutral ENSO phases). On the other hand, the minimum FLCs presence occurred during ENSO warm years, once more characterized by varying intensities. In this case, it is worth mentioning the FLCs occurrence in different cluster areas for the years 2002–2004, 2006 and 2009 (moderate-to-weak ENSO intensity), together with 2015 (very strong ENSO intensity), the year presenting the lowest FLCs presence, and also some neutral years, such as 2012–2013 and 2014.
To study the described correlation between the ENSO zone 3.4 and FLCs presence during September in the coastal Atacama’s different cluster areas, we fitted a linear regression for all the analyzed years (see Fig. 8b). This regression indicates (p < 0.001) that ENSO anomalies cause almost 50% of the variance in the interannual FLCs frequency during winter–early spring in the oceanic clusters (1 and 2), with an R2 of 0.48 and 0.49, respectively. For clusters 3 and 4, R2 significantly increases until reaching values of 0.61 and 0.71, respectively. Among these, cluster 3 corresponds to the most interesting area in terms of fog presence, variability, and Tillandsia presence. Here, ENSO phases correspond to 61% of the FLCs interannual variance, suggesting the existence of a potential causal relationship between ENSO (warm and cool) phases and the high FLCs season in the coastal Atacama. This assumption is particularly relevant with regard to the potential effects of the ENSO phenomenon on fog-dependent ecosystems.
We performed statistical analyses to describe the relationships between interannual FLCs variability and SST anomalies in areas closer to the Atacama for February and September. In this way, we correlated FLCs presence with SSTs for ENSO zone 1 + 2 (coastal Equatorial Pacific) and along the coast of the Atacama. For doing it, we used SST monthly averages estimated by Vargas (2019) based on MODIS product (L3) for the period 2002–2017 (the average SST was extracted for the area between the coastline and 120 km offshore, referring to the northern and southern boundaries of the processed GOES imagery). Our analysis revealed no significant relationship between FLCs (with and without the IPO decadal sign influence) and the ENSO zone 1 + 2. Likewise, no significant relationship was identified between FLCs and Atacama SSTs at the different analyzed daytimes. Therefore, during February and September local SSTs have no direct association with FLCs in the coastal Atacama. This finding emphasizes that FLCs formation is primarily related to regional rather than local dynamics. del Río et al. (2018) found a low (r = −0.40) but significant relationship between the local SST and the fog water yields collected in the Alto Patache station, especially during summer and under ENSO ( +) conditions. Our analysis suggests that the relationship between coastal SST and fog water is mainly explained by two interactions. First, the coastal SST affects the MBL temperature budget (Vuille et al. 2015) and its vertical structure (Garreaud et al. 2008). Second, this MBL structure affects the Liquid Water Content (LWC) of the Sc-cloud since cold (warm) conditions lead to low (high) LWC. This alteration affects the fog water yields significantly.
ENSO phases influence on the marine boundary layer growth
To complete the study of the FLCs presence interannual variability at the coastal Atacama Desert, we analyze the vertical structure of the marine boundary layer during different ENSO phases.
Figure 9 shows the factors affecting the MBL structure: SST, mean MBL potential temperature (θ), inversion layer (∆θ), free troposphere thermal gradient (γθ) and boundary layer height (h). These data were measured at Antofagasta Airport (23.45361° S, 70.44056° W) from 1995 to 2017 during the representative summer (February) and winter (September) months under warm, cool and neutral phases of ENSO. The ENSO phases are classified by the Oceanic El Nino Index (ONI), which describes SST anomalies in the 3.4 region. We name it a “cool ENSO phase” when ONI is < −0.5 C, a “neutral ENSO phase” when ONI stays between −0.5 C and 0.5 C, and a “warm ENSO phase” when ONI is > 0.5 C.
During summer conditions (February), we do not observe significant differences for the MBL height (h) between different ENSO phases. However, during warm ENSO phases, h presents higher variability than in cool ones. As expected, during February warm ENSO phases, SSTs are 2 K higher than during cool ENSO phases and 1 k higher than during neutral ENSO ones. Even though this SST rise does not substantially alter the MBL heat budget (theta), it significantly affects the thermal inversion that results in being 1.5 K warmer than during cool ENSO phases. Finally, the free tropospheric thermal gradient observed during warm ENSO phases is slightly higher than during cool ones. During summer, the somewhat higher free tropospheric thermal gradient and inversion layer promote Sc-cloud formation and maintenance.
