Shining a light on duckweed: exploring the effects of artificial light at night (ALAN) on growth and pigmentation

Plant collection and species description

We collected a large sample of floating duckweeds on 25th February 2022 from Busby’s Pond located in Centennial Park, Sydney, NSW, Australia (coordinates: −33.897873 151.228606). We placed the plants into plastic buckets with the pond water (pH = 7.4, Ammonia = 1.0 ppm, Nitrate = less than 0.25 ppm, Nitrite = less than 1.0 ppm; API Freshwater Master Test Kit) and allowed collected plants to acclimate to indoor conditions for 4 days.

After examination under a bifocal microscope, we determined the duckweed sample species to be Landoltia punctata (Les & Crawford, 1999). This species is native and common in New South Wales (PlantNET, 2023; NSW FloraOnline; https://plantnet.rbgsyd.nsw.gov.au/; accessed March 2022). The typical frond (a clonal plant) is made from one to three leaves: a mother leaf, a budding out new leaf, and sometimes also another new buddying leaf. The leaves are very variable in terms of their size, shape, and pigmentation (Les & Crawford, 1999; Baek, Saeed & Choi, 2021). The upper side of leaves is typically dark green, but the lower side can accumulate red pigment and appear dark brown or purplish (Kittiwongwattana, 2019).

Experimental setup

Figure 1A summarizes the experimental design. We conducted the experiment using 161 30 mL clear plastic cups. we filled all cups with 20 mL of pond water each. We labelled these cups with a number 1–161 and drew a water level line at 20 mL for future refills. We placed one randomly selected duckweed frond (2–3 connected leaves) into each cup. Then, we used an R script to randomly assign a cup into either the Control (N = 81) or ALAN (N = 80) group. We placed the cups into one of the two large holding boxes (Control and ALAN) described below.

The two holding boxes were 50 L in volume, made of black opaque plastic and had cardboard lids fitted on top to block any outside light at night, but allowing ventilation via a small gap under the overhanging edges of the lids (Fig. 1B). One box was used as a Control condition, and the other as an ALAN condition. The lid that covered the ALAN box had a small hole in it which was mounted with a lamp (LED lamp, Correlated Color Temperature (CCT) 3000K) for shining artificial light at night on ALAN-group plants. We placed the two boxes next to each over near a west-facing window (Fig. 1B), approx. 10 cm from the fully transparent large window glass panel and 6 m from an air conditioner located at the top of an internal wall. We swapped positions of the two boxes every few days.

At the beginning of each day, we took the lids off the boxes, and the lamp was switched off. At dusk, we put the lids back on, and turned the lamp on. Every few days, we refilled water lost to evaporation in the cups, as needed, and randomly rearranged the cups within each box to negate potential small differences in the light exposure or temperature within each box. Every few weeks, we measured the temperature, humidity, and light intensity with a digital thermometer, humidity meter, and a Photon app (https://growlightmeter.com/; v.0.4.2 for iPhone SE; free version downloaded March 2022), within each box at a few different locations.

The experiment was conducted between 27 February and 17 April 2022, which corresponds to the early autumn season at the experimental location. The exact geographic location of the experiment was: latitude of −33.92296868040449 and longitude of 151.2373484125205 (coordinate values derived from Google Maps).

During the day, the amount of light illumination in both boxes was, on average, 493 lux (Standard Deviation, SD = 144 lux). During the night, the illumination in the ALAN box was, on average, 40.5 lux (SD = 8 lux). During the night, the illumination in the Control box was, on average, 0.2 lux (SD = 0 lux). The difference in illumination between the Control and ALAN boxes at night was statistically significant (Wilcoxon test: W = 16, p = 0.021).

To answer our research questions, we estimated duckweed growth as weekly leaf counts per sample and as total leaf surface area per cup on Day 28. On Day 49, we measured leaf pigmentation levels as a percentage of dark-coloured leaf area on the underside of the leaves, where most red pigment accumulates in this duckweed species (Fig. 2). We provide detailed descriptions of these measurements below.

