Next Article in Journal
Cross-Boundary Sustainability: Assessment across Forest Ownership Categories in the Conterminous USA Using the Montréal Process Criteria and Indicators Framework
Next Article in Special Issue
Plastic Responses in Growth, Morphology, and Biomass Allocation of Five Subtropical Tree Species to Different Degrees of Shading
Previous Article in Journal
Aesthetic Quality Assessment of Landscapes as a Model for Urban Forest Areas: A Systematic Literature Review
Previous Article in Special Issue
Nondormant Acorns Show Higher Seed Dispersal Effectiveness Than Dormant Ones
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Forests
Volume 13
Issue 7
10.3390/f13070993
forests-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Climate Niche Modelling for Mapping Potential Distributions of Four Framework Tree Species: Implications for Planning Forest Restoration in Tropical and Subtropical Asia

by
Pimonrat Tiansawat
,
Stephen D. Elliott
* and
Prasit Wangpakapattanawong
Forest Restoration Research Unit (FORRU) and Environmental Science Research Center (ESRC), Faculty of Science, Chiang Mai University, 239 Huay Kaew Road, Muang District, Chiang Mai 50200, Thailand
*
Author to whom correspondence should be addressed.
Forests 2022, 13(7), 993; https://doi.org/10.3390/f13070993
Submission received: 3 May 2022 / Revised: 17 June 2022 / Accepted: 21 June 2022 / Published: 24 June 2022
(This article belongs to the Special Issue Forest Restoration and Secondary Succession)
Browse Figures
Versions Notes

Abstract

:
Selecting tree species to plant for forest ecosystem restoration is critical but problematic. Knowing tree species’ climatic niches can help, but such information is limited for most tropical tree species. Consequently, the research presented here explored the use of climate and species record location data, to map the potential distribution of four tree species across tropical and subtropical Asia based on their modelled climatic niches. All were framework species, of proven effectiveness for forest restoration: Choerospondias axillaris (Roxb.) B.L. Burtt and A.W. Hill, Ficus hispida L.f., Hovenia dulcis Thunb., and Prunus cerasoides Buch.-Ham. ex D. Don. Potential species distributions were projected onto maps of known climatic conditions using the maximum entropy algorithm in Maxent software to predict where the climate is conducive for including each species in forest restoration trials. The models predicted species location very well for F. hispida and H. dulcis and fairly well for P. cerasoides and C. axillaris. Climatically suitable areas for C. axillaris were located mostly north of the equator, whilst those of F. hispida, H. dulcis and P. cerasoides extended south of the equator. The importance of each climatic niche variable differed among species. Driest-month precipitation was critical for F. hispida, as well as mean dry-season temperature for C. axillaris and P. cerasoides and cold-season precipitation for H. dulcis. In addition to aiding species selection for forest restoration, potential distribution maps based on climatic niche models can indicate where novel species for tree plantations might be successful and where species might become invasive. Applying such techniques to a large number of species will be needed to significantly improve species selection for forest ecosystem restoration.

1. Introduction

Tropical forest-ecosystem restoration can be defined as management interventions on degraded forestlands to maximize biomass accumulation and the recovery of forest structural complexity, biodiversity, and ecosystem functioning within prevailing climatic and soil limitations [1]. Its significance for mitigating both biodiversity loss (e.g., [2,3]) and global climate change (by carbon sequestration [4]) is now well-recognized, under such global schemes as the Bonn Challenge (targeting restoration of 350 million hectares by 2030 [5]) and the UN’s “Decade on Ecosystem Restoration” (2021–2030) [6]. Its role in climate change mitigation first rose to prominence with the inclusion of “enhancement of carbon stocks” into the UN’s REDD+ initiative [7], with safeguards to ensure biodiversity conservation and community participation (safeguards [d] and [e] in UNFCC COP16 [8]). Conventional monoculture plantations of fast-growing, often exotic, tree species satisfy neither of these safeguards [9,10]. Forest ecosystem restoration sequesters carbon 40 times more efficiently than plantations and six times more efficiently than agroforestry systems [4]. Despite this, about two thirds of the area pledged under the Bonn Challenge and other planned national initiatives are (or will be) plantations and agroforests, with only one third undergoing more natural ecological restoration [4]. Consequently, the need to develop effective techniques to restore tropical forest ecosystems is urgent, including protocols for the selection of indigenous forest tree species that will thrive in the prevailing climatic conditions of deforested sites.
In northern Thailand, the framework species method (FSM) of forest restoration (Figure 1) has been adapted, tested, and successfully used to restore various forest types on “moderately” degraded sites, i.e., where natural regeneration is too sparse to achieve canopy closure within 3–5 years (i.e., Stage 3 degradation, sensu Elliott et al. [1]). Framework species are selected from among those that are characteristic of the target ecosystem for high rates of survival and growth in the exposed conditions of deforested sites. They should also have dense spreading crowns, capable of rapidly shading out competitive herbaceous weeds, and they should attract seed-dispersing wildlife to foster biodiversity recovery (by producing fleshy fruits or nectar-rich flowers at a young age). An additional advantageous characteristic in the fire-prone seasonal tropics is resilience after fire. Selected species are propagated in nurseries from locally collected seeds for 6–18 months. Then, they are planted in restoration sites and maintained for two to three years (weeding and fertilizer application three times each rainy season) until canopy closure. The technique was first developed in Queensland, Australia [11] and it has been trialed to restore various forest ecosystem types in Thailand [1], Cambodia [12], Indonesia [13] China [14], Tanzania [15], and the Philippines [16].
It has proved to be highly successful in achieving ecological objectives. With careful species selection, weeds are shaded out and canopy closure can be achieved within 2–3 years and biodiversity recovery is rapid. For example, the species richness of the bird community during restoration of evergreen forest in northern Thailand increased from about 30 before planting to 88 after 6 years, representing about 54% of bird species recorded in nearby mature forest [17], and the birds brought in tree seeds. Sinhaseni reported that 73 species of nonplanted trees recolonized the plot system within 8–9 years, most having germinated from seeds dispersed from nearby forest by birds (particularly bulbuls), fruit bats, and civets [18]. Furthermore, Kavinchan et al. and Jantawong et al. recently demonstrated remarkably rapid recovery of ecosystem carbon dynamics. Net inputs of carbon into the soil from litterfall, the overall accumulation of soil organic carbon, and the accumulation of above-ground carbon in the trees returned to levels that are typical of old-growth natural forest within 14–16, 21.5, and 16 years, respectively [19,20]. However, since the technique relies on attracting seed-dispersing animals to foster biodiversity recovery, its use is limited where seed sources (remnant forest) are too distant and/or where seed-dispersing animals have been extirpated [21]. Its adoption is also limited by the lengthy and in-depth research needed to identify framework tree species.
A critical step in adapting the framework species method to restore a greater diversity of forest types over a wider range of climatic zones is the selection of “candidate” framework species for trialing. This is particularly difficult where target (or “reference”) forest ecosystems have been extirpated from landscapes and where local records of indigenous species are sparse. Under such circumstances, matching the climatic niches of indigenous tree species with climate maps provides an alternative way to determine where trials of candidate frameworks the tree species are likely to be successful and can greatly facilitate species–site matching to increase the likelihood of success of forest restoration projects [23,24,25].
Herbarium labels are an underutilized source of information that could greatly facilitate and improve forest restoration practices, particularly species–site matching [26]. Modern labels usually have the GPS coordinates of the collection locality, whereas old labels usually have written descriptions. Satellite imagery and/or field checks can often be used to convert such descriptive notes into GPS coordinates, which can then be used for mapping and modelling species distributions. Old, large herbaria (mostly in Europe) usually store multiple specimens of each species from many localities, and are thus invaluable in determining tree distributions. Models of species’ climatic suitability, based on herbarium data, provide a way to expand forest restoration into wider areas [27].
In the study presented here, we mapped the potential distributions of four tree species, previously proven to be “excellent” framework species for restoring upland evergreen forest in northern Thailand [28], using location data from herbarium specimen labels and the Global Biodiversity Information Facility [29], climatic data for each recorded location, and the maximum entropy approach [30]. The maximum entropy algorithm (MaxEnt) allows the use of a small amount of presence-only occurrence data. Development of MaxEnt modelling is still being researched. Modelling species’ niches using MaxEnt relies on the quality of sample data and requires attention at many stages of model calibration [31]. In our study, MaxEnt was selected in preference to new modelling approaches [32,33] due to its 15 years of development and its superiority in terms of computational time and ease of use. The aim of this study was to assist with species selection for FSM trials by mapping the potential distribution of four known framework tree species, thus exploring their potential inclusion in forest restoration trials