On the contrary, under winter conditions (September), the MBL height does not change among different ENSO phases. However, as expected, we observe an increase of 1 K in the SST during warmer ENSO phases, leading to a slightly warmer MBL heat budget. The thermal inversion layer observed during September under warm ENSO phases does not shows significant changes. However, during cool and neutral phases, the inversion layer reaches its maximums values > 12 K. Under winter conditions, during cool ENSO phases, we observed that the free tropospheric thermal gradient doubles (0.045 K/m) compared to the warm ENSO phases. This variation diminishes the exchange between the MBL and the free troposphere, promoting Sc-cloud formation and maintenance. Our results agree with the ones obtained by Muñoz et al. (2011) for the same study area, where the authors describe the climatology of the MBL from 1979 to 1997. Also, our findings partly agree with studies based on observations obtained in the Northeast Pacific Ocean (Wood, 2012; Schubert et al. 1979). However, our ABL height observations differ from these reference studies due to the strong influence of topography characterizing our study site.
To stretch the differences of the MBL structure during cool, neutral and warm ENSO phases observed in February (Summer) and September (Winter), we described (Fig. 10) the vertical profiles of potential temperature (θ) and specific humidity (q) (12:00 UTC-08:00 LT) measured at Antofagasta Airport (23.45361° S, 70.44056° W) during the most extreme ENSO phases. These extreme ENSO phases correspond to the highest, lowest, and the most neutral SST anomalies in the ONI index observed from 1995 to 2017 during February and September.
For September, Fig. 10a shows that MBL well-mixed θ-profiles amount to 285 K during cool and neutral ENSO phases and to 288 K during warm ENSO phases. The three inversion layers (∆θ) present more differences than the mixed layer values. The cool ENSO phase presents an ∆θ of 13 K with h ~ 1000 m.
Conversely, the warm ENSO phase shows a slightly weaker ∆θ than the cold one (10 K) a hundred meters higher (h ~ 1100 m). Finally, the neutral ENSO phase presents a ∆θ similar to the warm one, yet at a lower height (h ~ 900 m). These thermal inversion layer values are consistent with the statistics reported in Fig. 9. The three troposphere vertical thermal gradients shown in the θ-profiles are similar among them, with values around 0.035 K m−1. The structure of the q-profiles shown in Fig. 10b agrees with the θ-profiles one. During warm ENSO phases, the MBL is well mixed and the moistest, with values ranging from 7 to 9 g kg−1. Likewise, during cool and warm ENSO phases, the inversion layer (Fig. 10b) can be found at 1000 and 1200 m, respectively. Finally, lapse rates of −0.001 g kg−1 m−1 in q-profiles are equally observed in the three ENSO phases.
During February, the state of the MBL is opposite to the winter one. Figure 10c shows less mixing in the MBL, reporting mixed layer values of 0.002 K m−1. Likewise, the inversion layer is remarkably weaker in the summer than winter, with values around ∆θ ~ 5 K which are challenging to identify due to the lower MBL mixing. However, the height of the inversion base observed in February is higher during the warm ENSO phase (h ~ 1250 m) than during the cool one (h ~ 750 m). Finally, during summer the free troposphere thermal gradient is low in all the three ENSO phases, with values around 0,005 K m−1. The latter is related to the summer decreasing subsidence (Fig. 9). Figure 10d shows the MBL q-profiles in each of the three ENSO phases. These profiles present a similar but moister mixing than in winter (~ 11 g kg−1). Here, the inversion layer observed in q-profiles (Fig. 10d) is less defined than in θ-profiles. However, we find the MBL at 1300 m during the warm ENSO phase and at 750 m during the cool one.
The MBL shows an opposite behavior concerning ENSO phases between September and February. This contrast is related to the FLCs frequency along the coastal Atacama Desert observed by the GOES satellite. For example, during September cool ENSO phases prompt the strongest and best-defined ∆θ. This phenomenon contributes to increasing fog frequency in two ways. First, the thermal inversion layer restricts the MBL below the top of the coastal mountains (Rutllant et al. 2003), avoiding entrainment from the free troposphere and the MBL mixing with dry inland air masses. Second, it enhances the condensation over the ocean due to low sea surface temperatures. In contrast, the warm ENSO phases in September might generate less fog frequency due to the decreased inversion layer strength, leading to more entrainment and less condensation over the ocean.