Leaf counts

During the experiment, we weekly counted the number of live leaves per cup in all the cups to collect data on duckweed growth and reproduction (via leaf buddying). Before measurements, one researcher removed the labels from the boxes so that the other researcher counting leaves would not know about the treatment group identity. Such “blinded” data collection aims to reduce unconscious bias during measurements.

One researcher wrote down the number of leaves and the corresponding cup number, as well as refilled the pond water in the cup as needed. Once we completed processing one box, the other researcher shuffled the cups around and placed them back in their original box. Then, we repeated the process with the other box. When the counting was finished, we placed the labels of each box back on, and entered the count data into a spreadsheet. We then cross-checked the entered data for errors and corrected any mismatches by re-checking cups as needed.

On experiment Day 42 and Day 49, along with the number of live leaves, we also recorded the number of dead leaves. One (blinded to the treatment group) researcher counted the number of discoloured (i.e., turned white / yellow, or completely falling apart) leaves that were floating on the water’s surface. Such dead leaves were rare before Day 42, and because of this, they were initially not counted. We then calculated total leaf counts for Day 42 and Day 49, by adding counts of live and dead leaves together, to represent the rates of growth accounting for duckweed aging and mortality.

Leaf surface area

On Day 28 of the experiment, we measured the total surface area of leaves within each cup. To do so, we built a cardboard stand that was set up to hold a phone firmly so that it would not shake when taking photos. We used a small ceramic base plate as a white background for taking plant images. The plate had four black dots in a square shape, each three cm apart. This was to allow the iPhone app LeafByte (Getman-Pickering et al., 2020) to accurately calculate the total leaf surface area by scaling it to the dot locations. To correctly identify each plant sample in the photo taken, we wrote the ID number of the sample with an erasable pen just outside the square formed by the four dots, within the camera vision range. We then gently removed all the plants, one by one, out of their cups (with the use of a thin brush), with a small amount of water to prevent desiccation-related damage(which occurs rapidly in small aquatic plants). We placed all leaves within the four dots area to take a photo, with the upper side of the leaves facing up. We took a photo of each sample. After this, we quickly returned the plants to their corresponding cup using a brush. We repeated this process for every experimental cup until we had a photo for every sample. We then used the LeafByte app to calculate the total leaf area of each sample and exported all the data from the app to an Excel spreadsheet.

Leaf pigmentation

We estimated pigmentation levels of the undersides of the leaves on Day 49 of the experiment. We had the same setup as for the measurements of the total area of the leaves on Day 28 but used a flat background plate, avoided transferring excess water for better image quality, and flipped the leaves upside down so that the bottom side of the leaves (which accumulates most anthocyanins (Kittiwongwattana, 2019)) was facing up. We analysed the images using an R package pliman v. 1.1.0 (Olivoto, 2022). This package can extract the colouration of the leaves by removing the background and classifying areas with different pigmentation levels into two classes “dark” (more pigments) and “light” (less pigmented). The algorithm is based on a general linear model (binomial family) fitted to the RGB values of each image, calibrated on a set of predefined color palette samples. This approach took into account that, within each sample, some leaves (or leaves parts) could be less pigmented, and others could be more pigmented. In our workflow, for each duckweed sample, we estimated the proportion of total leaf area classified as “dark”.

Statistical analyses

We conducted all statistical analyses in the R v.4.2.3 statistical environment (R Core Team, 2021). Analysis code and processed data are available on GitHub (https://github.com/lordofthepigs123/ALAN_duckweed and archived on Zenodo with doi:10.5281/zenodo.8017670). We compared average light exposure for each group of samples, time trends and differences in weekly leaf counts during the experiment, total leaf area on Day 28 and pigmentation on the underside of leaves on Day 49. Sample sizes were initially 80 for ALAN and 81 for Control group, with the first 80 from each group used for weekly counts. Two samples of plants were lost during taking photo measurements due to spilling them on the floor. Missing values were omitted from statistical analyses; thus, sample sizes are 80 and 80, per replicate, for most statistical tests.

