The Ciénaga Las Cucharillas natural reserve (18° 26′25.27″ N, 66° 08′08.39″ W) is an urban coastal palustrine-estuarine wetland dominated by freshwater herbaceous vegetation transitioning to mangroves. It is located on the western side of the San Juan Bay in the northern metropolitan area of Puerto Rico (Lugo et al., 2011; Figure 1). The current extent of Cucharillas covers 500 hectares which is all that remains of the historic wetland area. Studies suggest that as much as 90% of the historic mangrove area may have been lost in the 20th century from filling activity for urban and industrial use of the land (Martinuzzi et al., 2009; Lugo et al., 2011). Our study area covers 2 hectares of the wetland, which have been part of restoration efforts for the last 20 years by the community-based organization Corredor del Yaguazo.

The wetland is part of the Cucharillas drainage basin, located between two large river basins that drain to the bay: the río Bayamón and the río Piedras basins. The present hydrological regime was modified from colonial times to the present, including (a) drainage channels for agricultural use from the 17th century until the mid-20th century (Kennaway and Helmer, 2007); (b) the construction of the Malaria channel in the 1940s, bringing a direct flow of fresh water to the wetland from the upper and middle parts of the basin (Pumarada-O’Neill et al., 1991); and (c) restricted seawater exchange due to the dike effect of an outflow water pump structure at the mouth of the channel (Webb and Gómez-Gómez, 1998). For the metropolitan area of San Juan, the mean annual temperature is 25.7°C, and the average annual precipitation is 1,755 mm (Walter, 1971). Rainfall in the northern Caribbean has a bimodal pattern (Taylor, 2002). Puerto Rico has two wet seasons: a peak in July-December, another in May-June, and a dry season from January to April (Torres-Valcárcel et al., 2014).

Monthly rainfall measurements were collected from daily values from the National Climate Data Center (NCDC) Toa Baja Station (RQC00669415) and a meteorological station located in the wetland. We gathered water level and conductivity measurements in ten (10) measuring wells installed in the study area in 2015, establishing a spatial gradient from the coast (800 meters from the open water coast) to inland (300 meters from the freshwater Malaria channel) (Figure 1). We recorded conductivity (mS) monthly at the water table and at 2.5 m depth with an EcoSense EC300A handheld conductivity meter. Water levels were measured monthly with a measuring tape at each well and deployed water level data loggers (Onset corp.). The results shown here are the monthly averages. We performed ANOVA and Tukey test if significant differences were found between conductivity, water levels, precipitation data, and monitoring wells. Microtopography analysis was done in ArcGIS Pro 2.8 using DEM data sources of elevation surfaces at a 10 m resolution from the USGS National Geospatial Program.

To capture wet and dry season variability effects on the wetland changes in plant functional forms distribution throughout the 2,000 m² study area, we used a Phantom 3 (UAV) electrically powered quadcopter (DJI Company) with a Red-Green-Blue (RGB) camera in April 2018, October 2018, and January 2019. Missions were planned around optimal weather conditions, including no rain and low wind speed (less than 5 m/s). Each mission was flown at an altitude of 50 meters (2.2 cm/pixel resolution) in compliance with FAA regulations, and each flight lasted 18–20 min.

Orthomosaic images were analyzed using the program ArcGIS Pro. The first analysis was to create a vegetation index using only the visible RGB bands. Spectral vegetation indices were used to monitor and analyze spatio-temporal variations in vegetation structure, commonly based on infrared bands. The VARI (Visible Atmospherically Resistant Index) allows the calculation of vegetation indices using the visible bands (Gitelson et al., 2002; Raoufat et al., 2020):

Vegetation cover types were identified manually, creating feature classes, rasterized (converted from vector to a raster image), and adapted to the exact spatial resolution for pixel-by-pixel comparison.

Drone flights in April 2018 produced less usable imagery than the other flight dates because low light and high wind conditions were causing blurry imagery. When these images were stitched together, the output orthomosaic image had empty pixels where the original images were too blurry to analyze. Although this resulted in a loss of 859 m² of aerial imagery, this accounts for less than 4 % of the entire study area and results are still significant.

