Effect of Temperature and Precipitation on Global Mangrove Rhizophora Species Distribution

Corresponding Author: Aroloye O. Numbere, Department of Biology, Saint Louis University, St Louis, Missouri, 63101, USA Email: aroloyen@yahoo.com Abstract: Climate change influences the distribution of mangroves towards tropical or temperate regions of the world. We hypothesized that Rhizophora species organized along complex spatial and temporal temperature and precipitation gradient. To determine the affinity of different mangroves with climatic parameters, the response of eight Rhizophora species to seven climatic parameters (i.e., annual mean temperature, annual precipitation, maximum temperature of warmest month, minimum temperature of coldest month, temperature annual range and temperature of driest month and precipitation of warmest month) was analyzed with Arc GIS and Maxent software. For temperature and precipitation there was significant difference (p<0.001) among species location. Only R. selala (p<0.001) differed most from the rest taxa in temperature. The jackknife test in Maxent revealed that minimum temperature of coldest month (regularized training gain ~0.71), temperature annual range, (regularized training gain ~0.66) and annual mean temperature (regularized training gain ~0.36) were the most important environmental parameters that determine the occurrence of R. mangle, being the most dominant species. This outcome was further validated by an accuracy of 91%. This implies thattemperature was a better predictive parameter than precipitation in the distribution of R. mangle.


Introduction
Mangroves occupy more than 150, 000 km 2 , occur in over 123 countries and contain more than 73 species and/or hybrids (Bunt, 1992;Spalding, 1997;Spalding et al., 2010). Mangroves are divided into the Indo-West Pacific (IWP) and the Atlantic East Pacific (AEP) groups (Duke, 1990;Macnae, 1968). They originated from a warm environment (Plaziat et al., 2001) and their distribution is influenced by meteorological events (Alongi, 2008) such as temperature (Duke, 1992) and precipitation (Saenger and Snedaker, 1993). These climatic parameters influence influence their habitat specificity (Feller et al., 2010). Although, tolerance to warm conditions dictates their distribution, they sometimes drift to temperate regions where inhospitable weather threatens their existence (Hogarth, 1999). Global warming causes mangroves to the spread beyond their latitudinal limit (Ellison, 2000;Field, 1995). For example, New Zealand, southern Australia, southern Brazil and South Africa are countries that have temperate weather, but have mangroves. Tropical conditions are the best for mangroves, but excessive heat increase the rate of evaporation leading to increase in salinity (Gilman et al., 2008). This situation triggers the succession of salt tolerant mangrove species (e.g. Avicennia germinans) over low salt-tolerant species (e.g. Rhizophora) (Lugo, 1980). Increase in temperature affects hydrology, evapo-transpiration and water runoff (Dai et al., 2009;Solomon, 2007). Temperature greater than 35°C affects root structure, seedling establishment and photosynthesis in mangroves (Chakraborty, 2013). Continuous rise in temperature can lead to modification, increase in diversity and migration of species into subtropical salt marsh areas (Perry and Mendelssohn, 2009) and Arctic pole (Cavanaugh et al., 2014).
Precipitation regulates nutrient up-take and affects productivity (Snedaker, 1995) and survival (Tomlinson, 1998) of mangroves. Moderately warm and wet equatorial areas with high rainfall have rich supply of mangrove population (Record et al., 2013). However, increase in sea level (Lyu et al., 2014) generates excess water, which drowns mangroves (Gilman et al., 2008).
Similarly, extreme events like global cooling and warming (IPCC, 2014;Scherer and Diffenbaugh, 2014) lead to range shifts and the extinction of organisms (Hewitt, 2004a;Yokoyama et al., 2001). Mangrove propagules are dispersed by tidal currents, but land barriers prevent their movements (Duke, 1990) leading to a discontinuous distribution. This discontinuity causes intra-specific, morphologic and genetic variation in Rhizophora species (Dodd et al., 1998).
Rhizophora species were used for this study because they were have global dominance as compared to other mangrove species (Rabinowitz, 1978). They are found in latitude 28ºN and 28ºS in the AEP. They are halophilic and viviparous in nature, which give them the advantage of faster growth and wider distribution outside their geographic range. Rhizophora mangle is the most widespread of all Rhizophora species. They are also one of the species that have extended beyond the tropics into the temperate region (Afzal-Rafii et al., 1999). The present study examines the key climatic parameters determining the global distribution pattern of Rhizophora species. The objective of this study is to determine the climatic and environmental variables that have higher distribution of R. mangle species around the globe.