Leaf counts across 7 weeks

Using a Poisson mixed model predicting leaf counts within experimental cups across 7 weeks, we found a tendency for a positive effect of ALAN treatment on leaf counts (estimated difference between ALAN and Control = 0.079, 95% CI [−0.01, 0.18], p = 0.079; Standard Error, SE = 0.04, 95% CI [−0.005, 0.09]), as shown in Fig. 3. The number of leaves increased over time (slope = 0.38, 95% CI [0.36, 0.40], p < .001; SE = 0.38, 95% CI [0.36, 0.40]). The model’s total explanatory power was substantial (conditional R ² = 0.62, marginal R ² = 0.41; sensu Nakagawa & Schielzeth, 2013).

Duckweed plants had the same number of live leaves on Day 0 (start of the experiment) in ALAN and Control groups (Wilcoxon rank sum test: W = 3170, p = 0.906). From Day 7, there was a tendency for more live leaves in the ALAN group than in the Control group, but it was not consistently statistically significant across all weeks (Table 1).

The mortality, estimated as the number of dead leaves, was greater in the ALAN group on Day 42 (W = 4920. p = 0.000) and Day 49 (W = 4830, p = 0.000) (Figs. 4A and 4B). Total counts of leaves (live and dead) revealed higher overall growth and reproduction in the ALAN than in the Control group on Day 42 (W = 3808, p = 0.037; Fig. 4C), and a tendency in the same direction on Day 49 (W = 3745, p = 0.061; Fig. 4D).

The leaf numbers in the ALAN group were increasingly more variable than in the Control group as the experiment progressed (Table 1), especially on Day 49 (variance test: F = 2.007, p = 0.002, df = 79, 79). This pattern of greater variability in the ALAN group was also clear for the total counts of live and dead leaves on Day 42 (F = 1.886, p = 0.005, df = 79, 79) and Day 49 (F = 2.193, p = 0.001, df = 79, 79).

Total leaf surface area and pigmentation

Total leaf area for each sample measured on Day 28 showed a clear difference (W = 4020, p = 0.005; Table 1), indicating that plants in the ALAN group grew and produced new fronds faster than the Control group (Fig. 4E). As with the leaf counts, the leaf area of plants in the ALAN group was more variable than in the Control plants on Day 28 (F = 1.814, p = 0.009, df = 79, 79). Proportion of the underside leaf area with dark pigment content on Day 49 was higher in the ALAN group plants than in the Control plants (Fig. 4F; W = 4738, p = 0.000; Table 1), with higher variance observed in the ALAN group (F = 0.628, p = 0.041, df = 78, 79).

Finally, exploratory analyses showed a positive relationship between leaf count and total leaf area on Day 28 (Fig. 5A), with a strong correlation between these variables both for the ALAN group (r = 0.90, CI = [0.85, 0.94], p = 0.000) and for the Control group (r = 0.85, CI = [0.78, 0.91], p = 0.000). There was also a negative relationship between leaf count and the measure of dark pigmentation on the undersides of the leaves on Day 49 (Fig. 5B), with moderate correlation values for both the ALAN group (r = −0.49, CI = [−0.64, −0.30], p = 0.000) and for the Control group (r = −0.41, CI = [−0.57, −0.20], p = 0.0002).

Summary of main findings

Overall, our results indicate that duckweed plants exposed to ALAN tend to respond with enhanced growth rate, but potentially also some physiological stress. We also observed greater variability of growth in the ALAN group in terms of total leaf surface area and leaf counts. Finally, we found that the lower sides of duckweed leaves exposed to ALAN were often darker than in the control group, likely due to the accumulation of protective dark pigments, such as anthocyanins. The increase in dark pigmentation suggests that ALAN induces physiological stress-response mechanisms involved in protecting tissues from oxidative damage. Below, we discuss our findings in detail.

Increased duckweed growth under ALAN

Total leaf area per sample on Day 49 was larger in the ALAN group than in the Control group. This difference was also clear from day 28 for the duckweed growth and reproduction measured as the count of leaves per sample (Table 1; Fig. 3). This finding is in line with the work of Liu et al. (2018), who found higher dry mass production (an alternative measure of growth rate) in the same species (Landoltia punctata) under constant light. Increased dry mass was also found in Yin et al. (2015) using a different duckweed species, Lemna aequinoctialis, under continuous light exposure. However, a study by Gallego, Chien & Angeles Jr (2022) concluded that photoperiod had no effect on the relative growth rate of Landoltia punctata exposed to three different light sources. Yet, a possible factor explaining this result is that their experiment was only run over 16 days, with only three replicates per treatment combination, and the effects could be too small to be reliably detected.