Land cover categories were trees, herbs, shrubs, grasses, water, managed areas, and bare soil. The four vegetation-based categories represented plant growth forms. Using these land cover categories, the description of vegetation types focuses on physiological characteristics in response to environmental factors and ecosystem processes (Chapin et al., 1996). We sampled representative species for three plant types for ecophysiological analyses: (a) trees (Laguncularia racemosa and Avicennia germinans), (b) shrubs (Dalbergia ecastaphyllum), and (c) herbs (Acrostichum danaeifolium) (Table 1).

From 2018 to 2019, we conducted a periodic collection of vegetation specimens for classification. Individuals were pressed, dried, identified, and grouped into functional groups based on life forms. We used literature and herbarium specimens from the New York Botanical Garden,¹ University of Puerto Rico, Río Piedras² and Mayaguez campus³ to compare post-hurricane species to the pre-hurricane wetland composition.

Fully expanded and exposed adult leaves (at a 2-meter height in trees) of the four plant species representative of plant growth forms present in the wetland (trees, shrubs, and herbs) were sampled and measured throughout the seasons. L. racemosa (L.) C.F. Gaertn. and A. germinans (L.) L. are both considered salt-tolerant trees (Parida and Jha, 2010). D. ecastaphyllum (L.) Taub. is a climbing decumbent shrub, reaching 1–5 m in length and grows in non-forested areas forming monospecific stands (Acevedo-Rodríguez, 2005). A. danaeifolium Langsd. & Fisch. Is an herbaceous, rhizomatous fern common in brackish swamps, tolerant to soil saline conditions up to 45–50 mS/cm but requires freshwater for establishment (Medina et al., 1990a). We sampled during 2018-2019 in the wet (May to October) and dry (November – April) periods. We assessed leaf area and dry mass at least once each dry and wet period for the three plant types representative of the plant growth form (Table 1). At the end of the study, we measured δ¹³C and δ¹⁵N isotopes in the leaf tissue for assessing long-term water use efficiency (WUE).

The leaves were stored in a cooler to prevent water loss until analysis. Leaf area was measured with an LI-3100C Area Meter and later dried in a forced-air circulation oven at 60°C for dry mass determination. Specific leaf area (SLA; the area of a fresh leaf divided by its oven-dry mass) was calculated as an index of the construction cost of leaves. SLA tends to relate positively to relative growth among species (Pérez-Harguindeguy et al., 2016). In the leaf economics spectrum, SLA indicates habitat preferences and plant productivity in environments under stress (Medina et al., 1990b). The SLA is also related to light conditions, where high values indicate shaded leaves and low values suggest high light or open canopies.

Carbon isotope analyses (δ¹³C) were performed in leaf samples to reveal plant long-term WUE which has been associated with salt tolerance (Ball and Passioura, 1995). Intracellular CO₂ and plant WUE was calculated based on the Farquhar et al. (1989) equation: δ¹³C_leaf = δ¹³C_air – a – (b-a) C_i / C_a, where δ¹³C_air is the carbon isotope ratio of the CO₂ in the air (around 8.2 ‰); a is the fractionation by slower diffusion of δ ¹³C to δ ¹²C (4.4 ‰); b is the fractionation by ribulose biphosphate carboxylase against ¹³C (27 ‰); and C_a is the atmospheric CO₂ concentration which averaged 413 μmol/mol at the time of the study. Intrinsic WUE was derived from the Lambers et al. (2008) equation, where intrinsic WUE = (A_n/g_s).

Shapiro Wilk test was used to test normality in distribution. Analysis of variance (ANOVA) was used when data fits normality; if not, a nonparametric Wilcoxon/Kruskal Wallis test was used with the statistical program JMP Pro version 13.