2. Materials and Methods

2.1. Studied Species

Four species were selected for this study: Choerospondias axillaris (Roxb.) B.L. Burtt and A.W. Hill (Anacardiaceae), Ficus hispida L.f. (Moraceae), Hovenia dulcis Thunb. (Rhamnaceae), and Prunus cerasoides Buch.-Ham. ex D. Don (Rosaceae) (Table 1, Figure 2). All had been tested in multiple field trials (FORRU-CMU Database) to restore seasonal, upland, evergreen forest (>1000 m from the sea level) in northern Thailand, resulting in their being ranked as “excellent” framework species [28]. All had excellent or acceptable growth and survival over the first two growing seasons after planting out in deforested sites and suppressed weeds well beneath their crowns. All demonstrated excellent or acceptable resilience after forest fires and were attractive to seed-dispersing animals (by producing fleshy fruits or bird nest sites within a few years after planting [34].

2.2. Data Collection

Species location records were gathered from specimen labels at the herbarium of the Royal Botanic Gardens, Kew, United Kingdom and the Global Biodiversity Information Facility [29]. Specimen labels were photographed at Kew from September to November 2014 and the data were subsequently transcribed into a spreadsheet. For those labels without GPS coordinates, location descriptions were checked with US GeoNames [35] and Google Earth Pro (Google Inc., Mountain View, CA, USA) to determine most likely GPS coordinates. Location data from specimens that could not be matched with any location database were discarded.
Species location records in the GBIF datasets were downloaded using the “gbif” function in the “dismo” package [36] written in the R programing language [37]. Search terms were accepted species names [38]. Locations where the species had been artificially introduced were removed. The numbers of records from each of the two data sources are listed in Table 2 for each species. The native geographical ranges differed among the four species (Table 2).
Current climatic data (1960–1990) were obtained from Worldclim V. 1 [39]: a set of global climate layers with a 2.5 arc minute resolution (approximately 20 km2). Nineteen bioclimatic variables (Table 3) were downloaded: (1) annual mean temperature, (2) mean diurnal range (mean of monthly (max temp–min temp)), (3) isothermality, (4) temperature seasonality (standard deviation *100), (5) max temperature of warmest month, (6) min temperature of coldest month, (7) temperature annual range, (8) mean temperature of wettest quarter, (9) mean temperature of driest quarter, (10) mean temperature of warmest quarter, (11) mean temperature of coldest quarter, (12) annual precipitation, (13) precipitation of wettest month, (14) precipitation of driest month, (15) precipitation seasonality (coefficient of variation), (16) precipitation of wettest quarter, (17) precipitation of driest quarter, (18) precipitation of warmest quarter, and (19) precipitation of coldest quarter.

2.3. Species Location Record Preparation

For each species, duplicate records and locations in the ocean were removed. Then, to account for sampling bias, species location records were systematically subsampled [40]. Grid cells were created at one-degree resolution throughout each species’ known range, and then only one record per grid cell was randomly sampled (Figure S1). The numbers of records retained in each cleaning step are presented in Table 2.

2.4. Climatic Data Preparation

To generate a list of noncorrelated variables to use in the model, correlations among 19 variables were identified using Pearson’s correlation coefficient (Table S1). The analyses were carried out using the function “removeCollinearity” in the “virtualspecies” package [41] in R programing language. Pearson’s correlation coefficient was used to compute a distance matrix, which was then used to construct an ascendant hierarchical classification (Figure S2) of the 19 variables. Variables that were 85% correlated were grouped together. Single variables not correlated to any other variables were automatically selected, along with one randomly selected variable from each of the groups, resulting in a total of 10 noncorrelated variables being included in the model. The extent of the climatic variables that were used in MaxEnt modelling was unique for each species, according to the extent of species occurrence. Climatic data were converted to ASCII raster files.

2.5. Modelling

Climatic niche models were calibrated in Maxent 3.4.1 [42]. The operations performed were (i) calibration, (ii) cross-validation, and (iii) projection. Maxent calculated the frequency distributions of all climatic variable values, firstly for all species location record points (species climate frequency) and secondly for a sample of 10,000 randomly selected background points within the species’ range boundaries (geographical climate frequency). If a species shows no climatic preference within its range, then the two proportional frequency distributions would be very similar, because the species would be distributed randomly across its range. If a species is associated with a certain climatic variable value, then the species frequency, at that value, would be higher than the geographical frequency and vice versa. Therefore, the ratio between the two proportional frequency distributions can be used to derive the climatic suitability of each point on the map for growth of a species, within its range.
For model calibration, we applied a “regularization multiplier” of two. The regularization multiplier controls how closely fitted the output distribution is around species record locations. Values of less than 1 result in very tight fit around species record locations, whereas higher values result in wider dispersion [43]. Regularization multipliers of two to four are suggested to reduce “over-fitting” the models [44].
The models were subjected to 10-fold cross-validation. Species location records were randomly split into 10 equal-sized groups (called “folds”). For each species, 10 models were run, each one using data from nine folds (training) with the omitted fold used for validation (by determining if the models reliably predicted species presence at the record locations in the omitted folds). Each of the 10 models was run for up to 500 iterations.
The predictive accuracy of each model was estimated using the area under the receiver-operating characteristic curve (AUC). AUC is interpreted as how much the model can distinguish between species location and background points. An AUC close to 0.5 indicates that model performance is equivalent to random guessing. AUCs greater than 0.8 indicate high model performance [45]. The AUCs were generated and averaged over 10 models for each species.
To identify locations suitable for supporting each species (habitats matching the species’ climatic niche requirements) beyond its current range, the climatic conditions across the whole of tropical and subtropical Asia (latitude 11°S to 40°N and longitude 60°E to 160°E) were used. Maxent model outputs (cloglog outputs) were imported into the R programing language and potential distribution maps were generated. The results were re-grouped into four classes of climatic suitability as: (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable” (see [46]).
In addition, the relative contributions of each climatic variable to explain variance in the MaxEnt models were estimated. The values of each climatic variable on training species location and background data were randomly permutated. Then, the model was re-evaluated using the permutated data and decreases in training AUC were measured (permutation importance) [44]. The permutation importance values were normalized to percentages. A high value of permutation importance indicated that the model depended on the variable to a high degree [44] (see Appendix S1).