Data Collection and Analysis
Present-only data of mangrove Rhizophora species from 41 countries were retrieved from Global Biodiversity and Information Facility (GBIF; http://www.gbif) and Tropicos of the Missouri Botanical Garden (www.tropicos.org). Similarly, climatic factors (i.e. environmental data) which were major predictors of species were downloaded from BIOClIM (http://www.bioclim.org), a bioclimatic analysis and prediction system which uses surrogate terms derived from mean monthly climatic estimates, to approximate energy and water balance at a given location (Nix, 1986). BIOCLIM is a monthly temperature and rainfall values from the past to the current conditions between 1950 and 2000 prepared to generate more biologically meaningful variables (Hijmans et al., 2005). The climatic parameters downloaded (Table 1) are all associated with mangrove distribution patterns (Alongi, 2008;Duke et al., 1998;Gilman et al., 2007).

Data Processing and GIS Software
ArcGIS 9.1 (ESRI, 2006) was used to create shape files for the presence data in a species distribution map ( Fig. 1). Countries shape file was exported into Arc GIS with the "Add XY Data" command. The file was converted to raster with the "Hawths Analysis Tool" in extensions. The temperature and precipitation data ( Table 2) were then extracted from the shape file already created and the average and standard error calculated and analyzed in R environment, version 2.9.0 (R Core Development Team, 2008).

Maxent (Maximum Entropy Method)
Binary probability distribution in Maxent software version 3.1.0 (Phillips et al., 2006) was used to determine the most important environmental data. Maxent uses background data based on maximum entropy, which is the cumulative probability of species occurrence (Phillips et al., 2004). It uses the principle of environmental niche modeling, where the niche model was divided into a sample and an environment part, which are jointly used to predict species occurrence. Maxent is similar to the Gibbs probability distribution (qλ) equation and both maximize the product of the probabilities of the sample locations (Phillips et al., 2004;. Maxent uses the jackknife test to assess environmental variables that are important to species ecological niche and distribution (Phillips et al., 2004). Jackknife test gives the quality of the models and checks the suitability of the environment for each species. The probability of occurrence of species in Maxent is divided into two in-built suitability ranges of 0-49 (unsuitable) and 50-100 (good). Maxent also uses the Receiver Operating Characteristics (ROC) curve to calculate the Area Under the Curve (AUC). The AUC measures the speed with which the model encompasses the testing points and varies between 0 and 1, where 0 is the least and 1 the maximum obtainable point. Thus, the higher the training gains, the better the model and the lower the training gain the poorer the model.  (14)), that are in-built in Maxent were used to graphically present the most significant variable that determine the distribution of the species investigated.

Statistical Analysis and Model Assessment
The ROC and the statistical analysis were performed to validate the data and to test the significance of the variables investigated respectively. The ROC and AUC were not only used to assess the suitability of species occurrence, but were also used as response curve to determine the variables (i.e. temperature and precipitation) that had the highest impact on the distribution of Rhizophora species globally. The classifier for the species was plotted as "sensitivity versus 1-specificity". Sensitivity and specificity are defined as the proportion of correctly predicting presence and absence of species (Wilson et al., 2005). Swets (1988) classified AUC range as follows: 0.90-1.00(excellent); 0.80-0.90 (good); 0.70-0.80 (fair); 0.60-0.70 (poor); and < 0.60 (fail).
Pairwise comparisons using t-tests with pooled standard deviation were conducted for temperature and precipitation. A Tukeys HSD test was also done to determine the specific differences. A one-way ANOVA test was also performed to determine the level of significance among species using the residuals. All tests were done in R statistical environment (R Development Core Team, 2008). Normality, homoscedasticity, Shapiro-Wilk and Bartlett tests were performed to ensure that the data were normal and the variances equal (Logan, 2010).

