Greenness trends and carbon stocks of mangroves across Mexico

Mangroves cover less than 0.1% of Earth’s surface, store large amounts of carbon per unit area, but are threatened by global environmental change. The capacity of mangroves productivity could be characterized by their canopy greenness, but this property has not been systematically tested across gradients of mangrove forests and national scales. Here, we analyzed time series of Normalized Difference Vegetation Index (NDVI), mean air temperature and total precipitation between 2001 and 2015 (14 years) to quantify greenness and climate variability trends for mangroves not directly influenced by land use/land cover change across Mexico. Between 2001 and 2015 persistent mangrove forests covered 432 800 ha, representing 57% of the total current mangrove area for Mexico. We found a temporal greenness increase between 0.003[0.001–0.004] and 0.004[0.002–0.005] yr−1 (NDVI values ± 95%CI) for mangroves located over the Gulf of California and the Pacific Coast, with many mangrove areas dominated by Avicennia germinans. Mangroves developed along the Gulf of Mexico and Caribbean Sea did not show significant greenness trends, but site-specific areas showed significant negative greenness trends. Mangroves with surface water input have above ground carbon stocks (AGC) between 37.7 and 221.9 Mg C ha−1 and soil organic carbon density at 30 cm depth (SOCD) between 92.4 and 127.3 Mg C ha−1. Mangroves with groundwater water input have AGC of 12.7 Mg C ha−1 and SOCD of 219 Mg C ha−1. Greenness and climate variability trends could not explain the spatial variability in carbon stocks for most mangrove forests across Mexico. Site-specific characteristics, including mangrove species dominance could have a major influence on greenness trends. Our findings provide a baseline for national-level monitoring programs, carbon accounting models, and insights for greenness trends that could be tested around the world.