3. Results

3.1. Model Performance and Potential Distributions of Four Studied Species

The models predicted species location fairly well for P. cerasoides and C. axillaris, and very well for the other two species. Average AUCs ranged from 0.782 for P. cerasoides to 0.830 for H. dulcis (Table 4, Figure S3). The predicted distributions of suitable habitats according to climatic conditions are presented in Figure 3, Figure 4, Figure 5 and Figure 6. Overall, areas with climates highly suited for C. axillaris tended to be located north of the equator, including parts of China, Vietnam, Thailand, Myanmar, India, Bhutan, and Nepal (Figure 3), whilst those of H. dulcis, F. hispida, and P. cerasoides extended more equally both north and south of the equator (Figure 4, Figure 5 and Figure 6).
The model predicted that the most suitable climatic conditions for C. axillaris occur mostly in SE China and northern Vietnam, where most of the records are located (Figure 3). The population in Thailand appears to exist under suboptimal climatic conditions, towards the southern limit of the species’ range. Although the species has spread westward along the Himalayas, it has yet to be recorded in highly suitable climatic areas in Pakistan, Afghanistan, and Tajikistan, nor has it colonized highly suitable climatic areas in southern Indonesia and Papua New Guinea.
Areas with highly suitable climatic conditions for F. hispida were widely spread from southern China through Indomalaya to Papua New Guinea (latitudes 30° N and 11° S) (Figure 4), although the highest density of records was concentrated in southern China and Indochina. Most of the Philippines appears to have suitable climatic conditions for this species, but F. hispida has not yet been recorded there. The records in central India exist under apparently unfavorable climatic conditions, on the western edge of the species’ range.
The model predicted suitable climatic conditions for H. dulcis both inside and outside the species’ native range. Figure 5 shows two disjunct climatically suitable zones—a northern zone where the species is present and a southern zone where the species has yet to be recorded. The northern zone includes SE China, Taiwan, S. Korea, and Japan. The southern zone includes Malaysia, Indonesia, southern Philippines, and Papua New Guinea. The outlying population in northern Thailand appears to exist under suboptimal climatic conditions at the southernmost edge for the species’ current range.
The recorded range of P. cerasoides is centered in SW China and northern Thailand and Lao PDR, spreading westward along the Himalayas (Figure 6). Scattered records exist under suboptimal climatic conditions in India and Sri Lanka at the western edge of the species’ range. Suitable (but unoccupied) climatic areas south of Thailand occurred sparsely and mostly in highland areas of the Indonesian islands, Borneo, the Philippines, and eastern Papua New Guinea.

3.2. Importance of Climatic Variables

The variables that contributed most to explaining the variance in the climatic niche models varied among the species (Table 4, see Appendix S1). The mean temperature of the driest quarter was the most important variable for C. axillaris and P. cerasoides. The mean temperature of the driest quarter (bio9) was highly correlated with annual mean temperature, minimum temperature of the coldest month, and the mean temperature of the coldest quarter. Precipitation of the driest month was among the top three most important variables for C. axillaris and F. hispida. Precipitation in the driest month was highly correlated with precipitation in the driest quarter. For H. dulcis, the most important variable was precipitation in the coldest quarter.

4. Discussion

4.1. Interpreting the Maps

Based on a comprehensive review of existing records, this study mapped habitat suitability in tropical and subtropical regions for four framework tree species in greater detail than previously reported. The four studied species are previously proven effective at catalyzing forest ecosystem restoration in northern Thailand, when included in mixed-species plantings [28].
Using climatic data from recorded locations, we mapped beyond each species’ recorded range the potential distribution according to climatic conditions conducive for supporting growth of each species (i.e., fundamental climatic niches). Conventionally, outputs from such models have been interpreted as “probability of presence”. However, we interpreted the models’ outputs as an index of “climatic suitability”, since we used only climatic variables in their construction and climate is only part of a species’ fundamental niche. This means that areas on the maps, marked as climatically suitable, may be subject to other niche limitations (e.g., unsuitable soil conditions or lack of biotic factors, such as pollinators, seed dispersers, or other essential symbionts). Dispersal barriers, particularly oceans, are another obvious explanation of why species are absent from areas with a highly suitable climate, hence our substitution of “probability of presence” with “climatic suitability”. This view is supported by Merow at al. (2013) [47], who stated that interpreting Maxent’s predictions as indices of habitat suitability is useful for qualitative exploratory analyses (such as that of the present study).
The maps reliably show where the climate is not suitable for each species. Obviously, planting a species in such locations would be a waste of time and money. However, caution should be exercised when using the maps, to aid species selection decisions for restoration plantings. Although success is highly likely within the species’ now more clearly defined ranges (particularly when mixed with other framework tree species), additional factors must be taken into account when considering climatically suitable areas that are far outside a species’ range, i.e., soil preferences and the presence/absence of essential pollinating and seed-dispersing animals and microbial symbionts. Caution should be applied when considering climatic zones that have long been separated from historical ranges by dispersal barriers (e.g., sea), since where climate is non-limiting, the risk that a species becomes invasive (threatening local flora) is high. This is, of course, anathema to the principles of forest ecosystem restoration [1].

4.2. Predicted Potential Distribution of Choerospondias Axillaris and Contributing Variables

Suitable climatic conditions, to support the growth of C. axillaris, were predicted in both tropical and subtropical parts of the region north of the equator. This species preferred cooler areas of the subtropics, where mean dry season temperature ranged from 5 to 19 °C. This species can occupy habitats with a wide range of precipitation, suggesting a certain degree of drought tolerance. In Nepal, for example, Guatum ranks this species as moderately drought-tolerant [48].
Topography and suboptimal climatic conditions along the Himalayas may prevent C. axillaris from expanding its range west into highly suitable climatic areas in Pakistan, Afghanistan, and Tajikistan. To the south, the Javan sea and a wide band of hostile climatic conditions in northern Indonesia and East Malaysia may have prevented expansion of this species’ range into more climatically suitable areas on Java, the Lesser Sunda Islands, Irian Jaya, and Papua New Guinea. C. axillaris produces fleshy fruits that are very attractive to a diverse range of terrestrial and arboreal vertebrate frugivores [49]. This contributes to its framework species status, since such animals disperse seeds into restoration sites from nearby forest remnants, thus promoting biodiversity recovery. However, it may also limit the species’ capability to expand its range over long distances, across large areas where a suitable habitat for such animals does not exist.
C. axillaris is not known to be invasive [50].

4.3. Predicted Potential Distribution of Ficus Hispida and Contributing Variables

F. hispida is very widespread, with the present range and geographic extent of its predicted climatic niche closely matching previous reports [51]. Of the four species studied, F. hispida had the widest climatic suitability range. In decreasing order of importance, precipitation of the driest month, precipitation seasonality, and mean temperature of the dry season most explained the species’ current geographic range. F. hispida inhabits areas where variability in monthly precipitation is low or high, relative to mean annual precipitation (precipitation seasonality ~21%–95%). The species may be highly drought-tolerant, since it is predicted to grow where precipitation of the driest month falls to ~8 mm.
This species occupies most of the areas where the models predicted that climatic conditions fall within the species’ climatic niche, with the notable exception of the Philippines. The models predicted that climatic conditions throughout the Philippines were suitable for F. hispida. However, the species has not yet been recorded in the Philippines. Its absence from the Philippines may be explained by the fact that the country was not connected to the rest of the Sunda Shelf islands when sea levels fell during ice ages [52]. Figs of F. hispida are highly coevolved with fig pollination wasps (Ceratosolen solmsi marchali Mayr in Chalcidoidae: Agaomidae) [53]. Another possible explanation for the absence of F. hispida from the Philippines may be that other factors, i.e., nonclimatic factors, may be detrimental to the survival of the pollinators.
F. hispida is not known to be invasive, although other Ficus species are, e.g., F. macrocarpa [50].