Around 33% of the study area was devoid of standing vegetation (Figure 2), which decreased considerably in July 2018 and January 2019. In April, trees were the predominant cover with 8,221 m², followed by grasses, 2,532 m², shrubs, 1,545 m², herbaceous vegetation, 1,368 m², and managed areas with 778 m² of the study area (Figure 3; Table 2).

In July 2018, trees continued as the dominant cover with 10,929 m², followed by grasses, 2,955 m², herbaceous cover 2,763 m², shrubs 2,077, and managed spaces with 312 m². In January 2019, trees occupied 10,345 m², a decrease related to the removal of trees due to rehabilitation efforts, herbaceous cover 5,363 m², shrubs 2,084 m², grasses 1,767 m², and managed areas 1,677 m² within the study area.

Precipitation data exhibited the bimodal pattern for the island, June (316 mm) and August (281 mm) being the wettest months. The driest month was March (60.9 mm; Figure 4). At the phreatic level, we found higher conductivity values closer to the coastline with a sharp drop around 1 km inland, with conductivity values ranging spatially from 10 to 35 mS. We found marine intrusion at a depth of 2.5 meters in most of the wells (Figure 5). This trend prevailed through the wet period (September to November) and to the beginning of the dry period (January 2019; Figure 4). An elevation profile of the microtopography shows that most of the wetland is below sea level – 0.4 meters below sea level at its lowest point and only 0.14 meters above sea level at its highest elevation (Figure 6).

From May 2018 to March 2019, we collected a total of 58 species which we categorized into 32 families. The predominant plant growth form was herbs with 34 species, followed by 10 species of trees, 10 vines, and 2 species of shrubs and 3 grasses. We found a proportion of 0.17 of woody species to non-woody (10 woody vs. 48 non-woody species), and 47 native species and 11 nonnatives (4.3 ratio). There were 42 dicotyledons and 12 monocotyledons. Of the collected species, 46 were perennial species, and 12 were annual (Table 3).

Trees had a lower leaf area than herbaceous and shrub species, where Avicennia and Laguncularia leaves were smaller than Acrostichum and Dalbergia (Table 4). The heaviest leaves were from Dalbergia; Laguncularia and Avicennia trees had similar foliar weights to each other. Leaves for all species followed the same weight area relationship, indicating a uniform sampling of adult leaves, allowing for comparison within dates (Figure 7). SLA for species varied significantly between dates, showing decreasing patterns post-hurricane in Avicennia, and increasing SLA values in Dalbergia. Laguncularia remained constant, except for June 2019 which SLA values decreased significantly from the rest of the dates (average 47.2 g/cm²). The pattern of SLA was Acrostichum > Dalbergia ≥ Avicennia > Laguncularia.

Leaf carbon (C) concentration varied from 36 to 58%, with a mean of 45 ± 7 % in all species. Dalbergia had the highest C concentration with 53 ± 3 %, and Laguncularia had the least with 37 ± 0.8 %. Tukey-Kramer comparison showed significant differences between species, except between Acrostichum and Laguncularia. Leaf nitrogen (N) concentration ranged from 0.91 to 3.95 %, with a mean of 2.4 ± 0.8 %. Dalbergia had the highest N concentration with 3 ± 3.1%, and Laguncularia had the least with 1.1 ± 0.1 %.

Leaf C isotope ratios spanned a range of over 6 ‰ with a mean of −28.7 ± 2.1 ‰. Avicennia showed a large range of values from −32.2 to −28.2 ‰, (average −29.8 ± 1.9 ‰) while the rest of the species had smaller ranges (standard deviations <0.8 ‰). Acrostichum and Dalbergia C stable isotopes were similar and differed significantly from Laguncularia and Avicennia. N isotope ratios spanned a range of over 8 ‰ with a mean of 3.79 ± 2.7 ‰. Avicennia had the highest values with 6.6 ± 1.2 ‰ and Dalbergia the lowest with 0.3 ± 0.7 ‰. Plant WUE varied significantly among plant life-forms where trees had lower values than herbaceous species.