Introduction
Mangroves cover 0.1% of the Earth's surface (Atwood et al 2017, Hamilton andFriess 2018), are within the most productive ecosystems of the world (average 1023 Mg of Carbon ha −1 ; Donato et al 2011), and are highly vulnerable to global environmental change (Alongi 2015, Osland et al 2016. Previous studies have focused on the vulnerability of mangrove forests to sea level rise (SLR) and is expected that this environmental change will influence variables dependent to hydroperiods and salinity, that ultimately impact mangrove forest structure and function (Woodroffe 1990, Mckee et al 2007, Lovelock et al 2015. Recently, there has been increasing attention on how climate variability could impact latitudinal migration of mangroves and their interactions with other coastal plant communities (Cavanaugh et al 2014, Saintilan et al 2014, Osland et al 2017. Climate variability, such as changes in air temperature and precipitation, influence the temporal and spatial patterns of ecosystems processes (e.g. photosynthesis, respiration, evapotranspiration) that control rates of mangroves growth and their geographical distribution (Twilley et al 1999, Alongi 2015, Ward et al 2016, Cavanaugh et al 2018. Due to the relevance of mangrove forests for the global carbon cycle, it is critical to quantify the effects of climate variability on their ecosystem processes and carbon stocks around the world (Atwood et al 2017).
Greenness of vegetation around the world has been used as proxy for photosynthesis activity (Myneni et al 1997, Zhou et al 2014, net primary productivity (Park et al 2016) and water use efficiency (Zhang et al 2015). Studies at the global scale have shown positive greenness trends as consequence of CO 2 fertilization, nitrogen deposition, water availability, and climate variability (Forkel et al 2015, Zhu et al 2016. At the regional scale, studies have focused on quantifying greenness trends across the Northern hemisphere (Park et al 2016), tropical rain forests (i.e. Amazonas and Congo; Hilker et al 2014, Zhou et al 2014, Guan et al 2015 and grasslands (Trujillo et al 2012). Recently, the sensibility and range of mangroves distribution in the Americas were explained by greenness trends (Cavanaugh et al 2018). Other greenness studies on mangrove forests have been site-specific and they have identified land cover changes (Rahman et al 2013), mangrove degradation (Alatorre et al 2016, Ishtiaque et al 2016, Flores-Cárdenas et al 2017, disturbances by chilling events (Zhang et al 2016) and biomass change (Fuller and Wang 2014). Despite these efforts, there is still lacking information at country-specific scale to better understand the influence of climate variability on greenness trends of mangroves and to provide insights for changes in phenology, productivity, and ultimately carbon stocks.
Studies at global and continental scales have used coarse resolution data to predict potential climate variability effects over distribution, biodiversity (∼50 km grids; Osland et al 2017), and spatial variability of above ground biomass in mangroves (∼1 km grids; Hutchison et al 2014, Rovai et al 2016. Synthesis studies at the global scale have highlighted the role of mangrove forests for the Earth-system and the global carbon budget (Giri et al 2011, Hutchison et al 2014, Hamilton and Casey 2016, Ward et al 2016, Atwood et al 2017, Hamilton and Friess 2018, but countryspecific information is needed to refine continentalto-global estimates and identify region-specific trends. Recent examples at the regional-scale and countrylevel include studies in the Indo Pacific Region (Donato et al 2011, Murdiyarso et al 2015, Richards and Friess 2016, Kenya (Gress et al 2017) and the United States (Hinson et al 2017). Despite these efforts it is important to provide and synthesize new information across other countries; especially, for those that have high country-specific carbon stocks and country-specific mangrove area such as Mexico (Atwood et al 2017, Hamilton andFriess 2018).
Our overarching goals were to quantify (a) greenness trends and their spatial variability, and their relationships with mean air temperature and total precipitation; and (b) the overall relationships between greenness and carbon stocks (i.e. AGC and SOCD respectively) in mangrove forests across Mexico. Mexico is a megadiverse country and a global hotspot for conservation priorities (Myers et al 2000) with high potential for implementation of REDD +initiatives (Vargas et al 2017). It has 755 555 ha of mangroves (Valderrama-Landeros et al 2017) where trees can be taller than 35 m or smaller than 1 m as a consequence of gradients of local biophysical factors (i.e. hydroperiod, topography, salinity; López-Portillo and Ezcurra 2002, Ezcurra et al 2016. Furthermore, mangrove forests in Mexico have high countryspecific carbon stocks and country-specific mangrove area, placing them as the top sixth of the world (Atwood et al 2017, Hamilton andFriess 2018). We based our analyses on unique country-specific mangrove cover maps between years 2001 and 2015 and we defined categories of mangroves considering their latitudinal distribution, mean air temperature, main input of freshwater and their location along the coast (i.e. Gulf of California, Pacific Coast, Gulf of Mexico, Caribbean Sea).
Here we asked three interrelated questions. (b) If there are greenness trends, how do they relate to air temperature or precipitation trends (i.e. climate variability)? We expected that positive greenness trends may be related to increases in air temperature (as main factor regulating canopy phenology and photosynthesis; Alongi 2015). Furthermore, changes in local precipitation may not explain greenness trends, as inland precipitation events (and consequently lateral freshwater inputs) could reduce salinity stress and increase local canopy photosynthesis.
(c) Are carbon stocks (i.e. AGC and SOCD) related to trends of mangrove greenness or climate variables during the study period? We hypothesize that mangrove areas with significant greenness increase could have higher primary production and consequently have larger AGC and SOCD.

Study area
Mangrove forest across Mexico are distributed between latitudes 27°50′ to 14°30 N. The main mangrove species are: Rhizophora mangle L. (red mangrove), Avicennia germinans (L) L. (black mangrove), Laguncularia racemosa (L.) C.F. Gaertn (white mangrove) and Conocarpus erectus L. (buttonwood mangrove). We defined four categories of mangroves considering: latitudinal distribution, mean air temperature, main input of freshwater (i.e. by rivers, with surface water as main input or groundwater as main input and due to karst topography), and their location throughout the coast (i.e. Gulf of California, Pacific Coast, Gulf of Mexico, Caribbean Sea). Defining mangrove categories across Mexico is relevant because the country has a large diversity of geomorphological features along its coast (Lankford 1977). Consequently, these characteristics influence local hydrology and define quality and quantity of freshwater, nutrients and sediments that influence mangroves' development and productivity (Twilley et al 1999). We considered as reference the categories developed by Lankford (1977)  Arid category includes all mangroves above the Tropic of Cancer (23°26′ N) and mangroves in the Baja California Peninsula developed over six physiographic regions (see table 1S). These mangroves are usually located within narrow watersheds. Northern areas have small rivers and poor drainage, while southern areas have larger rivers with seasonal drainage. This category shows abrupt changes between mountain areas and coastal plains. Some areas are open and exposed to intermediate to high wave energy, with higher ebb velocity and predominantly semidiurnal tide. Surface water input is the main source of freshwater (Lankford 1977). Annual surface water discharge for this category is 20 838 hm 3 yr −1 ; representing nearly 7% of the total surface water discharge in Mexico (CONAGUA 2016; figure 1).
HUsw-Pa category includes all mangroves below the Tropic of Cancer (23°N Latitude) until the political limits of Mexico with Guatemala, along the Pacific Coast. This category includes mangroves over 6 physiographic regions along the central and south Pacific Coast (table 1S). Those mangroves are in narrow watersheds with abrupt changes between the land and coastal plains. This category has many rivers with small river basin and seasonal flow. All the coastal areas are open and exposed to high wave energy and high ebb velocity (Lankford 1977), where surface water input is the main source of freshwater. Annual surface water discharge for this category is 58 617 hm 3 yr −1 ; representing nearly 19% of the total surface water discharge in Mexico (CONAGUA 2016; figure 1).
HUsw-Gf category includes all mangroves in the Gulf of Mexico, from the political limits of Mexico with United States until the last river founded in the Yucatan Peninsula (close to Campeche City). This category includes mangroves over 6 physiographic regions (table 1S). Those mangroves are in wide watersheds with many rivers with large drainage basins. The wave and tide energies are low, except during hurricanes or northern climatological events. Tides are predominantly diurnal, but rivers are the main source of freshwater (Lankford 1977). Annual surface water discharge is 224 032 hm 3 yr −1 , representing nearly 72% of the total surface water discharge in Mexico (CON-AGUA 2016; figure 1).
Humid mangroves with groundwater as main input of water (HUgw). This category includes mangroves of Yucatan Peninsula (YP) in two physiographic regions (table S1 is available online at stacks. iop.org/ERL/14/075010/mmedia). Those mangroves develop over carbonate platforms, where groundwater is the main source of freshwater. The wave and tide energies are low, except during hurricanes, northern climatological events, or areas on shelf margin reefs. Tides in this category can be diurnal or semidiurnal (Lankford 1977). Conservative estimates of groundwater outflow to the coastal areas are ∼211 462 hm 3 yr −1 (Null et al 2014); this flow of freshwater is nearly similar to 70% of the total surface water discharge in the rest of the coastal areas in Mexico.

Persistent mangrove forest (PMF) coverage across Mexico
We aimed to quantify greenness trends for mangroves without a direct influence of land use/land cover change (LULCC) during the study period. Therefore, we identified PMF across Mexico between 2001 and 2015. Those mangrove areas were estimated from available cartographic sources for 2000 (Giri et al 2011(Giri et al ), 2005(Giri et al , 2010 and 2015 from the Mexican Mangrove Monitoring System (Valderrama-Landeros et al 2017). We used fraction analysis to standardize all sources (i.e. maps) at 1 km of spatial resolution (S section 1). The PMF coverage was used to extract the Normalized Difference Vegetation Index (NDVI), climate variables (i.e. air temperature and precipitation) and carbon stocks (i.e. AGC and SOCD) for each mangrove category.

Detection of greenness of mangroves
We used monthly composites of NDVI from 2001 to 2015 at 1 km of spatial resolution of the MOD13A3 product from the Moderate Resolution Imaging Spectroradiometer (MODIS). Briefly, NDVI is the ratio between red and infrared wavelengths where values close to 1 represent higher greenness in vegetation, while values close to zero are degraded vegetation We used 180 composites for each of the 9 tiles covering Mexico (h07v05, h07v06, h07v07, h08v05, h08v06, h08v07, h09v05, h09v06, j09v07; https:// reverb.echo.nasa.gov/). These composites were resampled and mosaicked using the MODIS Reprojection Tool (https://lpdaac.usgs.gov/tools). We selected the best quality and reliability for the NDVI composites using the Time-series Generator Software (TiSeG; Colditz et al 2008) and applied a linear interpolation for gap filling (see S section 2). Finally, NDVI time series were independently analyzed for each mangrove category.

Climate variables
We used mean monthly air temperature (°C) and total monthly precipitation (mm) from Daymet at 1 km of spatial resolution from 2001 to 2015 (Thornton et al 2017). Mean air temperature was estimated as the average of maximum and minimum air temperature for each pixel (thereafter referred as temperature). Temperature and precipitation data were projected at the same spatial features within persistent mangrove coverage and corresponding mangrove categories.

Above ground carbon and soil organic carbon density
We used country-specific information and extracted values for AGC (Cartus et al 2014). For SOCD, we used a product at 30 cm of depth (Guevara et al 2017) because we needed a comparable soil depth for all mangroves categories, and this is a recommended depth to compare different mangroves areas of the country (Adame et al 2013a, Ezcurra et al 2016). Both carbon products were one-time maps (i.e. AGC from 2007; SOCD from 1991-2010). Products were standardized per mangrove category according to results from a synthesis study (Herrera-Silveira et al 2016; see S section 2), and reprojected and resampled for the same spatial features within PMF coverage and corresponding categories.

Statistical analyses and annual mangrove greenness and climate variability
We used average monthly values during the study period to identify the annual mangrove greenness and climate variability for every mangrove category. We showed these results considering the greenness seasonality of mangroves. After that, we performed crosscorrelation analysis to identify time-lags between the annual mangrove greenness peaks and climate variability peaks. We used the identified time-lags to adjust the annual greenness values with the climate variability . This category has narrow watersheds, northern areas have small rivers and poor drainage. Southern areas larger rivers with seasonal drainage. Annual river discharge of 20 838 hm 3 yr −1 , (∼7% of the total river discharge in Mexico). (b) Humid Mangroves with Surface Water Input along the Pacific Coast (HUsw-Pa). This category has narrow watersheds with abrupt changes between the land and the coastal plains. Annual river discharge of 58 617 hm 3 yr −1 (∼19% of the total river discharge in Mexico); (c) Humid Mangroves with Surface Water Input along the coast of the Gulf of Mexico (HUsw-Gf). This category has wide watersheds with many rivers with large drainage basin. Annual river discharge of 224 032 hm 3 yr −1 (∼72% of the total river discharge in Mexico); (d) Humid Mangroves with Groundwater Input over the Gulf of Mexico and Caribbean Sea (HUgw). Mangroves in this category are over carbonate bedrocks, where groundwater is the main source of freshwater.
values. We did regression models between 'adjusted' greenness and climate variables to identify the main climate variable that controls the annual greenness for every mangrove category.

Greenness and climate variability trends and their spatial variability
We used monthly mean values of NDVI, temperature and precipitation to identify trends of greenness and climate variability. Trend detection analyses were performed using the non-parametric analysis Theil-Sen Regression, with 95% confidence intervals and deseasonalized data in 'openair' R package. Theil-Sen regression uses medians to calculate slopes between n − 1 point in the time series. Its robustness is based on bootstrap simulations to derive p-values, slope estimates and uncertainties (Wilcox 2004, Carslaw andRopkins 2012). We show results of aggregated trends for greenness and climate variability per mangrove category.
We used the Annual Aggregated Time Series (AATS) method to identify the spatial variability of greenness trends (SVGT) and climate variables in 'greenbrown' R package, with 95% confidence interval. AATS aggregates seasonal time series values to annual values, and uses the sum of linear square residuals to estimate the breakpoints and the slopes in the time series (Bai andPerron 2003, Forkel et al 2013). AATS has a good performance for NDVI time series affected by seasonality, and is a conservative method for potential false positive or false negative trends (Forkel et al 2013). We report the SVGT per mangrove category. Spatial variability of climate variables trends are not shown, but we used them to identify linear relationships between SVGT and climate variables with carbon stocks.
We summarized the distribution of SVGT with histograms representing the percentage of significant greenness trends (GTP) for every category. We calculated the GTP by dividing the values of SVGT by the mean value of NDVI in each category.
Annual mangrove greenness and climate variability Overall, lower greenness values were evident from April to May (range 0.53-0.72), while greenness peaked between October and January (range 0.67-0.80) ( figure 2(a)). Temperature showed the higher values from April to October ( figure 2(b)), while precipitation from July to October (figure 2(c)). We observed time-lags from two to five months between the highest values of temperature and precipitation with the highest values of greenness. In all cases, temperature and precipitation peaked before greenness (figures 3(a) and (b)). For ARsw, we found that temperature peaked three months before maximum greenness, while maximum precipitation peaked two months before maximum greenness. In contrast, for HUgw temperature peaked five months before than maximum greenness, while maximum precipitation peaked four months before maximum greenness (figures 3(a) and (b)). Overall, temperature was able to better represent the variability of greenness (ARsw r 2 =0.72; HUsw-Pa and HUsw-Gf r 2 =0.74; HUgw r 2 =0.69; in all cases p < 0.01; figure 3(a)), than precipitation across categories (ARsw r 2 =0.30; HUsw-Pa r 2 =0.40; HUsw-Gf r 2 =0.44; HUgw r 2 =0.41; in all cases p < 0.01; figure 3(b)).

Greenness and climate variability trends and their spatial variability
Greenness trends had a significant increase for the mangrove categories developed over the Gulf of California and Pacific Coast. These categories are ARsw and HUsw-Pa (ARsw, p<0.001, figure 4(a); HUsw-Pa, p<0.001; figure 4(b)), with rates from 0.004 in ARsw to 0.003 in HUsw-Pa. We did not find significant greenness trends for the HUsw-Gf and HUgw developed over the Gulf of Mexico and Caribbean Sea (figures 4(c) and (d), respectively). However, these two categories were the only ones with a significant increase in temperature (figures 4(g) and (h)), with a range from 0.10°C to 0.11°C. ARsw was the only category with a significant decrease in precipitation (0.35 mm yr −1 ; p<0.001; figure 4(i)).
SVGTs showed that PMF areas over the Gulf of California and Pacific Coast (ARsw and HUsw-Pa) have a significant greenness increase (i.e. positive trend) during the study period, while areas in the Gulf of Mexico and Caribbean Sea (HUsw-Gf and HUgw) showed a mix of significant increase and decrease
We did not find significant relationships between the SVGT, climate variability and carbon stocks for most categories. HUsw-Gf was the only one with a negative relationship between negative greenness trends and AGC (r 2 =0.38, p > 0.01); which was mainly observed for the Pom Atasta Lagoon ( figure 5(h)). HUsw-Pa had a positive relationship between AGC and SOCD (r 2 =0.54, p>0.01).

Discussion
Greenness and climate variability trends and their spatial variability We found positive greenness trends for mangroves categories developed over the Gulf of California and Pacific Coast (ARsw and HUsw-Pa). Mangroves over the Gulf of Mexico and Caribbean Sea (HUsw-Gf and HUgw) did not show significant greenness trends; however, SVGTs were site-specific. Trends of climate variability were not consistent with greenness trends. These results suggest that the dominance of mangrove species could be one of the main factors on greenness trends, together with site-specific environmental factors such as quantity/quality of freshwater input, ocean water exchange, extreme events (i.e. frequency and magnitude of tropical storms and hurricanes) and human-induced changes (i.e. preferential flow paths of water).
ARsw and HUsw-Pa receive less than 30% of the total annual river discharge in Mexico (79 455 hm 3 year; CONAGUA 2016). Fifty-two percent of the ARsw rivers are dammed, and those dams could store ∼84% of the total annual river discharge (∼17 500 hm 3 yr −1 ; CONAGUA 2016). For HUsw-Pa, 28% of the rivers are dammed and dams can store ∼26% of the total annual river discharge (15 240 hm 3 yr −1 , CON-AGUA 2016). These two categories are open to landocean exchanges influenced by medium to high wave energy (Lankford 1977). The combination of these factors could increase salt-water intake on mangrove soils and enhance the dominance of Avicennia germinans. This is the most tolerant mangrove species to higher salinities and extreme temperatures in the Americas . Positive greenness trends on these categories could relate with physiological adaptations of A. germinans. It is known that Avicennia spp. can increase water retention and leaf thickness when salinity and aridity increase (Nguyen 2017). Therefore, this could have direct implications for the canopy reflectance and NDVI values. First, the infrared (IR) wavelength is sensitive to the water and air retained in the mesophyll structure of the leaves (Peñuelas and Filella 1998); consequently, higher water content in leaves (i.e. water retention) results in an increase of IR reflectance and higher greenness values. Second, A. germinans modifies leaf angles according to the sun conditions to reduce leaf temperature (i.e. heliotropism; Krauss et al 2008); consequently, this modification can increase overall canopy reflectance. Greenness trends could relate with physiological adaptations of dominant mangrove species, as well their canopy interaction at different wavelengths. Resilience of mangrove species to environmental conditions and human impacts could also enhance greenness trends. A. germinans is one of the most resilient species to recent climate variability with the ability to expand its global distribution to the Poles and to encroach over temperate coastal saltmarshes (Cavanaugh et al 2014, Madrid et al 2014, Saintilan et al 2014, Krauss et al 2014a. Its expansion should relate with potential greenness increase, as consequence of trees recruitment and canopy growth. Human-induced changes could drive differential impacts on the mangrove community that eventually may enhance greenness trends. For example, Marismas Nacionales part of HUsw-Pa category (figure 5(e)), has experienced a differential impact on its mangrove community, losing nearly 8000 ha of mangrove in the last 30 years (mainly Laguncularia racemosa; Kovacs et al 2001, Valderrama et al 2014 and potentially enhancing the presence and development of A. germinans. Those impacts are a consequence of human-induced preferential flow paths of water from the ocean to the land (i.e. Cuatlá Channel opened in the 1970s with an average opening of ∼50 m, but currently the opening has ∼800 m; Google Earth). This preferential flow path increased the salinity on the mangrove soils and could enhanced the differential impacts on the mangrove community (Kovacs et al 2001). Our results showed that positive greenness trends in mangroves over the Gulf of California and Pacific Coast may relate with mangrove species composition and differential impacts of sitespecific characteristics and human-induced changes.
HUsw-Gf and HUgw did not show significant greenness trends. These categories receive the major amount of freshwater in the country (∼200 000 hm 3 yr −1 per category); where HUsw-Gf is influenced by rivers and HUgw by groundwater. Forty-four percent of the HUsw-Gf rivers are dammed with a capacity to store ∼13% of the total annual river discharge (∼29 500 hm 3 yr −1 ; CONAGUA 2016). There are not dams in HUgw but conservative estimates of groundwater extraction are ∼2400 hm 3 yr −1 ; representing nearly ∼1% of the potential water outflow to the coastal zone (INEGI 2010, Null et al 2014. Sediments and nutrient inputs are the main difference between both categories. HUsw-Gf have the major river discharge across Mexico and likely increase the inputs of sediments, nutrients and particulate matter  environmental and anthropogenic disturbances (e.g. hurricanes, SLR, reduction of freshwater, pollution in sediments or water, etc) that could influence a decrease in forest grow and productivity.
Hurricanes affect the canopy of mangroves (Doyle et al 1995, Adame et al 2013b, Zhang et al 2016. During the study period there were 17 tropical storms and hurricanes (TS-H) affected HUgw mangroves (24% of them were major, category 3 or more in Saffir-Simpson scale). HUsw-Gf mangroves were impacted by 24 TS-H, with no major hurricanes. There were 23 TS-H for ARsw (less than 5% were major) and 17 for HUsw-Pa (12% were major; NOAA 2018a). All categories, except HUsw-Pa, had an increase> 4% on the TS-Hs frequency between 2001 and 2015, compared to the period of 1980-2000. HUgw had the higher frequency increase (9%), while the HUsw-Pa the lower decrease (∼18%). Therefore, we propose that hurricane impacts could cancel out any potential greenness increase over the Gulf of Mexico and Caribbean Sea, mainly via the negative effect of canopy defoliation during storm events (Doyle et al 1995, Zhang et al 2016.
We highlight that other factors could also influence greenness trends. For example, SLR and its effect on phosphorus (P) availability may have a differential effect on the mangroves productivity and carbon stocks based on site-specific characteristics  in Mexico have experienced an average sea level rise >2.34 mm yr −1 (NOAA 2018b) that may enhance soils fertility by increasing P availability (Krauss et al 2006, Castañeda-Moya et al 2011 and intensify the positive greenness trends detected for ARsw and HUsw-Pa. However, other mangrove areas (e.g. HUgw) could experience canceling effects by SLR and P availability, and could explain the slow root and leaf turnover rates experienced by those mangroves (Castañeda-Moya et al 2011, 2013. We conclude that SLR and P availability could enhance grow and be reflected on greenness, but the timing of the response will depend on site-specific characteristics.

Annual variability of greenness and climate drivers and their relationship
Our results show that mangroves across Mexico have a seasonal cycle with higher greenness during winter months, as seen across North America (Zhang et al 2016) and at the local scale in mangroves of Mexico (Pastor-Guzman et al 2015, Flores-Cárdenas et al 2017). Furthermore, we found that annual greenness peaks have a time-lag of two to five months after the peak of temperature and precipitation, respectively. This time-lag may be related to higher stress conditions for mangroves during warmer months (Chen and Ye 2014). Mangroves are tropical and subtropical vegetation stressed by constant water salinity inputs, this could limit the photosynthesis activity during warmer months that may delay the greenness increase. The time-lag effect of precipitation may be related to the time that water requires to move from high watershed areas to the coast, and the time involved in dissolving nutrients before they are available to mangroves. We propose that global-scale studies of greenness ( consider that greenness patterns of mangroves may be decoupled from patterns of terrestrial ecosystems as they can be influenced by regional (i.e. teleconnections) or local factors inherent to coastal wetlands.

Carbon stocks and greenness
Our analyses show that the spatial variability of greenness and climate trends are not related to carbon stocks in almost all mangroves of Mexico. AGC and SOCD did not show significant spatial relationships for almost all categories. Globally, SOCD patterns in mangroves do not overlap with AGC patterns (Atwood et al 2017), likely as a consequence of site-specific factors for the origin of soil organic carbon (Ezcurra et al 2016). HUsw-Gf mangroves have more than 40% of AGC than other categories, likely because HUsw-Gf represent the most developed and structured mangroves across Mexico (Day et al 1987, Kauffman et al 2016. HUgw showed the highest amount of SOCD with 58% more than the other categories, possible because this category is less influenced by spring water pulses (i.e. large river discharge). We postulate that in HUgw the surface SOCD could have lower lateral transport rates to the coastal ocean than in the other categories, and consequently it could have a longer residence time at the 0-30 cm depth. Many mangrove areas in this category are P limited that could increase the SOCD  increase in the future, but some other areas could experience a decrease (e.g. Gulf of Mexico and Caribbean Sea). Consequently, there is a need to develop reference frameworks for long-term monitoring projects of carbon stocks in mangroves of Mexico and implementation of REDD+initiatives (Vargas et al 2017).

Conclusion
Our results quantified greenness trends and their spatial variability across PMF without direct influence of LULCC. Our results showed a greenness increase for mangroves developed over the Gulf of California and Pacific Coast, mainly dominated by A. germinans. In contrast, site-specific biophysical factors could influence the response of mangroves across the Gulf of Mexico and Caribbean Sea. Overall, greenness trends were not consistently influenced by trends in temperature or precipitation. We propose that the combination of environmental factors such as quantity/quality of freshwater input, storms, anthropogenic influence, and site-specific characteristics could have more influence on greenness trends than climate variability alone. Finally, greenness and climate variability trends are not directly related to carbon stocks for most mangrove areas of Mexico. Our findings provide a baseline to develop regional monitoring programs, carbon accounting models, and could be tested across other mangrove forests around the world.