4.4. Predicted Potential Distribution of Hovenia Dulcis and Contributing Variables

High precipitation favors H. dulcis. The first two most contributing variables were precipitation of the coldest and warmest quarter. H. dulcis requires humidity during the cold season (>87 mm with optimum value of 558 mm) and summer (range ~448–2127 mm with optimum value of 526 mm). In addition, mean temperatures in the wettest quarter ranging from 18 to 25 °C during the rainy season were suitable for this species. Northern Thailand marks the southern limit of H. dulcis’ known distribution [54].
Unsuitable climatic conditions in central Thailand and dispersal limitation may prevent spread of the species southwards. The species has not yet been recorded in highly climatically suitable areas, predicted in the Malay Peninsula, Borneo, western Philippines, and most parts of Indonesia. Seed dispersal of this species is by medium-sized birds such as wedge-tailed green pigeons [55,56]. Even though the range of wedge-tailed green pigeons stretches from southern China to Indonesia, dispersal distance by the birds may not be far enough to deliver seeds escaping unsuitable climate conditions of central Thailand.
H. dulcis is well-known as an invasive species in Brazil and Tanzania [57,58,59,60]. In Brazil, black capuchin monkeys consume the fruit and may facilitate seed dispersal [58]. Figure 5 illustrates the potential of this species to become invasive in Malaysia, the Philippines, and Indonesia. Bergamin and his team modelled the distribution of H. dulcis worldwide under various climate change scenarios [60]. They predicted that the species range is likely to contract slightly at low and medium latitudes, but with a strong potential for range expansion in the northern boreal zone [60]. They stated that “it is essential to implement policies to prevent H. dulcis introduction in suitable areas worldwide, as well as local population control, especially in biodiversity hotspots. We concur that this species should not be introduced for forest restoration projects outside of its native range, despite it being an excellent framework tree species in northern Thailand” [28].

4.5. Predicted Potential Distribution of Prunus Cerasoides and Contributing Variables

P. cerasoides grows in low-temperature habitats. Mean temperature of the driest quarter was the most important climatic variable. The upper limits of suitable ranges of mean temperature of dry season and summer were relatively low: 19 and 26 °C, respectively. Furthermore, suitable climatic conditions included medium temperature variability within a month relative to the year (~43%–63%). In Thailand, relatively low temperatures are found at high elevations (>1000 m asl.), resulting in the occurrence of P. cerasoides in upland evergreen forest, evergreen and deciduous forest, and evergreen and pine forest, particularly in disturbed areas and the forest edge [61]. Our findings suggest that high temperature and high temperature fluctuations limit the distribution of this species.
Despite highly suitable climatic conditions in northern Myanmar, P. cerasoides has not been recorded there, abruptly truncating the recorded distribution of the species at the national border between Myanmar, China, and Thailand. This suggests that a lack of collection effort in Myanmar may be responsible for the lack of records there. Its occurrence in the mountains of northern Myanmar is therefore highly likely.
The species has been introduced to Zimbabwe [62], Australia [63], New Zealand [64], and Japan [65]. It is not known to be invasive [50].

4.6. Study Limitations and Future Direction for Species Distribution Modelling for Forest Restoration

One of the limitations of our study may have been sampling bias in herbarium specimen collection. Herbarium specimens are more likely to be collected closer to roads and herbaria [66]. Furthermore, the density and distribution of the species records are likely to be influenced by both geographical accessibility and political history. For example, the GBIF database contained 0.08 records of vascular plant species per square kilometer in Myanmar compared with 0.25 records per square kilometer in China [29]. This may indicate lower sampling effort in Myanmar in comparison to China. The missing data on species presence affect the differentiation of the environmental conditions between the occurrence points and the background points [67]. In our study, we addressed sampling bias by (1) systematic subsampling and (2) restricting the calibration of the model to the extent of species occurrence data and obtaining the prediction by projecting potential distributions across tropical and subtropical Asia [67]. In this study, we removed duplicate records and used systematic subsampling to account for sampling bias [40]. It is important to remove duplicate records because they lead to overfitting of the models and overestimating habitat suitability in unsuitable sites. In addition, we systematically sampled the occurrence data, to account for sampling bias. However, this subsampling method may underestimate the contribution of highly suitable areas of occurrence, where the high density of records truly reflected the commonness of the species in those areas.
Further studies may include model calibration with more environmental variables and field validations of model predictions. Adding other environmental variables to the models may produce more accurate predictions of species potential distributions, e.g., soil variables [68,69] such as soil type, pH, and nitrogen content [70]. Because some soil properties are likely to be influenced by climatic conditions [71], correlations among the variables should still be considered when selecting variables to include in the models. In addition, independent field validation of models is scarce, especially in forest restoration ecology. To test whether species can grow where models predict suitable habitat, classic transfer experiments should be performed—transplanting seedlings from their home range to the predicted suitable habitat.
Potential distributions of more species should be determined to improve future forest restoration projects. Furthermore, modeling how the distribution of species niches might change as a result of climate change is also important (e.g., [25,72,73]) to help climate-proof future forest restoration initiatives. This includes site selection, the selection of indigenous tree species, and the potential need for assisted species migration.

5. Conclusions

The climatic niches of four tree species were modelled using MaxEnt. Potentially suitable habitats based on climatic conditions were mapped both inside and outside their native ranges across tropical and subtropical Asia. The four species studied—Choerospondias axillaris, Ficus hispida, Hovenia dulcis, and Prunus cerasoides—had already been classified as excellent tree species for restoring evergreen seasonal forests [28]. The important variables contributing to their distribution differed among species. Precipitation in the driest month contributed most to determining the distribution of F. hispida. Mean temperature of the dry season contributed the most to the species distribution of C. axillaris and P. cerasoides, while for H. dulcis the precipitation of the cold season was the most important contributing factor. The predictive potential distribution maps can be used to facilitate species selection in restoration trials and to determine where not to plant, because a species may become invasive. While this study establishes the usefulness of the techniques used, their application to a much large number of species is needed, to significantly improve species selection for forest restoration initiatives.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/f13070993/s1, Figure S1: An example of systematic subsampling of the occurrence data; Table S1: Pearson’s correlation coefficient between each pair of climatic variables of tropical and subtropical Asia region; Figure S2. An ascendant hierarchical clustering classification of 19 bioclimatic variables into 10 internally correlated groups (see Table 2 for full variable names); Figure S3: The area under the receiver-operating characteristic curve (AUC) of each species.; Appendix S1: Analysis of variable contributions of each species.

Author Contributions

Conceptualization, S.D.E. and P.T.; methodology, P.T.; formal analysis, P.T.; investigation, P.T.; data curation, P.T.; writing—original draft preparation, P.T.; writing—review and editing, P.T., S.D.E. and P.W.; visualization, P.T. and S.D.E.; supervision, S.D.E. and P.W.; funding acquisition, P.T. All authors have read and agreed to the published version of the manuscript.

Funding

The Institute for the Promotion of Teaching Science and Technology (IPST), Thailand, for financial support under the Research Fund for DPST Graduate with First Placement 2015 (Grant number 028/2558). This research work was partially supported by Chiang Mai University.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data in this study are available on request from the first author.

