Fog makes a significant contribution to the hydrology of a wide range of important terrestrial ecosystems. The amount and frequency of fog immersion are likely to be affected by rapid ongoing anthropogenic changes but the effects of these changes remain relatively poorly understood compared to changes in rainfall.

Here, we present the design and performance of a novel experiment to actively manipulate low lying fog abundance in an old-growth tropical montane cloud forest (TMCF) in Peru - the Wayqecha Amazon Cloud Curtain Ecosystem Experiment (WACCEE). The treatment consists of a 30m high, 40m wide mesh curtain suspended between two towers and extending down to the ground, and two supplementary curtains orientated diagonally inwards from the top of each tower and secured to the ground upslope. The curtains divert and intercept airborne water droplets in fog moving upslope, thereby depriving a ~ 420 m² patch of forest immediately behind the curtains of this water source.

The treatment caused a strong reduction in both air humidity and leaf wetness, and an increase in vapour pressure deficit, above the canopy compared to a nearby unmodified control plot. This effect was most pronounced during the nighttime (20:00 - 05:00). Below-canopy shifts within the treatment were more subtle: relative humidity at 2 m height above the ground was significantly suppressed during the daytime, while soil moisture was apparently elevated. The treatment caused a small but significant increase in air temperature above the canopy but a decrease in temperature in and near the soil, while mixed effects were observed at 2 m height above the ground. Above-canopy radiation was slightly elevated due to the treatment, which was mainly caused by a notable increase relative to the control during the dry season.

We employ the WACCEE infrastructure to understand in situ impacts of fog reduction within a pristine TCMF, but the basic principle of the method is extremely versatile. Further application of the method in other systems where fog plays a major role in ecosystem processes could improve our understanding of the ecological impacts of this important but understudied climate driver.

Above-canopy meteorology

Three weather stations were installed at the WACCEE experimental site in May 2022, two stations were installed inside the FE plot at approximately 20 m height above the canopy and one station was installed at approximately 20 m height adjacent to the experiment on a third aluminum tower 40 m from the FE plot. The weather stations included a CS320 digital thermopile pyranometer, a HygroVUE5 temperature and relative humidity (RH) sensor enclosed in a RAD06 radiation shield (Campbell Scientific Ltd, Loughborough, UK). Data was recorded on a CR1000X measurement and control datalogger and the stations were powered by a SP30 30W solar panel (Campbell Scientific Ltd, Loughborough, UK). Temperature, RH and radiation sensors were calibrated with each other prior to installation. On one station installed inside the WACCEE plot and the station installed adjacent to the experiment, two PHYTOS 31 leaf wetness sensors (METER Group, Munich, Germany) were installed perpendicular to the prevailing wind at a 45° angle. One leaf wetness sensor was installed facing upwards to estimate real leaf wetness conditions capturing both fog and precipitation water. The other leaf wetness sensor was installed facing downwards to mostly capture fog water, although there may be indirect inputs from rainfall via droplet drip and splash.

Within-canopy meteorology

Hourly air temperature and humidity were sampled with 10 sensors per plot (Hygrochron, Maxim Integrated, San Jose, CA, USA), suspended at 2 m height above the ground. Loggers were shielded to remove any confounding influences from direct radiation and precipitation.

Soil-level meteorology

Hourly temperature and soil moisture were sampled with 5 sensors per plot (TMS-4 datalogger, Tomst s.r.o., Michelská, Czech Republic). Temperature was recorded at 3 levels: 15 and 2 cm above the soil surface and 6 cm beneath the soil surface. Moisture was recorded with the time domain transmission method and recorded in raw form as electromagnetic pulses, then converted to volumetric soil moisture with a calibration curve derived for “Loamy sand A” (Wild et al. 2019) in line with previous work from the study site (Halbritter et al. 2024).

Vapour pressure deficit was calculated from temperature and relative humidity by subtracting actual vapour pressure from saturation vapour pressure using Teten’s equation (Teten 1930). Since fog only forms once relative air humidity reaches saturation point (100%), we estimated fog formation from relative humidity data by categorizing relative humidity as either equal to 100% or <100%.

To understand differences in climatic conditions between the CON and FE plots, we fitted linear mixed effect models using the lme4 package in R statistical software (Bates et al. 2015). Each climatic variable was considered as the response variable, with the treatment effect considered as a fixed effect. Date and hour were included as random intercept variables to control for variation in climatic conditions across time, whilst the treatment effect was also included as a random slope effect with hour to test for differences in response with time of day. The maximal model was compared with both a null model and a model without the treatment random slope effect based on the Akaike Information Criterion scores, with the model with the lowest AIC selected following Zuur et al. (2007). To compare between treatments, we fitted equivalent models to all climatic variables with the exception of the categorized relative humidity data where we used a generalised linear mixed model with a binomial error distribution with logit link function to account for the binary nature of the data. Model selection was undertaken in the same manner as the linear mixed effect models described above.

To understand variation in diurnal variation in climatic variables, a generalised additive model was fitted using the mgcv R package (Wood 2017), with treatment and season included as fixed effects.

To understand how leaf wetness varies with relative humidity, temperature and VPD, we fitted a general linear model with binomial family distribution with treatment and season included as fixed effect variables. Models were fitted using the stats R package (R Core Team 2023). All analyses were done in R version 4.2.3 (R Core Team 2023).