During February, warm ENSO phases have an opposite effect compared to the one observed in September, producing increased fog frequency as cool ENSO phases do during September. This change is explained by a combination of SST, subsidence and topographic factors. First, under warm ENSO phases, the MBL and the free troposphere are the moistest and the warmest. Therefore, the warmest and moistest marine air masses induce the MBL to grow and reach the coastal summits (1300 m a. s. l.). Simultaneously, the coastal summits force MBL air masses to lift, cooling down and condense. These interpretations agree with previous studies that show a higher advective fog frequency during winter and a higher orographic fog frequency during summer in the study area (Cereceda et al. 2002, 2008).
The FLCs interannual variability reveals the role played by ENSO in the cloud presence over the coastal Atacama Desert. Here, ENSO phases directly affect the interaction among the marine boundary layer, the free troposphere and the sea surface temperature (Schubert et al. 1979). Figure 11 summarizes our interpretation of this phenomenon under four scenarios: (a) cool ENSO phases in September, (b) warm ENSO phases in September, (c) cool ENSO phases in February and (d) warm ENSO phases in February. In September and under a cool ENSO phase, high FLCs arose from the strong advection of the Sc-clouds decks in the study area. These advected clouds remain steady until reaching the coastal cordillera, where cool SSTs contribute to keeping the same ocean conditions. The low surface temperatures also increase the thermal inversion layer, which results in a boundary layer height of around 1000 m. Again in September, but under a warm ENSO phase, the FLCs presence is due to the combination of higher SSTs and a weaker inversion layer, which promotes more entrainment from the dry and free troposphere. This entrainment contributes to dissipating clouds and decreasing the boundary layer height. In February and under a cool ENSO phase, the FLCs presence is related to lower temperature and subsidence than in September. Together with a diminished inversion layer, these conditions provoke higher entrainment and, consequently, a lower boundary layer height. Finally, in February and under a warm ENSO phase, we find a higher FLCs presence compared to a neutral ENSO phase, revealing the role played by the topography. In this case, the MBL is the warmest and moistest, while higher surface fluxes increase its height. This thicker boundary layer allows warm and moist oceanic air masses to cool while lifting above the coastal cordillera, enhancing their condensation.
Insights into Tillandsia Landbeckii past and present distribution
The FLCs presence is strongly related to water availability and, therefore, to fog-driven ecosystems such as the Tillandsia landbeckii lomas vegetation. Studies have shown that T. landbeckii plants derive from fog not only moisture but also key nutrients such as nitrogen and phosphorus (Gonzalez et al. 2011a; Latorre et al. 2011; Jaeschke et al. 2019). These inputs provide sustenance for an ecological community of consumers (González et al. 2011b). Hence, it stands to reason that the T. landbeckii ecosystems distribution is sensitive to FLCs spatiotemporal variations, acting as a “biosensor” of fog availability. The long-term history of the evolution of this highly specialized species and its actual distribution pattern may hint that, while being adapted to a minimum average fog occurrence during the entire growing season, these species do not tolerate long-term stochasticity and large variation of such fog events. The data reported in Table 1 support this assumption by showing that the area between 15 and 25 km inland (clusters 2 and 3), which is the Chilean coastal region harboring T. landbeckii loma vegetation (e.g., Koch et al. 2020), also presents the lower variation of FLCs occurrence.
According to the previously described results, the weak decreasing (increasing) trend of fog presence at altitudes above (below) the 1000 m a. s. l. may have already significantly affected the survival of these ecosystems over the past 2–3 decades. This repercussion may be evident especially along the margins of the Tillandsia distribution, which could be susceptible areas to such changes. For example, the Tillandsia lomas at Cerro Oyarbide recently show significant mortality along the lower margins (approximately between 900 and 1050 m a. s. l.) of its distribution (Fig. 12). Moreover, the altitude range of this phenomenon (around the 1000 m a. s. l.) appears to be relevant to the atmosphere–biosphere relationship due to the increasing LCL (Lifting Condensation Level) altitude, the downward cloud base altitude trend (Muñoz et al. 2011, 2016), and the resulting variations in exposures to wetter-coastal or drier-desertic conditions (Schween et al. 2020; García et al. 2021). This critical relationship between FLCs distribution and Tillandsia dieback is worth to be further studied. Finally, the radiocarbon analysis of the affected vegetation bands provided information about the time of their death, thus a better understanding of the temporal distribution of these past changes. Indeed, some of these preserved bands date back to several thousand years and are located at elevations where no Tillandsia ecosystems are found today (Latorre et al. 2011). Therefore, the herein presented data of increased stochasticity of the ENSO climate system may explain the large-scale dieback of Tillandsia lomas in the Atacama Desert.