Acknowledgments

We thank Kate Hardwick for her support with this project and for her help in coordinating with the Royal Botanic Gardens, Kew; Timothy Utteridge for his guidance during data collection at the Kew Herbarium; and Justin Moat for his technical advice. We thank interns of the Forest Restoration Research Unit for helping with label transcription. We also thank the British Council for partial travel support, under the Researcher Links Travel Grant initiative. This research was funded by the Institute for the Promotion of Teaching Science and Technology (IPST), Thailand under the Research Fund for DPST Graduate with First Placement 2015 (Grant number 028/2558). This research work was partially supported by Chiang Mai University.

Conflicts of Interest

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; or in the writing of the manuscript.

References

  1. Elliott, S.; Blakesley, D.; Hardwick, K. Restoring Tropical Forests: A Practical Guide; Royal Botanic Gardens, Kew: London, UK, 2013; p. 344. [Google Scholar]
  2. Catterall, C.P.; Kanowski, J.; Wardell-Johnson, G.W. Biodiversity and New Forests: Interacting Processes, Prospects and Pitfalls of Rainforest Restoration. In Living in a Dynamic Tropical Forest Landscape; Stork, N.E., Turton, S.M., Eds.; Wiley-Blackwell: New York, NY, USA, 2009; pp. 510–525. [Google Scholar]
  3. Tucker, N.I.J.; Simmons, T. Restoring a rainforest habitat linkage in north Queensland: Donaghy’s Corridor. Ecol. Manag. Restor. 2009, 10, 98–112. [Google Scholar] [CrossRef]
  4. Lewis, S.L.; Wheeler, C.E.; Mitchard, E.T.A.; Koch, A. Restoring natural forests is the best way to remove atmospheric carbon. Nature 2019, 568, 25–28. [Google Scholar] [CrossRef] [PubMed]
  5. IUCN. Restore Our Future: The Bonn Challenge. Available online: https://www.bonnchallenge.org/ (accessed on 18 May 2020).
  6. UNEP; FAO. The UN Decade on Ecosystem Restoration 2021–2030. Available online: https://wedocs.unep.org/bitstream/handle/20.500.11822/30919/UNDecade.pdf?sequence=11 (accessed on 6 July 2020).
  7. United Nations. Report of the Conference of the Parties on Its Thirteenth Session. In Proceedings of the Conference of the Parties on Its Thirteenth Session, Bali, Indonesia, 3–15 December 2007. [Google Scholar]
  8. United Nations. Report of the Conference of the Parties on Its Sixteenth Session. In Proceedings of the Conference of the Parties on Its Sixteenth Session, Cancun, Mexico, 29 November–10 December 2010. [Google Scholar]
  9. Lamb, D. Large-Scale Forest Restoration; Routledge: London, UK, 2014. [Google Scholar]
  10. Barlow, J.; Gardner, T.A.; Araujo, I.S.; Ávila-Pires, T.C.; Bonaldo, A.B.; Costa, J.E.; Esposito, M.C.; Ferreira, L.V.; Hawes, J.; Hernandez, M.I.M.; et al. Quantifying the biodiversity value of tropical primary, secondary, and plantation forests. Proc. Natl. Acad. Sci. USA 2007, 104, 18555–18560. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  11. Goosem, S.; Tucker, N.I.J. Repairing the Rainforest, 2nd ed.; Wet Tropics Management Authority and Biotropica Australia Pty. Ltd: Cairns, Australia, 2013; p. 153. [Google Scholar]
  12. Sobon, K. Identifying Framework Tree Species for Restoring Forest Ecosystems in Siem Reap Province; Royal University of Agriculture: Phnom Penh, Cambodia, 2013. [Google Scholar]
  13. Harrison, R.; Swinfield, T. Restoration of logged humid tropical forests: An experimental programme at Harapan Rainforest, Indonesia. Trop. Conserv. Sci. 2015, 8, 4–16. [Google Scholar] [CrossRef] [Green Version]
  14. Weyerhaeuser, H.; Kahrl, F. Planting trees on farms in Southwest China: Enhancing rural economies and the environment. Mt. Res. Dev. 2006, 26, 205–208. [Google Scholar] [CrossRef] [Green Version]
  15. Marshall, A. (Udzungwa Forest Project, Morogoro, Tanzania). Restoration Project in Tanzania. Personal communication, 2021.
  16. WeForest. WeForest Project Philippines: Restoring the Sarangani Forest. Available online: https://www.weforest.org/project/philippines (accessed on 16 May 2020).
  17. Toktang, T. The Effects of Forest Restoration on the Species Diversity and Composition of a Bird Community in Doi Suthep-Pui National Park Thailand from 2002–2003; Chiang Mai University: Chaing Mai, Thailand, 2005. [Google Scholar]
  18. Sinhaseni, K. Natural Establishment of Tree Seedlings in Forest Restoration Trials at Ban Mae Sa Mai, Chiang Mai Province; Chiang Mai University: Chiang Mai, Thailand, 2008. [Google Scholar]
  19. Jantawong, K.; Elliott, S.; Wangpakapattanawong, P. Above-ground carbon sequestration during restoration of upland evergreen forest in northern Thailand. Open J. For. 2017, 7, 157–171. [Google Scholar] [CrossRef] [Green Version]
  20. Kavinchan, N.; Wangpakapattanawong, P.; Elliott, S.; Chairuangsri, S.; Pinthong, J. Use of the framework species method to restore carbon flow via litterfall and decomposition in an evergreen tropical forest ecosystem, northern Thailand. Kasetsart J.-Nat. Sci. 2015, 49, 639–650. [Google Scholar]
  21. Di Sacco, A.; Hardwick, K.A.; Blakesley, D.; Brancalion, P.H.S.; Breman, E.; Cecilio Rebola, L.; Chomba, S.; Dixon, K.; Elliott, S.; Ruyonga, G.; et al. Ten golden rules for reforestation to optimize carbon sequestration, biodiversity recovery and livelihood benefits. Glob. Chang. Biol. 2021, 27, 1328–1348. [Google Scholar] [CrossRef]
  22. Jantawong, K.; Kavinchan, N.; Wangpakapattanawong, P.; Elliott, S. Financial analysis of potential carbon value over 14 years of forest restoration by the framework species method. Forests 2022, 13, 144. [Google Scholar] [CrossRef]
  23. Gastón, A.; García-Viñas, J.I.; Bravo-Fernández, A.J.; López-Leiva, C.; Oliet, J.A.; Roig, S.; Serrada, R. Species distribution models applied to plant species selection in forest restoration: Are model predictions comparable to expert opinion? New For. 2014, 45, 641–653. [Google Scholar] [CrossRef]
  24. Thomas, E.; Alcázar Caicedo, C.; Moscoso, L.G.; Vásquez, A.; Osorio, L.F.; Salgado-Negret, B.; Ramírez, W. The importance of species selection and seed sourcing in forest restoration for enhancing adaptive capacity to climate change: Colombian tropical dry forest as a model. CBD Tech. Ser. 2017, 89, 122–132. [Google Scholar]
  25. Fremout, T.; Thomas, E.; Gaisberger, H.; Van Meerbeek, K.; Muenchow, J.; Briers, S.; Gutierrez-Miranda, C.E.; Marcelo-Peña, J.L.; Kindt, R.; Atkinson, R.; et al. Mapping tree species vulnerability to multiple threats as a guide to restoration and conservation of tropical dry forests. Glob. Chang. Biol. 2020, 26, 3552–3568. [Google Scholar] [CrossRef] [PubMed]