Results
The result indicates that there was a significant difference in temperature (F 7,1,700 , = 8.8, P<0.001) and precipitation (F 7,1,700 = 19.2, P<0.0001) amongst the Rhizophora species. To determine specific differences, results from the pairwise t-test (Holm, 1979) indicated that R. selala was the most significantly different taxa (P<0.001). The Tukeys HSD further shows that R. selala was the most significantly different species in terms of temperature and precipitation.
The mean temperature tolerance range of species was 23-26°C, while the mean precipitation tolerance range was 1100-2600 mm in the areas sampled. Rhizophora species that occur in the wettest areas (i.e., precipitation > 2000 mm) are R. racemosa, R. × apiculata and R. × harrisonii. In the same vein, Rhizophora species that occur in the driest areas are R × selala, R × Lamarckii and R. × stylosa. This result indicates that the hybrid species have more affinity for low precipitation. Rhizophora. × harrisonii and R × lamarckii had more affinity for higher temperature (i.e. 26°C), whereas R × selala had affinity for the lowest temperature (23°C) in this study. Because of the low sample size of other Rhizophora species (i.e. < 10 %, Table 1), R. mangle (Fig. 1), which had over 70% occurrence and more training stations (Table 2) was analyzed separately.
Therefore, Maxent results indicate that the most important climatic variable in the distribution of R. mangle was minimum temperature of coldest month (regularized training gain 0.71). This outcome was based on the jackknife test, which revealed that when this important parameter was removed from the environmental data it reduced the performance of the entire model. Two other important parameters were temperature annual range (regularized training gain 0.66) and annual mean temperature (regularized training gain 0.38) ( Table 2). However, values of maximum temperature of warmest month, annual precipitation, precipitation of warmest month and precipitation of driest month were all low (Table 2). This result is further shown in Fig. 2 where the dark blue bars in the legend indicates the contributions of the environmental variables alone while the bluish green bar shows the performance of the model without the environmental variables and the red shows the results when all the variables are put together. The AUC of the training data (red curve) was 0.88, while AUC for the test data (blue curve) was 0.91 (Fig. 3) in line with the prediction of DeLong et al. (1988). Since the percentages of both the training and test data were above 80%, it was assumed that there is a high level of confidence in the prediction of species distribution as influenced by temperature and precipitation.

Discussion
The influence of temperature and precipitation on species distribution (Cavanaugh et al., 2014) indicates that climate change will influence the diversity and global distribution of mangroves (Said, 2003). Therefore the emergence of unprecedented temperature regime is significant towards the survival of organisms and ecosystems (Diffenbaugh and Charland, 2016). This study indicates that temperature and precipitation influence Rhizophora species distribution. Rhizophora species have low tolerance for temperatures below 0°C and above 38°C (Duke and Allen, 2006). Similarly, precipitation below 40 mm is intolerable to Rhizophora species. Nevertheless, precipitations in the warmest and coldest quarters are the highest predictors of mangrove distribution (Record et al., 2013). The most important predictor of the distribution of R. mangle from this study was the minimum temperature of the coldest month. This is in line with other studies that showed that current mangrove distributions are limited by the 16°C isotherm of the coldest month (Ellison, 1994;Gilman et al., 2008). Historically, Rhizophora species originated from a hot environment in the Tethys Sea during the Eocene period (Plaziat et al., 2001). This is why they have affinity for hot tropical climates, but global warming is forcing them to adapt to cooler temperature climes around the world, and move away from their latitudinal limits (Feller et al., 2010). Presently mangroves grow outside their geographic ranges in botanic gardens, nurseries and green houses, where favorable temperatures and precipitations are simulated. Although, temperature was a better predictor of R. mangle distribution than precipitation in this study, the fact remains that precipitation is also important to the growth and development of mangroves (Saenger and Bellan, 1995) as recorded in our study (i.e. 1000 to 2500 mm). Rhizophora species are less halophilic as compared to Avicennia germinans that are more halophilic (Lugo, 1980). During drought low precipitation cause increase in salinity levels (Snedaker, 1995), which leads to the migration of Rhizophora species to areas of lesser salinity (Lugo, 1980). In contrast, increased precipitation leads to increased run-off, which decreases salinity levels in mangrove environment (Gilman et al., 2008). Thus, during sea level rise Rhizophora species may have higher survival than other mangrove species (Pachauri, 2015) given their physiological adaptation and tolerance to flooding (Greenway and Gibbs, 2003;He et al., 2007).
Rhizophora mangle dominance is hinged on their ability to survive diverse climatic and environmental conditions. Since minimum temperature of coldest month influence their distribution, it implies that cold condition will limit their expansion while warm condition will accelerate their expansion into warmer environments. During future climate change mangroves will undergo positive range shift by migrating (Root et al., 2003) to warmer environments. This means species that cannot shift their ranges (Schloss et al., 2012) because of habitat destruction, geographic isolation and low genetic fitness will go into extinction (Parmesan and Yohe, 2003). For instance, many parts of the world such as Eurasia, North America, tropical South America and tropical North Africa are experiencing > 1.0°C warming over the past half century (IPCC, 2013). In addition, global warming is causing some areas to move into a regime where coolest years are warmer than the hottest years (Diffenbaugh and Scherer, 2011). This has resulted to low inter-annual variability in the tropics and has created new temperature regimes (Anderson, 2011;Diffenbaugh and Scherer, 2011). What all these mean for the mangroves is that there will be an explosion in population growth.
The databases used for this study applied the principle of synonymy or name matching (Holm-Nielsen et al., 1988) to retrieve species information (i.e. name, coordinates and references) from websites. Temperature and precipitation were used because they are the key climatic parameters used in ecological niche modeling for forecasting species distribution (Ellison, 2002;Gilman et al., 2008). Some limitations of georeferencing and data collection were observed (Dahdouh-Guebas, 2011;Record et al., 2013), but were corrected (Yesson et al., 2007) before they were used for analysis. In addition, other parameters that might contribute to species distribution (e.g. soil moisture, radiation, humility and salinity) will be considered in future work.

Conclusion
Temperature and precipitation enahnace the survival of mangroves, but temperature had better predictive power than precipitation in mangrove global distribution.. The distribution of Rhizophora mangle, the most dominant Rhizophora species, is mostly influenced by minimum temperature of coldest month, temperature annual range and annual mean temperature. Out of all the Rhizophora species, R.selala may likely have more tolerance to future climate change.

Acknowledgement
Special thanks go Dr. Jason Knouft, Department of Biology of Saint Louis University for his advice on the project. We thank Dr. Eric Westhus for assistance in R statistics and Dr. Life Kaanagbara for reading the manuscript. We thank the Saint Louis University, Missouri for making its GIS lab available for this study. Thanks go to the University of Port Harcourt, Nigeria for granting the lead author a study at the Saint Louis University.

Funding Information
The researchers did not receive any funds from any funding agency for this project. The authors at the Saint Louis University, Missouri GIS laboratory and the field station in Niger River Delta, Nigeria, independently carried out the study.

Author's Contributions
Aroloye O. Numbere: Contributed in experimental design, performed data analysis, interpreted the results and did the manuscript writing.
Gerardo R. Camilo: Developed the hypothesis and initiated the conceptual design, supervised the project and was involved in all aspect of the work.

Ethics
This article is original and contains unpublished material. The corresponding author confirms that all of the other authors have read and approved the manuscript and no ethical issues involved.