Increased variability in duckweed growth under ALAN

The plant samples in the ALAN group were more variable for all growth measures (Figs. 3 and 4). For example, for total leaf counts on Day 49, there are four outliers in the ALAN group, the largest of which had 34 leaves, while the Control group’s largest value was 21 leaves. This indicates that ALAN treatment caused some of the plants to respond more strongly than other plants within the same treatment group, which could be due to genetic differences and/or phenotypic plasticity. It is known that many organisms become more variable under stressful conditions (Ghalambor et al., 2007). Further, we observed higher and more variable leaf death rates in the ALAN group compared to the control group (Table 1; Figs. 4A and 4B), which could also be attributed to increased stress. Stress-induced expression of variability in phenotypic traits can potentially influence adaptive changes leading to population-level adaptation to stressful conditions (Hoffmann & Hercus, 2000).

Increased pigmentation under ALAN

Plants in the ALAN group had more pigmentation on the underside of the leaves (Fig. 4F; Table 1). The dark pigment is most likely anthocyanin, which is known to protect tissues from oxidative damage (Agati et al., 2021). Anthocyanin has been found in duckweed fronds exposed to heavy metal, and organic pollutants (Jayasri & Suthindhiran, 2017; Sharma & Kaur, 2020). High light intensity can also trigger anthocyanin biosynthesis, helping plants cope with oxidative stress (Tattini et al., 2005). Our results indicate that artificial light exposure at night can also induce this type of stress response, despite the relatively low light intensity used. At the same time, we observed that samples with the highest proportion of dark-pigmented areas were growing slower than samples with less pigmented leaves in both ALAN and Control treatments (Fig. 5B). Thus, it is possible that excessive production of dark pigment may slow down the growth of duckweed plants.

Limitations

Our experiment had five limitations. First, the light intensity of our ALAN treatment could be considered relatively weak when compared to lights used in cities. For example, streetlights have a maximum of 4800 lux (Bennie et al., 2016), and the light we used had a maximum of 200 lux. However, the illumination quickly gets weaker further from the light source, so our experimental design could have been realistic for locations further away from strong light sources or plants growing in dimly lit areas at night. Second, we also did not have an automated system to turn on and off the lights at exactly the same time every day, which might have added some daily variation in exposures. Third, daily changes in weather made the system more vulnerable to uncontrolled events (variations in daylight intensity and temperature) that could have affected the results. However, these limitations also make our experiment closer to conditions in real urban environments. Fourth, we had only two holding boxes and a single ALAN light source in our experiment, which ideally should be replicated using multiple boxes and light sources. Finally, the software we used for quantifying the pigmentation levels of the underside of the leaf only measured the proportion of darker areas rather than the full-colour spectrum. However, our approach to measuring relative levels of dark-pigmented leaf areas was robust to differences in light conditions given the light exposure when taking the images of plants was not even, which could have influenced the results of colour spectrum analyses.

Future research

Future work can be done in at least three areas, building on our results. First, investigating responses of the spotted duckweed Landolita punctata to different light wavelengths, illumination levels, and exposure durations of ALAN. Second, by carrying out equivalent experiments across different strains and duckweed species. The dark pigmentation in the leaves of duckweed is a species-specific trait, with species differing in their ability to produce dark pigments (Tippery & Les, 2020). Thus, it would be important to test the stress responses of other species of duckweed exposed to ALAN in terms of the production of pigments and other mechanisms related to antioxidant defences. future research could also quantify responses at genomic and transcriptomic levels and look at the potential carry-over effects in progeny fronds. Finally, results of our study could be used in a meta-analysis alongside other works on this topic to create a bigger picture of the effects of ALAN on plant growth and physiology.