  26. Hardwick, K.A.; Fiedler, P.; Lee, L.C.; Pavlik, B.; Hobbs, R.J.; Aronson, J.; Bidartondo, M.; Black, E.; Coates, D.; Daws, M.I.; et al. The role of botanic gardens in the science and practice of ecological restoration. Conserv. Biol. 2011, 25, 265–275. [Google Scholar] [CrossRef] [PubMed]
  27. Sinclair, S.J.; White, M.D.; Newell, G.R. How useful are species distribution models for managing biodiversity under future climates? Ecol. Soc. 2010, 15, 8. [Google Scholar] [CrossRef]
  28. Elliott, S.; Navakitbumrung, P.; Kuarak, C.; Zangkum, S.; Anusarnsunthorn, V.; Blakesley, D. Selecting framework tree species for restoring seasonally dry tropical forests in northern Thailand based on field performance. For. Ecol. Manag. 2003, 184, 177–191. [Google Scholar] [CrossRef]
  29. Derived Dataset GBIF.org Filtered Export of GBIF Occurrence Data. Available online: https://doi.org/10.15468/dd.cg253q (accessed on 2 May 2022).
  30. Phillips, S.J.; Dudik, M.; Schapire, R.E. A Maximum Entropy Approach to Species Distribution Modeling. In Proceedings of the International Conference on Machine Learning (ICML’04), Banff, AB, Canada, 4–8 July 2004. [Google Scholar]
  31. Lissovsky, A.; Dudov, S. Species-distribution modeling: Advantages and limitations of its application. 2. MaxEnt. Biol. Bull. Rev. 2021, 11, 265–275. [Google Scholar] [CrossRef]
  32. Hao, T.; Elith, J.; Lahoz-Monfort, J.J.; Guillera-Arroita, G. Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models. Ecography 2020, 43, 549–558. [Google Scholar] [CrossRef] [Green Version]
  33. Kaky, E.; Nolan, V.; Alatawi, A.; Gilbert, F. A comparison between Ensemble and MaxEnt species distribution modelling approaches for conservation: A case study with Egyptian medicinal plants. Ecol. Inform. 2020, 60, 101150. [Google Scholar] [CrossRef]
  34. Chiang Mai University Forest Restoration Research Unit. How to Plant a Forest: The Principels and Practice of Restoring Tropical Forests; Biology Department, Science Faculty, Chiang Mai University: Chiang Mai, Thailand, 2005. [Google Scholar]
  35. GeoNames Team. GeoNames. Available online: http://www.geonames.org/ (accessed on 21 July 2018).
  36. Hijmans, R.J.; Phillips, S.; Leathwick, J.; Elith, J. Dismo: Species Distribution Modeling. Available online: https://CRAN.R-project.org/package=dismo (accessed on 2 May 2022).
  37. R Core Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2019. [Google Scholar]
  38. IPNI. International Plant Names Index. The Royal Botanic Gardens, Kew, Harvard University Herbaria and Libraries and Australian National Botanic Gardens; Available online: http://www.ipni.org (accessed on 2 May 2022).
  39. Hijmans, R.J.; Cameron, S.E.; Parra, J.L.; Jones, P.G.; Jarvis, A. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 2005, 25, 1965–1978. [Google Scholar] [CrossRef]
  40. Fourcade, Y.; Engler, J.O.; Rödder, D.; Secondi, J. Mapping species distributions with MAXENT using a geographically biased sample of presence data: A performance assessment of methods for correcting sampling bias. PLoS ONE 2014, 9, e97122. [Google Scholar] [CrossRef] [Green Version]
  41. Leroy, B.; Meynard, C.N.; Bellard, C.; Courchamp, F. virtualspecies, an R package to generate virtual species distributions. Ecography 2015, 39, 599–607. [Google Scholar] [CrossRef]
  42. Phillips, S.J.; Dudík, M.; Schapire, R.E. Maxent Software for Modeling Species Niches and Distributions. Available online: https://biodiversityinformatics.amnh.org/open_source/maxent/ (accessed on 27 September 2019).
  43. Phillips, S.J. A Brief Tutorial on Maxent. Available online: https://biodiversityinformatics.amnh.org/open_source/maxent/ (accessed on 27 September 2019).
  44. Radosavljevic, A.; Anderson, R.P. Making better Maxent models of species distributions: Complexity, overfitting and evaluation. J. Biogeogr. 2014, 41, 629–643. [Google Scholar] [CrossRef]
  45. Phillips, S.J.; Anderson, R.P.; Schapire, R.E. Maximum entropy modeling of species geographic distributions. Ecol. Model. 2006, 190, 231–259. [Google Scholar] [CrossRef] [Green Version]
  46. Liu, Y.; Huang, P.; Lin, F.; Yang, W.; Gaisberger, H.; Christopher, K.; Zheng, Y. MaxEnt modelling for predicting the potential distribution of a near threatened rosewood species (Dalbergia cultrata Graham ex Benth). Ecol. Eng. 2019, 141, 105612. [Google Scholar] [CrossRef]
  47. Merow, C.; Smith, M.J.; Silander, J.A., Jr. A practical guide to MaxEnt for modeling species’ distributions: What it does, and why inputs and settings matter. Ecography 2013, 36, 1058–1069. [Google Scholar] [CrossRef]
  48. Guatum, K.H. Lapsi (Choerospondias axillaris) emerging as a commercial non-timber forest product in the hills of Nepal. In Forest Products, Livelihoods and Conservation: Case Studies of Non-Timber Forest Product Systems—Asia; Kusters, K., Belcher, B., Eds.; Center for International Forestry Research: Bogor Regency, Indonesia, 2004; Volume 1, pp. 114–129. [Google Scholar]
  49. Chanthorn, W.; Brockelman, W. Seed dispersal and seedling recruitment in the light-demanding tree Choerospondias axillaris in old-growth forest in Thailand. ScienceAsia 2008, 34, 129–135. [Google Scholar] [CrossRef] [Green Version]
  50. Invasive Species Specialist Group. Global Invasive Species Database; Invasive Species Specialist Group: Rome, Italy, 2020. [Google Scholar]
  51. Rojo, J.P.; Pitargue, F.C.; Sosef, M.S.M.; Ficus Hispida, L.F. Plant Resources of South-East Asia No 12: Medicinal and Poisonous Plants; de Padua, L.S., Bunyapraphatsara, N., Lemmens, R.H.M.J., Eds.; PROSEA Foundation: Bogor, Indonesia, 1999; Volume 12. [Google Scholar]
  52. Voris, H.K. Special paper 2: Maps of Pleistocene sea levels in Southeast Asia: Shorelines, river systems and time durations. J. Biogeogr. 2000, 27, 1153–1167. [Google Scholar] [CrossRef] [Green Version]
  53. Yang, D.; Peng, Y.; Song, Q.; Zhang, G.; Wang, R.; Zhao, T.; Wang, Q. Pollination biology of Ficus hispida in the tropical rainforests of Xishuangbanna, China. Acta Bot. Sin. 2002, 44, 519–526. [Google Scholar]
  54. Maxwell, J.F. Botanical Notes on the Flora of Thailaod: 4. Nat. Hist. Bull. Siam Soc. 1994, 42, 4. [Google Scholar]
  55. Kopachon, S.; Suriya, K.; Hardwick, K.; Pakaad, G.; Maxwell, J.F.; Anusarnsunthorn, V.; Blakesley, D.; Garwood, N.C.; Elliott, S. Forest restoration research in northern Thailand, 1. The fruits, seeds and seedlings of Hovenia dulcis Thunb. (Rhamnaceae). Nat. Hist. Bull. Siam Soc. 1996, 44, 12. [Google Scholar]
  56. Hitchcock, D.; Elliott, S. Forest restoration research in northern Thailand, III : Observations of birds feeding in mature Hovenia dulcis Thunb. (Rhamnaceae). Nat. Hist. Bull. Siam Soc. 1999, 47, 149–152. [Google Scholar]
