**3. Results**

### *3.1. Germination and Fertilization Observations and Experiments*

*S. lepifera* exhibits a typical fern life cycle with generational alternation between the gametophyte and sporophyte stages. We observed that the sporangia of *S. lepifera* produce single-celled meiotic spores in April and October each year (Figure 1). The spores germinate in about 60–70 days via the vittaria-type germination pattern. The center of each spore (Figure 1A) contains oil droplets, and germination starts with the exertion of a rhizoid (Figure 1B–E). The apical chloroplast-bearing cells opposite the rhizoid development (Figure 1E) become enlarged, and the emerging prothallus (Figure 1F) develops several rhizoids on its ventral surface. The structure of the gametophyte is simple but with differentiation of the dorsal and ventral surfaces. In particular, when the gametophyte matures, there are many antheridia (Figure 1G) and archegonia (Figure 1H) on the ventral surface. Antheridia generate sperm that must swim in the water and sense the chemical attractants secreted by the archegonia to complete the fertilization.

Warm temperatures and high moisture appear essential for fertilization to occur in *S. lepifera*. Specifically, we found that fertilization (evidenced by production of a sporophyte) had visually-assessed higher success rate among plants grown at 25 ◦C than 18 ◦C. Moreover, we found that watering every day had an important impact on the formation of sporophytes, possibly related to maintaining high humidity (Figure 1I). In fact, we found that gametophytes that were watered only once weekly became malformed and yielded unproductive sporophytes.

**Figure 1.** Spore germination process of *Sphaeropteris lepifera*. (**A**) Spore of *S. lepifera* (center of each spore contains oil droplets). Scale bar = 50 μm. (**B**–**E**) Spore germination. Scale bar = 100 μm. (**F**) Prothallus. Scale bar = 100 μm. (**G**) Antheridium. Scale bar = 100 μm. (**H**) Archegonium. Scale bar = 100 μm. (**I**) Young sporophyte. Scale bar = 10 cm.

#### *3.2. Ecological Niche Modeling and Importance of Environmental Variables*

Our ENM result for *S. lepifera* showed that the most suitable habitat was found on northern Taiwan, along the central and western coast of Guangdong, on coastal Guangxi and Yunnan, and on near shore islands south of these provinces, such as Nan'ao. Areas of moderate suitability included coastal mountains. The AUC was 0.983 (Table 2, Figure 2A) suggesting that our model was robust. Notably, among reintroduction sites, only Shaoguan was predicted as environmentally unsuitable.

Our estimates of the relative contributions of the sampled environmental variables to the MaxEnt model showed that mean diurnal range (Bio02) contributed the most to the model (29.3%) followed by temperature seasonality (Bio04, 24.7%), precipitation of the coldest quarter (Bio19, 13.5%), and mean annual precipitation (Bio12, 11.7%) (Table 2). Precipitation of the coldest quarter was also the most important variable based on permutation (66.6%).

The jackknife test of variable importance (Figure 2B–D) showed that, overall, models showed roughly equal gains when any one variable was excluded. However, the highest training and test gain (Figure 2B,C) for a variable used in isolation were for mean diurnal range (Bio02), followed by temperature seasonality (Bio04), mean temperature of the coldest quarter (Bio11), isothermality (Bio03) and precipitation of the warmest quarter (Bio18). The first two of these are in agreemen<sup>t</sup> with the measured percentage contributions, the mean temperature of the coldest quarter is consistent with the permutation result, and the others are broadly congruen<sup>t</sup> with the inference that temperature and precipitation variability are critical for the species. The AUC calculations under the jackknife test (Figure 2D) showed a similar pattern as the training and test gain, except that the mean annual temperature (Bio01) also showed notable predictive capability in isolation.


**Table 2.** Contributions of the environmental variables to the MAXENT models using the 19 bioclimatic variables as well as slope, aspect, land cover, and population.