  57. Zenni, R.D.; Ziller, S.R. An overview of invasive plants in Brazil. Braz. J. Bot. 2011, 34, 431–446. [Google Scholar] [CrossRef] [Green Version]
  58. Hendges, C.; Fortes, V.B.; Dechoum, M. Consumption of the invasive alien species Hovenia dulcis Thumb. (Rhamnaceae) by Sapajus nigritus Kerr, 1792 in a protected area in southern Brazil. Rev. Bras. De Zoociências 2012, 14, 255–260. [Google Scholar]
  59. Penny, J. Plant 100: Hovenia Dulcis Thunb. (Rhamnaceae); Department of Plant Sciences, University of Oxford: Oxford, UK, 2020. [Google Scholar]
  60. Bergamin, R.S.; Gama, M.; Almerão, M.; Hofmann, G.S.; Anastácio, P.M. Predicting current and future distribution of Hovenia dulcis Thunb. (Rhamnaceae) worldwide. Biol. Invasions 2022. [Google Scholar] [CrossRef]
  61. The Forest Restoration Research Unit. Tree Seeds and Seedlings for Restoring Forests in Northern Thailand; Kerby, J., Elliott, S., Blakesley, D., Anusarnsunthorn, V., Eds.; Biology Department, Science Faculty, Chiang Mai University: Chiang Mai, Thailand, 2000; p. 152. [Google Scholar]
  62. Mapaura, A.; Timberlake, J. (Eds.) A Checklist of Zimbabwean Vascular Plants; Southern African Botanical Diversity Network Report No. 33 Sabonet; Southern African Botanical Diversity Network (SABONET): Pretoria, South Africa; Available online: http://opus.sanbi.org/bitstream/20.500.12143/5966/1/Mapaura_et_al_2004.pdf (accessed on 2 May 2022).
  63. Atlas of Living Australia. Species Page. 2020. Available online: https://bie.ala.org.au/species/NZOR-6-116741 (accessed on 14 August 2020).
  64. Herbarium, A. Prunus cerasoides Records. 2020. Available online: https://scd.landcareresearch.co.nz/ (accessed on 17 May 2020).
  65. Kobayashi, Y. Cherry blossoms in Japan. Jpn. Agric. Res. Q. 1983, 17, 5. [Google Scholar]
  66. Daru, B.H.; Park, D.S.; Primack, R.B.; Willis, C.G.; Barrington, D.S.; Whitfeld, T.J.S.; Seidler, T.G.; Sweeney, P.W.; Foster, D.R.; Ellison, A.M.; et al. Widespread sampling biases in herbaria revealed from large-scale digitization. New Phytol. 2018, 217, 939–955. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  67. Phillips, S.J. Transferability, sample selection bias and background data in presence-only modelling: A response to Peterson et al. (2007). Ecography 2008, 31, 272–278. [Google Scholar] [CrossRef]
  68. Velazco, S.J.E.; Galvão, F.; Villalobos, F.; De Marco Júnior, P. Using worldwide edaphic data to model plant species niches: An assessment at a continental extent. PLoS ONE 2017, 12, e0186025. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  69. Zuquim, G.; Costa, F.R.C.; Tuomisto, H.; Moulatlet, G.M.; Figueiredo, F.O.G. The importance of soils in predicting the future of plant habitat suitability in a tropical forest. Plant Soil 2020, 450, 151–170. [Google Scholar] [CrossRef] [Green Version]
  70. Dubuis, A.; Giovanettina, S.; Pellissier, L.; Pottier, J.; Vittoz, P.; Guisan, A. Improving the prediction of plant species distribution and community composition by adding edaphic to topo-climatic variables. J. Veg. Sci. 2013, 24, 593–606. [Google Scholar] [CrossRef]
  71. Thuiller, W. On the importance of edaphic variables to predict plant species distributions–limits and prospects. J. Veg. Sci. 2013, 24, 591–592. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  72. Trisurat, Y.; Shrestha, R.P.; Kjelgren, R. Plant species vulnerability to climate change in Peninsular Thailand. Appl. Geogr. 2011, 31, 1106–1114. [Google Scholar] [CrossRef] [Green Version]
  73. Khan, A.M.; Li, Q.; Saqib, Z.; Khan, N.; Habib, T.; Khalid, N.; Majeed, M.; Tariq, A. MaxEnt modelling and impact of climate change on habitat suitability variations of economically important Chilgoza Pine (Pinus gerardiana Wall.) in South Asia. Forests 2022, 13, 715. [Google Scholar] [CrossRef]
Forests 13 00993 g001 550
Figure 1. How the framework species method accelerates forest ecosystem restoration. Reprinted with permission from Stephen D. Elliott, first published in Jantawong et al. 2022 [22]. Dotted lines indicate positive feedback loops, by which planted framework tree species intensify key mechanisms of natural regeneration: weed suppression, seed–rain enhancement, and creation of conditions conducive to the establishment of recruit tree species.
Figure 1. How the framework species method accelerates forest ecosystem restoration. Reprinted with permission from Stephen D. Elliott, first published in Jantawong et al. 2022 [22]. Dotted lines indicate positive feedback loops, by which planted framework tree species intensify key mechanisms of natural regeneration: weed suppression, seed–rain enhancement, and creation of conditions conducive to the establishment of recruit tree species.
Forests 13 00993 g001
Forests 13 00993 g002 550
Figure 2. Foliage and fruit of four studied species: (a) Choerospondias axillaris (Roxb.) B.L. Burtt and A.W. Hill, (b) Ficus hispida L.f., (c) Hovenia dulcis Thunb., and (d) Prunus cerasoides Buch.-Ham. ex D. Don.
Figure 2. Foliage and fruit of four studied species: (a) Choerospondias axillaris (Roxb.) B.L. Burtt and A.W. Hill, (b) Ficus hispida L.f., (c) Hovenia dulcis Thunb., and (d) Prunus cerasoides Buch.-Ham. ex D. Don.
Forests 13 00993 g002
Forests 13 00993 g003 550
Figure 3. Choerospondias axillaris: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Figure 3. Choerospondias axillaris: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Forests 13 00993 g003
Forests 13 00993 g004 550
Figure 4. Ficus hispida: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Figure 4. Ficus hispida: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Forests 13 00993 g004
Forests 13 00993 g005 550
Figure 5. Hovenia dulcis: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Figure 5. Hovenia dulcis: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Forests 13 00993 g005
Forests 13 00993 g006 550
Figure 6. Prunus cerasoides: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Figure 6. Prunus cerasoides: confirmed records (brown circles) and potential current climate suitability, modelled from climatic variables (green)—darker green represents higher climatic suitability. Four classes of climatic suitability are (i) <0.1 = “unsuitable”, (ii) 0.1–0.3 = “low suitability”, (iii) 0.3–0.6 = “suitable”, and (iv) >0.6 = “highly suitable”.
Forests 13 00993 g006
Table
Table 1. Information of the four species. Overall, they are excellent framework tree species for forest restoration.
Table 1. Information of the four species. Overall, they are excellent framework tree species for forest restoration.
SpeciesFamilyElevation (m asl) 1Framework Species Assessment 2
LowerUpperSurvival 3Growth 4Weed