### *3.3. Community Composition*

We observed that the forests of Nan'ao Island mainly comprised mixed stands of *Acacia confusa* Merr. and *Pinus massoniana* D. Don with wild shrubs from the families Rosaceae, Rutaceae, and Theaceae. According to the Drude abundance analysis of the community composition of *S. lepifera* (Table 3), the tree layer is mainly composed of *Machilus chinensis*(Benth.) Hemsl., *Schefflera heptaphylla* (L.) Frodin, *Cunninghamia lanceolata* (Lamb.) Hook. and *Acacia confusa* and is about 8–13 m in height. The shrub layer is dominated by *Psychotria rubra* (Lour.) Poir. and is about 1.2–1.5 m in height. The herb layer consists of *Eleutherococcus trifoliatus* (L.) S.Y. Hu, *Blechnum orientale* L., *Boehmeria nivea* (L.) Gaudich., *Cyclosorus parasiticus*(L.) Farw., *Deparia lancea* (Thunb.) Fraser-Jenk., *Dryopteris fuscipes* C. Chr., *Pteris fauriei* Hieron., *Alocasia odora* (Roxb.) K. Koch, *Adiantum flabellulatum* L., *Lygodium japonicum* (Thunb.) Sw., *Mussaenda pubescens* W.T. Aiton, and *Liriope spicata* (Thunb.) Lour.

### *3.4. Preliminary Reintroduction Experiment*

The survey in February 2021 revealed that 14 of 40 introduced individuals of *S. lepifera* survived at the site of the original population on Nan'ao Island compared with 27 of 40 and three of 40 at the two nearby sites. In addition, Maofeng Mountain in Guangzhou, Renhua County in Shaoguan, and Bajia Town in Yangchun five, 10, and five survived, respectively, of the 20 planted. The environmental conditions of the site are shown in Table 4.

**Figure 2.** (**A**) ROC curve for MaxEnt prediction; (**B**) Result of the jackknife test of variable importance using training gain; (**C**) Result of the Jackknife test using test gain; (**D**) Result of the jackknife test using AUC on test data.

Although the sample size is small and the environmental variability limited due to geographic distance among reintroduction sites, it may be noteworthy that the site with the highest survival rate (after averaging among the three adjacent Shantao sites), Shaoguan (50%), represents the extremes of all values shown to be important based on the various modes of testing within the ENM framework (Figure 3). Moreover, in several cases, the site with the second-highest survival rate, Shantao (38% average rate), is usually closer to the same extremes as Shaoguan compared to the other key variables. This is the case with annual precipitation, isothermality, and precipitation of the warmest quarter. This may further highlight the importance of these specific four variables for the survival of reintroduced plants. However, in some cases Shaoguan and Shantao are at opposite extremes of the environmental variables surveyed. This includes for mean diurnal range, temperature seasonality, mean temperature of the coldest quarter, and precipitation of the

coldest quarter. These four variables are all related to temperature. Notably, at Shantao, meticulous managemen<sup>t</sup> practices included mitigating the effects of temperature, especially during winter, possibly suggesting the feasibility of successful reintroduction even under less than ideal temperature conditions.

**Table 3.** All species recorded in quadrats established around adult individuals of *Sphaeropteris lepifera* on Nan'ao Island.


Note: Soc, society; Cop, copiosa; Sp, sparsal; Sol, solitariae; Uni, unicum.

According the MaxEnt results, we divided predictions for *S. lepifera* into five categories: currently present, high habitat suitability, moderate suitability, low suitability, and unsuitable (Figure 4). Much of the area throughout Guangdong Province is moderately to highly suitable for large-scale reintroduction. This is in contrast to Guangxi Province, which has a very similar flora to Guangdong, sharing ca. 92% of plant genera [48], and in both provinces, the flora has strongly tropical and subtropical characteristics. However, within Guangxi, mountainous terrain may affect critical environmental factors, such as temperature seasonality and precipitation of the coldest quarter, thus limiting the moderately suitable habitat for *S. lepifera* to areas along the border with Vietnam. Additionally, the MaxEnt model predicted that there are some places in Hong Kong that are highly suitable for introduction.


**Table 4.** Comparison of environmental conditions between the reintroduced area and Nanao Island.

**Figure 3.** Survival rates of reintroduced seedlings at study sites based on the most important environmental variables according to assessments within the MaxEnt ENM framework. Survival rates were averaged across the three adjacent sites in Shantou, where the original population occurs.

**Figure 4.** Prediction of habitat suitability for *Sphaeropteris lepifera* in eastern Asia.