Suppression 5		Fire
Resilience 6		Attraction to Seed Dispersers 7
Choerospondias axillaris (Roxb.) B.L. Burtt and A.W. HillAnacardiaceae4601600EEAEFlowering and fruiting from 4th year. Fruits attract seed-dispersing mammals.
Ficus hispida L.f.Moraceae601525EAEEFigs from 3rd year attract seed-dispersing birds/squirrels.
Hovenia dulcis Thunb.Rhamnaceae10251300EEEEFruit and infructescence attract seed-dispersing birds but flowers late: >8 years after planting.
Prunus cerasoides Buch.-Ham. ex D. DonRosaceae10501750EEEAFlowers, fruit, and bird nests within 3 years. Fruits attract seed-dispersing birds.
1 CMU-B Herbarium Database; 2 Elliott et al., (2003) [28]; 3 E > 70% (mean at end of 2nd growing season); 4 E > 2:0 m, A = 1:5–1.99 m (mean height at end of 2nd growth season); 5 E > 33%, A 16.5%–33% (mean reduction in weed cover beneath tree crown at end of 2nd growth season); 6 E > 70%, A = 50%–69.9% (mean survival after fire); 7 Forest Restoration Research Unit (2005) [34].
Table
Table 2. Species and the number of data points at different steps of data preparation.
Table 2. Species and the number of data points at different steps of data preparation.
SpeciesHerbarium RecordsGBIF Records with Coordinate Data 1Records within the Native RangeSubsampled RecordsNative Geographical Extent of Location DataRange Description
C. axillaris57361183116Longitude 77–142° E
Latitude 18–40° N		India to China
Southeast/south-central China to Thailand
F. hispida601221756236Longitude 76–154° E
Latitude 20°S–38° N		India to Australia
Australia to Southeast/south-central China
H. dulcis70637391101Longitude 77–136° E
Latitude 11–36° N		India to Japan
Japan and China to Thailand
P. cerasoides4621313968Longitude 76–113° E
Latitude 6–31° N		India to Thailand
South China to Thailand
Total23424321481524
1 1 September 2019.
Table
Table 3. Nineteen climatic variables and their abbreviations. After removing collinearity among the variables, 10 variables were used in the model (variables with ✓).
Table 3. Nineteen climatic variables and their abbreviations. After removing collinearity among the variables, 10 variables were used in the model (variables with ✓).
AbbreviationVariables (Unit)NoteUsed in Modelling
Bio1Annual mean temperature (°C)
Bio2Mean diurnal range (mean of monthly) (°C)max temp–min temp✓
Bio3Isothermality (%) 1(Bio2/Bio7) × 100✓
Bio4Temperature seasonality (°C) 2standard deviation
Bio5Max temperature of warmest month (°C)
Bio6Min temperature of coldest month (°C)
Bio7Temperature annual range (°C)Bio5–Bio6
Bio8Mean temperature of wettest quarter (°C) ✓
Bio9Mean temperature of driest quarter (°C) ✓
Bio10Mean temperature of warmest quarter (°C) ✓
Bio11Mean temperature of coldest quarter (°C)
Bio12Annual precipitation (mm)
Bio13Precipitation of wettest month (mm) ✓
Bio14Precipitation of driest month (mm) ✓
Bio15Precipitation seasonality (%) 3coefficient of variation✓
Bio16Precipitation of wettest quarter (mm)
Bio17Precipitation of driest quarter (mm)
Bio18Precipitation of warmest quarter (mm) ✓
Bio19Precipitation of coldest quarter (mm) ✓
1 A value of 100 indicates the diurnal temperature range is equivalent to the annual temperature range. A value less than 100 represents a smaller level of temperature variability within an average month relative to the year. 2 The larger the standard deviation, the greater the variability of temperature. 3 The ratio of the standard deviation of the monthly total precipitation to the mean monthly total precipitation. A measure of the variation in monthly precipitation totals over the course of the year.
Table
Table 4. Area under the receiver-operating characteristic curve (AUC), averaged over 10 models with one standard deviation and the importance of permutation of bioclimatic variables contributing to climatic niche models of each species. The first three most important variables and the average suitable range and optimum value of those variables are presented.
Table 4. Area under the receiver-operating characteristic curve (AUC), averaged over 10 models with one standard deviation and the importance of permutation of bioclimatic variables contributing to climatic niche models of each species. The first three most important variables and the average suitable range and optimum value of those variables are presented.
SpeciesAverage AUCPermutation Importance (Normalized Percentage) and Suitable Range of Variables
(1 Standard Deviation)The FirstSuitable Range (Optimal Value)The SecondSuitable Range (Optimal Value)The ThirdSuitable Range (Optimal Value)
C. axillaris0.783 (0.062)Mean temperature of driest quarter (36.3%)4.7–18.8 °C (13.9 °C)Precipitation of driest month (19.3%)8.8–81.6 mm (22.5 mm)Precipitation of wettest month (11.8%)161.0–397.7 mm (215.3 mm)
F. hispida0.811 (0.039)Precipitation of driest month (20%)≥7.9 mm (14.3 mm)Precipitation seasonality (17.4%)21–95% (72.4%)Mean temperature of driest quarter (°C) (15.4%)11.3–27.0 °C (25.6 °C)
H. dulcis0.830 (0.059)Precipitation of coldest quarter (32.3%)≥87 mm (558 mm)Precipitation of warmest quarter (24.2%)448.0–2127.2 mm (526.0 mm)Mean temperature of wettest quarter (9.6%)18.5–25.1 °C (23.1 °C)
P. cerasoides0.782 (0.079)Mean temperature of driest quarter (59.5%)4.3–18.8 °C (11.1 °C)Mean temperature of warmest quarter (27.9%)12.7–25.6 °C (18.4 °C)Isothermality (9.6%)43.3–62.6% (49.2%)
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Tiansawat, P.; Elliott, S.D.; Wangpakapattanawong, P. Climate Niche Modelling for Mapping Potential Distributions of Four Framework Tree Species: Implications for Planning Forest Restoration in Tropical and Subtropical Asia. Forests 2022, 13, 993. https://doi.org/10.3390/f13070993

AMA Style

Tiansawat P, Elliott SD, Wangpakapattanawong P. Climate Niche Modelling for Mapping Potential Distributions of Four Framework Tree Species: Implications for Planning Forest Restoration in Tropical and Subtropical Asia. Forests. 2022; 13(7):993. https://doi.org/10.3390/f13070993

Chicago/Turabian Style

Tiansawat, Pimonrat, Stephen D. Elliott, and Prasit Wangpakapattanawong. 2022. "Climate Niche Modelling for Mapping Potential Distributions of Four Framework Tree Species: Implications for Planning Forest Restoration in Tropical and Subtropical Asia" Forests 13, no. 7: 993. https://doi.org/10.3390/f13070993

APA Style

Tiansawat, P., Elliott, S. D., & Wangpakapattanawong, P. (2022). Climate Niche Modelling for Mapping Potential Distributions of Four Framework Tree Species: Implications for Planning Forest Restoration in Tropical and Subtropical Asia. Forests, 13(7), 993. https://doi.org/10.3390/f13070993

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop