Skip to main content
PMC home page
PLOS ONE logoLink to PLOS ONE
. 2021 May 20;16(5):e0252063. doi: 10.1371/journal.pone.0252063

Global tropical dry forest extent and cover: A comparative study of bioclimatic definitions using two climatic data sets

Jonathan Pando Ocón 1,*,#, Thomas Ibanez 2,#, Janet Franklin 3,‡, Stephanie Pau 4,‡, Gunnar Keppel 5,‡, Gonzalo Rivas-Torres 6,7,8, Michael Edward Shin 1, Thomas Welch Gillespie 1,#
Editor: RunGuo Zang9
PMCID: PMC8136719  PMID: 34015004

Abstract

There is a debate concerning the definition and extent of tropical dry forest biome and vegetation type at a global spatial scale. We identify the potential extent of the tropical dry forest biome based on bioclimatic definitions and climatic data sets to improve global estimates of distribution, cover, and change. We compared four bioclimatic definitions of the tropical dry forest biome–Murphy and Lugo, Food and Agriculture Organization (FAO), DryFlor, aridity index–using two climatic data sets: WorldClim and Climatologies at High-resolution for the Earth’s Land Surface Areas (CHELSA). We then compared each of the eight unique combinations of bioclimatic definitions and climatic data sets using 540 field plots identified as tropical dry forest from a literature search and evaluated the accuracy of World Wildlife Fund tropical and subtropical dry broadleaf forest ecoregions. We used the definition and climate data that most closely matched field data to calculate forest cover in 2000 and change from 2001 to 2020. Globally, there was low agreement (< 58%) between bioclimatic definitions and WWF ecoregions and only 40% of field plots fell within these ecoregions. FAO using CHELSA had the highest agreement with field plots (81%) and was not correlated with the biome extent. Using the FAO definition with CHELSA climatic data set, we estimate 4,931,414 km2 of closed canopy (≥ 40% forest cover) tropical dry forest in 2000 and 4,369,695 km2 in 2020 with a gross loss of 561,719 km2 (11.4%) from 2001 to 2020. Tropical dry forest biome extent varies significantly based on bioclimatic definition used, with nearly half of all tropical dry forest vegetation missed when using ecoregion boundaries alone, especially in Africa. Using site-specific field validation, we find that the FAO definition using CHELSA provides an accurate, standard, and repeatable way to assess tropical dry forest cover and change at a global scale.

Introduction

Tropical dry forest has been estimated to comprise 42% of all tropical forests and is believed to be one of the world’s most endangered biomes [1–3]. Tropical dry forests provide the ecosystem services needed to support millions of subsistence farmers in some of the world’s poorest areas, and higher population densities are driving the demand for energy and land leading to higher tropical deforestation rates in dry forest than humid forest [4, 5]. Additionally, tropical dry forests harbor unique and diverse ecological communities and their deforestation contributes to the steady erosion of Earth’s biodiversity [6–8]. To protect this critically endangered and valuable resource, we need reliable estimates on the extent of the tropical dry forest biome and understand the degree of uncertainty around those estimates.

Terrestrial biomes have long been associated with climatic range limits [9, 10]. These foundational early descriptions of global biome patterns are complemented by a growing understanding of the relationship between climate and vegetation form and function [11–14]. Recent advances in global climate data sets are improving our understanding of the extent and distribution of biomes. WorldClim, the most widely used climate dataset in biogeography, provides global, gridded climate data at up to 1 km resolution interpolated from a network of weather stations [15], while the Climatologies at High-resolution for the Earth’s Land Surface Areas (CHELSA) offer 1 km resolution climate data based on a quasi-mechanistically statistical downscaling of global circulation models [16]. Furthermore, global data sets for estimated potential evapotranspiration (PET) can now map aridity indices at 1 km resolution [17, 18].

Bioclimatic definitions are useful for estimating biome extent and forest cover at different spatial scales and to predict the dominant vegetation without the influence of humans or disturbances such as fire [14, 19]. However, bioclimatic definitions cannot differentiate between vegetation types, such as savannas, shrublands, woodlands, and deciduous to evergreen forests, whose climatic limits overlap [14, 20–22]. Thus, climatic definitions allow delimiting the potential extent of biomes, but not detailed mapping of vegetation type boundaries.

There have also been significant advances in spaceborne remote sensing of forest cover [23–26]. Forest cover for the first two decades of the 21st century can be mapped globally based on forest cover change data sets that contain forest cover and percent forest cover at a 30 m resolution from Landsat imagery for 2000 to 2020 [27]. Such global forest cover products can be used to evaluate the extent and change of forest cover within the climatic boundaries of different biomes and forest types, providing crucial information for conservation and land use planning [28, 29].

There has been a growing number of efforts to map forest extent and forest cover change in tropical dry forest regions, especially to identify dry forest regions with a high conservation priority at a global spatial scale [4, 8, 23, 29, 30]. Most biogeographical studies and conservation evaluations of tropical dry forest biomes at a global spatial scale use World Wildlife Fund (WWF) terrestrial ecoregions to establish the potential extent of this biome [3, 23, 27, 31, 32]. Currently, the WWF has defined 53 tropical and subtropical dry broadleaf forest ecoregions.

Previous global analyses of tropical dry forest cover have, however, also shown that the tropical dry forest vegetation type (e.g. canopy height ≥ 3 m, closed canopy ≥ 40%, drought-adapted species) has been misidentified within 500 m pixels as tropical and subtropical grasslands, savanna and shrubland, woodland and scrub; and desert and xeric shrubland ecoregions [23]. Thus, more precise knowledge of the extent and location of tropical dry forest biome is needed in order to track forest cover change and better understand threats (deforestation, fire, climate change) and the conservation status (endangered species, old growth, protected areas) of this tropical forest biome.

The range boundaries of the tropical dry forest biome should coincide with climatic thresholds related to temperature, mean annual rainfall, seasonality, and water deficit [17, 20, 33]. However, there are several bioclimatic definitions of tropical dry forest that can be used to assess the global extent of this biome independently from WWF ecoregion data [1, 8, 17, 30, 34]. Widespread consensus on the bioclimatic parameters defining the tropical dry forest biome extent is yet to be established [35] and no studies compare the predicted global or regional extent of the forest biome based on different definitions of bioclimatic suitability [5].

There are a number of ways to validate and refine bioclimatic assessments of the tropical dry forest biome. Results can be compared to local vegetation maps [33, 36], high resolution (< 1 m) remote sensing imagery [17], and field inventories or georeferenced specimens [8, 33, 36]. Field plots, however, offer the most robust approach to validating bioclimatic models because they provide high resolution data on location, species composition, and structure over a standardize area [37]. Furthermore, field ecologists are often familiar with a site’s natural history and have identified the vegetation and function as a tropical dry forest. Global studies of plant and forest diversity [38] and structure [31] now contain over 100 field plots from tropical dry forests, although certain regions may be underrepresented.

This study has three objectives related to the global extent of the tropical dry forest biome and vegetation type. First, we compare the extent of tropical dry forest predicted by four bioclimatic definitions (Murphy and Lugo, Food and Agriculture Organization, DryFlor, aridity index) for the biome using two different climatic datasets (WorldClim and CHELSA) and compare these to widely-used WWF ecoregions. We would expect that there should be high agreement (e.g. 70%) between individual bioclimatic definitions and WWF ecoregions regardless of climate datasets used. Second, we identify if WWF ecoregions or bioclimatic definitions have the highest agreement with field plots defined as tropical dry forest. We would expect that there is high agreement among WWF ecoregions, bioclimatic definitions, and field plots. Third, we use the best definition to calculate tropical dry forest vegetation cover and change from 2000 to 2020. Global tropical dry forest cover estimates should be similar to other published global estimates.

Materials and methods

Climatic definitions of tropical dry forest

We computed the bioclimate envelope (or the potential extent) of tropical dry forest located between 30˚N and 30˚S using four different definitions (Table 1). For all these definitions, we first subsetted frost-free areas (> 0˚C) using minimum temperature of coldest month with mean annual temperature > 17˚C [1]. We defined tropical dry forest biome according to the four bioclimatic definitions by subsetting areas of relevant climate variables according to the thresholds or ranges (Table 1), and then overlaid (intersected) those subsetted areas to delimit where all climatic criteria co-occur.

Table 1. Overview of bioclimatic definitions (Murphy and Lugo, Food and Agriculture Organization of the United Nations [FAO], DryFlor, Aridity Index) of the tropical dry forest biome.

Bioclimatic Definition Source Annual Precipitation, Dry Season
Murphy and Lugo Murphy and Lugo 1986 250–2000 mm, 4–7 months ≤100 mm
FAO Sunderland et al. 2015 500–1500 mm, 5–8 months ≤100 mm
DryFlor Banda et al. 2016 ≤1800 mm, 3–6 months ≤100 mm
Aridity Index* Bastin et al. 2017 0.2 < aridity index < 0.65
Open in a new tab

* Calculated by taking the ratio of mean annual precipitation to potential evapotranspiration.

Climate data sets

We used two different climatologies to compute the potential extent of tropical dry forest biome, WorldClim [15, https://www.worldclim.org/] and CHELSA [16, http://chelsa-climate.org/]. Both climatologies have a spatial resolution of 30 arc seconds, or ~1 km, but differ in the way they were computed. We used the 2nd version of the WorldClim climatology, released in 2017. This climatology was derived from weather station measurements (1970–2000) interpolated using thin‐plate splines with covariates including elevation, distance to the coast and three satellite‐derived covariates: maximum and minimum land surface temperature as well as cloud cover, obtained with the MODIS satellite platform [15]. Trabucco and Zomer (2019) provided aridity index and PET computed using the WorldClim climatology [18, https://cgiarcsi.community/2019/01/24/global-aridity-index-and-potential-evapotranspiration-climate-database-v2/]. The second climatology, CHELSA, was derived from a quasi-mechanistically statistical downscaling of the ERA interim global circulation model with a GPCC bias correction [16]. This climatology is based on averaged climatic condition between 1979–2013. CHELSA incorporates topoclimate (e.g., orographic rainfall and wind fields), which is highly relevant for islands. We used the ENVIronmental Rasters for Ecological Modeling R-package [39] to compute the PET and aridity index using the CHELSA climatology.

WWF ecoregions

The WWF ecoregions are polygon shapefiles that included 867 land units classified into 14 different ecoregions [40]. Each represents large units of land or water containing a geographically distinct assemblage of species, natural communities, and environmental conditions [40]. WWF has identified 53 ecoregions that fall within the Tropical and Subtropical Dry Broadleaf Forest category (Fig 1) (S1 Appendix).

Fig 1. Distribution of 540 tropical dry forest plots and the World Wildlife Fund’s tropical and subtropical dry broadleaf forest ecoregion.

Fig 1

Open in a new tab

Field data sets

Biodiversity data, including vegetation plots, are valuable for addressing macro-ecological questions about community patterns and processes [41, 42], as well as global conservation problems in the global change era [43]. We tested the occurrence of the tropical dry forest biome based on bioclimatic definitions against 540 verified locations of tropical dry forest vegetation compiled from two primary sources (Fig 1). First, we searched Web of Science (v.5.32), Scopus, and Google Scholar databases for peer-reviewed articles published between January 1990 and September 2019. We queried titles, abstracts, and keywords for the following terms: tropical*dry* forest* plots*. We selected the peer-reviewed articles based on three criteria: 1) plots classified as dry forest (closed-canopy) following author’s classifications, and not savannas, woodlands (open-canopy), riparian or flooded forests, 2) articles needed to include published latitude and longitude of the plot location to within 1 km, and 3) forest had to be composed of drought tolerant tree species native to the region. Second, we searched four global forest data repositories for sites classified as tropical dry forest. Our search using the Global Biodiversity Information Facility, as well as ForestPlots.net, yielded much of the same plot data we had collected from our list of peer-reviewed studies. The Dryad data repository, however, yielded 40 additional dry forest plots from three peer-reviewed studies of dry forest across Latin America [44–46]. Additionally, we collected 150 sites from the United States Geological Survey’s Forest Inventory and Analysis program covering tropical dry forest from the United States and territories (Hawai’i, Puerto Rico and US Virgin Islands). We used Google Earth to verify the coordinates were within areas with closed canopy forest and to establish seasonality using built-in, time-lapse imagery collected since 1984 across varying seasons. These plot locations were used to determine whether the tropical dry forest bioclimatic definitions accurately capture observed locations of the tropical dry forest vegetation type (S2 Appendix).

Global forest cover and change

Originally developed to provide high-resolution global maps of forest cover change from 2000 to 2012 using Landsat 7 imagery, the Global Forest Change data set has grown to include time series analysis of Landsat 5, 7 and Landsat 8 imagery now covering forest cover and forest loss from 2000 to 2020 [18]. Each pixel has a spatial resolution of 1 arc second, or roughly 30 m. We calculate mean, minimum, and maximum percent canopy cover over each field plot at 1 km resolution to provide estimates of tropical dry forest cover and variability across the landscape in 2000 (S2 Appendix). Within the bioclimatic envelope of tropical dry forest biome that best matched the field plots, we identified forest areas as areas with a forest cover ≥ 40% [17, 23] for the year 2000, 2001 and 2020, and calculated forest cover change between 2001 and 2020.

Data analyses

Spatial coverage

We divided global results into six regions (Africa, North and Central America, South America, South Asia, and South East Asia and Pacific). We further sub-divided regions using biodiversity hotspots [47, 48] (S4 Appendix) and countries [49] (S5 Appendix).

Bioclimatic potential for tropical dry forest

We used geospatial analysis tools to compile spatial data and develop binary raster maps (dry forest, non-dry forest) by applying the four bioclimatic definition of dry forest. Free and open-source Python software was used to manipulate raw data, primarily using Geospatial Data Abstraction Library and RasterIO. We used the cloud-based geospatial analysis platform Google Earth Engine [50] to analyze forest cover [27]. All analyses were compiled in a WGS84 projection and code used in our analysis is available in S6 Appendix.

Statistical analyses

Regions, biodiversity hotspots, and country level data on area of tropical dry forest biome potential extent were examined for a normal distribution using one-sample Shapiro-Wilk normality test for small samples (≤ 30) and Kolmogorov-Smirnov tests (> 30). Non-parametric (Wilcoxon’s rank sum test) tests were used to identify significant differences in area among bioclimatic definitions and between the WorldClim and CHELSA data sets for regions and biodiversity hotspots. Parametric (paired T-tests) were used to identify significant differences in area among bioclimatic definitions and between the WorldClim and CHELSA data sets for countries.

Comparisons with WWF ecoregions, bioclimatic definitions, and field plots

We calculated the area of each WWF ecoregion and compared results with four bioclimatic definitions (S1 Appendix). The agreement between WWF ecoregions, bioclimatic definitions, and field plots were measured using the percentage of dry forest plots that have been classified as dry forest according to the WWF ecoregions and different bioclimatic definitions (i.e. the percentage of true positives). However, we lack non-tropical dry forest plots such as tropical rain forests to calculate the percentage of false positives (i.e. non dry forest plots classified as dry forest) and true negative (i.e. non dry forest plots not classified as dry forest). Maximizing only the percentage of true positives might lead to selecting the best bioclimatic definition that predicts the larger extent of dry forest. Thus, we examine the percentage of true positives as a function of the extent of dry forest predicted by the different bioclimatic definitions. Given that the percentage of true positives likely depends on our sample of field plots, we used bootstrapping to compute a 95% confidence interval (CI) around the estimated values. Bootstrapping was done using 1,000 iterations and the lower and upper bounds of the CI were estimated using the quantiles 0.025 and 0.975, respectively.

Tropical dry forest vegetation cover and change

We selected the bioclimatic definition that had the highest agreement with field plots at a global spatial extent and calculated forest cover in 2000 and gross loss from 2001–2020 based on the global forest change data set using a forest cover threshold of ≥ 40% to define closed canopy forest [17, 23, 27]. We also calculated similar results for (i) all tropical forests, (ii) the WWF tropical and subtropical dry broadleaf forest ecoregions, and (iii) the consensus (overlap) of the four bioclimatic definitions (aridity index, Murphy and Lugo, FAO, DryFlor). We also estimate the extent of forest cover using thresholds of ≥ 10% [17] and ≥ 60% [51, 52] for the bioclimatic definition that had the highest agreement with field plots at a global spatial for global comparisons (S3 Appendix).

Results

Comparisons of bioclimatic definitions of tropical dry forest biome

Estimates of tropical dry forest biome extent based on bioclimatic definitions varied (Table 2). The global extent estimated from the aridity index, Murphy and Lugo, and FAO climate definitions were relatively similar (~15,000,000 km2), while DryFlor covered the smallest area estimated ~10,000,000 km2 (Table 2).

Table 2. Estimate of tropical dry forest biome area (km2), globally and in different regions, based on four bioclimatic definitions (Murphy and Lugo, Food and Agriculture Organization of the United Nations [FAO], DryFlor, Aridity Index) using two projections.

Regions Murphy & Lugo FAO DryFlor Aridity Index
Global 15,300,143 15,514,946 10,370,038 15,777,797
16,123,939 15,177,193 10,820,627 14,376,146
Africa 6,825,248 7,480,815 5,005,193 8,000,948
7,237,338 7,700,711 5,243,736 7,420,821
North and Central America 590,609 689,325 284,154 1,166,918
652,516 811,289 326,525 1,028,800
South America 5,736,592 3,134,372 3,958,243 2,605,596
6,042,593 3,774,250 4,023,236 1,866,315
South Asia 437,803 1,261,394 112,170 1,983,914
481,186 1,230,178 153,111 1,417,872
South East Asia/Pacific 1,709,888 1,687,644 1,010,275 2,020,421
1,710,303 1,660,763 1,074,017 2,642,338
Open in a new tab

WorldClim data presented above CHELSA. Highest extents in bold.

The aridity index using WorldClim and Murphy and Lugo using CHELSA estimated the largest extents of the tropical dry forest biome by region followed by the aridity index and FAO using CHELSA (Table 2). The boundaries of the tropical dry forest biome extent were more homogenous (smoother) for WorldClim than CHELSA data (Figs 2 and 3). Consensus maps showed that the area predicted to support tropical dry forest biome based on all four bioclimatic definitions (aridity index, Murphy and Lugo, FAO, DryFlor) was smaller than the combined non-consensus areas predicted by at least one bioclimatic definition. WorldClim contained larger areas of a consensus for all four bioclimatic definitions than CHELSA (Figs 2E and 3E).

Fig 2. Global distribution of tropical dry forest biome using WorldClim.

Fig 2

Open in a new tab

Based on (a) aridity index, (b) Murphy and Lugo, (c) Food and Agriculture Organization of the United Nations (FAO), (d) DryFlor bioclimatic definitions using WorldClim and (e) overlap of all four bioclimatic definitions (aridity index, Murphy and Lugo, FAO, and DryFlor).

Fig 3. Global distribution of tropical dry forest biome using CHELSA.

Fig 3

Open in a new tab

Based on (a) aridity index (b) Murphy and Lugo, (b) Food and Agriculture Organization of the United Nations (FAO), (c) DryFlor bioclimatic definitions using CHELSA and (e) overlap of all four bioclimatic definitions (aridity index, Murphy and Lugo, FAO, and DryFlor).

On a regional scale, there was no significant difference in areas estimated by climatic data sets used (Wilcoxon’s rank sum tests p > 0.05). Murphy and Lugo’s definition estimated the largest tropical dry forest biome extent within biodiversity hotspots and countries (S4 and S5 Appendices). However, there was no significant difference between climatic data sets used in biodiversity hotspots (Wilcoxon’s rank sum tests p > 0.05). At the country level, there was a significant difference in climatic data set used for all four bioclimatic definitions (paired t-test: Murphy and Lugo p = 0.004, FAO p = 0.012, DryFlor p = 0.018, aridity index p = 0.005).

Comparisons to WWF ecoregions

The 53 WWF Ecoregions defined as tropical and subtropical dry broadleaf forest extend over an area of 2,918,256 km2 with South Asia having the largest extent and Africa the smallest extent (Table 3). Globally, the aridity index and FAO definitions using WorldClim had the highest overlap with WWF Ecoregions (57%) followed by FAO using CHELSA (56%), and Murphy and Lugo using CHELSA (44%) (Table 3). DryFlor had the lowest overlap. There was high variation among regions, however. Murphy and Lugo and the aridity index had the highest overlaps within most WWF tropical and subtropical dry broadleaf forest ecoregions by region.

Table 3. Extent of World Wildlife Fund’s Tropical and Subtropical Dry Broadleaf Forest (TSDBF) ecoregion (km2) and overlap (%) with bioclimatic definitions (Murphy and Lugo, Food and Agriculture Organization of the United Nations [FAO], DryFlor, Aridity Index).

Regions TSDBF Ecoregion Murphy & Lugo FAO DryFlor Aridity Index
Global 2,918,256 43% 57% 22% 57%
44% 56% 22% 35%
Africa 185,624 70% 69% 3% 48%
60% 56% 1% 13%
North & Central America 511,057 45% 63% 23% 67%
46% 64% 24% 41%
South America 642,243 58% 47% 31% 43%
60% 51% 28% 32%
South Asia 983,944 9% 60% 2% 82%
11% 61% 2% 56%
South East Asia/Pacific 593,608 75% 53% 49% 24%
76% 49% 51% 4%
Open in a new tab

WorldClim data presented above CHELSA. Highest agreement with ecoregion in bold.

Validation using field plots

We identified 540 field plots that met our search criteria. Most plots were between 0.03 and 0.1 ha, with diameter at breast height ranging from 1 cm to 10 cm. Species lists were available for 525 sites, however, full species density data were only available at 487 sites (S2 Appendix). Mean canopy cover estimates (1 km) were available for 534 sites (no forest cover data was available for the Mariana or Marquesas islands) and averaged 37.8% canopy cover, ± 25.9 in 2000 while maximum canopy cover estimates averaged 48.1 canopy cover, ± 30.1 (S2 Appendix).

Only 40% of the tropical dry forest field plots fell within the WWF tropical and subtropical broadleaf forest ecoregion boundaries (Table 4). This ranged from a low of 0% overlap in Africa to a high of 67% in South East Asia/Pacific. At a global scale, FAO bioclimatic definition using CHELSA had the highest agreement (81%) with field plots followed by Murphy and Lugo using CHELSA (75%). FAO (70%), Murphy and Lugo (66%) and the aridity index (64%) using WorldClim accounted for the next three highest overlaps with field plots. DryFlor had the lowest overlap (33–37%).

Table 4. Number of field plots and percent agreement with field plots (%) and World Wildlife Fund’s Tropical and Subtropical Dry Broadleaf Forest (TSDBF) ecoregions and four bioclimatic definitions (Murphy and Lugo, Food and Agriculture Organization of the United Nations [FAO], DryFlor, Aridity Index).

Regions Field Plots TSDBF Murphy & Lugo FAO Dryflor Aridity Index
Global 540 40% 66% 70% 33% 64%
75% 81% 37% 6%
Africa 54 0% 76% 83% 52% 41%
81% 76% 44% 11%
North & Central America 348 43% 63% 74% 26% 78%
77% 89% 32% 4%
South America 48 33% 79% 62% 58% 42%
79% 77% 58% 8%
South Asia 35 57% 91% 91% 34% 77%
89% 94% 31% 14%
South East Asia/Pacific 48 67% 52% 31% 42% 8%
50% 31% 42% 0%
Biodiversity Hotspots
Caribbean Islands 227 49% 62% 73% 28% 87%
85% 90% 37% 1%
Cerrado 17 0% 100% 53% 94% 6%
100% 76% 94% 6%
Coastal Forests of East Africa 12 0% 83% 92% 17% 75%
67% 75% 25% 8%
Guinean Forests of West Africa 10 0% 100% 80% 90% 0%
100% 80% 50% 0%
Indo Burma 7 100% 100% 43% 100% 29%
100% 43% 100% 0%
Madagascar & Indian Ocean 1 0% 0% 0% 0% 100%
0% 0% 0% 100%
Mesoamerica 38 89% 66% 84% 16% 82%
24% 76% 11% 29%
New Caledonia 5 100% 80% 40% 80% 0%
80% 100% 80% 0%
Polynesia/Micronesia 26 62% 42% 23% 31% 4%
35% 15% 27% 0%
Tropical Andes 5 20% 40% 40% 40% 40%
0% 80% 20% 40%
Wallacea 3 100% 67% 100% 33% 33%
100% 67% 33% 0%
Western Ghats & Sri Lanka 10 0% 100% 60% 100% 10%
100% 90% 100% 0
Open in a new tab

WorldClim data presented above CHELSA. Highest agreement with field plots in bold.

Most of the percentage of true positive were not correlated with the estimate of tropical dry forest biome extent (Fig 4). Thus, the better definition (e.g. the one which maximize the percentage of true positives) was not necessarily the definition that predicted the larger extent of the tropical dry forest biome. Globally the FAO-CHELSA was the best bioclimatic definition (80.9%), however, best definitions differ between the different regions. FAO-CHELSA was the best definition for North, Central (88.4%). There was no clear difference between FAO and Murphy and Lugo bioclimatic definitions in Africa, South America, and South Asia as evidenced by overlaps between confidence intervals. The percentage of true positives for bioclimatic definitions were lower for South East Asia/Pacific where the TSDBF definition performed better (65.6%).

Fig 4. Estimates of the percentage of true positives and biome extent using bootstrapping with 95% confidence intervals.

Fig 4

Open in a new tab

Points represent the average values and segments the lower and upper bounds of the intervals (i.e. the quantiles 0.025 and 0.975, respectively) using 1,000 iterations. Dotted black lines represent significant (p < = 0.05) linear relationships between the estimate of tropical dry forest biome area and the percentage of true positive.

Estimates of tropical dry forest cover and change

Using the FAO bioclimatic definition and CHELSA data set to estimate closed forest cover and change since 2000 (Fig 5), we estimated a total of 4,931,414 km2 of closed canopy forest in 2000 and 4,369,695 km2 of closed canopy forest in 2020. We estimate a potential gross loss of closed canopy tropical dry forest amounting to 561,719 km2 (11%) between 2001 and 2020. Tropical dry forest was estimated to account for 25% of all closed canopy forest cover in the tropics in 2000 (Table 5), which is considerably larger than WWF tropical and subtropical dry broadleaf forest ecoregion estimates (782,492 km2 or 4% of all tropical forests) and consensus maps based on all four definitions with either climatic dataset. Our estimates of tropical dry forest cover are greater than other global estimates that range from 1,048,700 km2 of tropical dry forest (≥ 40% closed canopy) [23] to 3,380,000 km2 [17].

Fig 5. Forest cover and change from FAO bioclimatic definition, CHELSA climate data set, and closed canopy cover.

Fig 5

Open in a new tab

(a) Global, (b) Africa, (c) North and Central America, (d) South America, (e) South Asia, and (f) Southeast Asia & Asia Pacific.

Table 5. Comparisons of best methods for estimating closed canopy (≥ 40% closed canopy) tropical dry forest extent and forest cover (km2) in 2000 and 2020 between 30˚N and 30˚S.

Definition Source Forest Cover 2000 Estimated Cover 2020 Gross Loss 2001–2020 Percent Gross Loss
Pantropics Global Forest Cover 20,064,659 17,962,101 2,102,558 10.48%
FAO CHELSA 4,931,414 4,369,695 561,719 11.39%
TSDBF Wild Wildlife Fund 782,492 670,336 112,156 14.33%
Consensus (all) CHELSA 20,758 19,166 1,592 7.67%
Consensus (all) WorldClim 332,336 305,973 26,363 7.93%
Open in a new tab

FAO, Food and Agriculture Organization of the United Nations; CHELSA, Climatologies at High-resolution for the Earth’s Land Surface Areas; TSDBF, tropical and subtropical dry broadleaf forest.

Discussion

WWF ecoregions

There was low overlap between WWF tropical and subtropical dry broadleaf forest ecoregions and bioclimatic definitions of tropical dry forest biomes (<57%) and field plots (40%) at a global scale. Tropical dry forest field plots occurred in other WWF ecoregions, including tropical and subtropical moist broadleaf forests (14% of the plots), tropical and subtropical coniferous forests (11%), tropical and subtropical grassland, savanna and shrubland (10%), and deserts and xeric shrublands (9%). This is of concern because the WWF ecoregions are widely used standard for delineating global biomes and estimating forest cover, canopy height, biomass, stand density, and anthropogenic disturbance [3, 23, 29, 31, 32, 53] (S1 Appendix).

We found that large areas of the tropical dry forest biome will be missed when using these WWF boundaries and that caution should be used when applying WWF boundaries for country scale analyses. At the country scale, higher resolution vegetation maps that identify forests, woodlands, shrublands, and savannas should be used. While WWF ecoregions may be appropriate for global or macro-scale analyses of tropical forests biomes, our results suggest that they may not be appropriate for analyzing tropical dry forest vegetation types or the impacts of climate change on tropical dry forests. Instead, the bioclimatic definitions of tropical dry forest from FAO with CHELSA datasets should be used.

Bioclimatic definitions of tropical dry forest

According to our comparison with field plots data, the FAO and Murphy and Lugo bioclimatic definitions of tropical dry forest extent performed best. There were 358 field plots (66%) that overlapped with both the FAO and Murphy and Lugo bioclimatic definitions (80 plots FAO only, 47 plots Murphy and Lugo only) and 54 field plots (10%), primarily in regions with long dry seasons (dry season > 9 months) or regions with over 2000 mm of mean annual precipitation such as Pacific islands, that did not overlap with either definition. The FAO definitions included drier regions with more seasonal forests while the Murphy and Lugo definition encompassed more moist regions with less seasonal forests. Thus, the FAO definitions appear to be best for identifying the core of tropical dry forest biome at a global spatial scale, while the Murphy and Lugo definition may perform best for identifying transitional areas between dry and moist or rain forest biomes. The DryFlor definitions worked well in the Neotropics but did not perform well at a global and regional spatial scales. The aridity index does not account for precipitation seasonality and appears to include the drier extremes of what would be considered tropical dry forest. Indeed, although forests that occur within an area with aridity index < 0.65 can be considered “tropical dry forests” because they occur in the tropics and clearly occur in dry conditions, the absence of a minimum precipitation threshold and seasonality criteria clearly misses a number of well-known tropical dry forest sites such as tropical dry forests of Central America and the Caribbean [2]. Future studies and comparisons of tropical dry forest biome should be undertaken using FAO bioclimatic definitions.

Climatic data sets

WorldClim and CHELSA climate data sets did not significantly change estimates of tropical dry forest area based on all four bioclimatic definitions at the regional and biodiversity hotspot scales. However, using different climate data sets did identify different extents and areas at the country level (Figs 2 and 3). As analyses proceed to smaller spatial extents from regions to countries, the climatic data set used results in significantly different estimates of area. Thus the data set used becomes more important at small spatial extent. The CHELSA dataset generally had higher mean annual precipitation estimates than WorldClim but there was little difference in temperature estimates. The CHELSA data best matched field verified locations of tropical dry forest extent and offered some advantages over interpolated WorldClim data when assessing the future impacts of climate change. Since CHELSA is a downscaled global circulation models, it can easily be used to estimate the future climatic variables a 10 km and over last 4 million years at 10 km resolution [54]. The CHELSA data set also had the highest agreement with field plots and appears to be the most appropriate climate data set to use for studying tropical dry forests and possibly other tropical ecosystems.

Tropical dry forest cover and changes

Our estimate of closed canopy (≥ 40%) tropical dry forest in 2020 is 4.9 million km2, based on the best performing parameters (FAO bioclimatic definition using CHELSA) and this is larger than previous estimates. Miles et al. (2006) estimated 1,048,700 km2 of tropical dry forest (≥ 40% closed canopy) using MODIS imagery (500 m), while Bastin et al. (2017) estimated 3.38 million km2 of closed canopy (≥ 40%) dry forest in the tropics. Miles et al. (2006) estimate thus appears very conservative and more in line with our estimates in WWF ecoregions, but well above estimates of closed canopy forest based on consensus of all four climatic definitions. Although global comparisons of studies are difficult, it is clear that tropical dry forest vegetation types are experiencing a decline. Estimates based on all definitions of tropical dry forest (Table 5) are between 7% and 15% from 2000 to 2020 with closed canopy tropical dry forests experiencing high rates of deforestation and gross forest cover loss (S3 Appendix).

Limitations

There is currently a lack of global comparative ground validation for our FAO CHELSA tropical dry forest biome and vegetation maps and field validation or very high resolution remote sensing validation is needed in the future [17]. While a growing number of tropical dry forest plots have been established worldwide, there is also a great deal of spatial bias in the number, extent, and density of tropical dry forest field plots. Our data set of 540 tropical dry forest plots were biased towards North and Central America, which accounted for 64% of the plots. Among the North and Central American plots, 65% were located in the Caribbean [8, 33]. There were relatively few comparative tropical dry forest plots from Madagascar (1 plot), eastern India, Indonesia, and Australia. Furthermore, there are a number of countries such as Angola, China, and Zambia that appear to have tropical dry forest biome and vegetation type but no field plots (S5 Appendix). This may be due to our search criteria for tropical dry forest plots which may have missed local names of dry forest vegetation types (e.g. monsoon forest), however, these areas deserve a high priority for establishing standardized plots in the future.

Almost all global climate data sets contain inaccuracies especially on islands with few weather stations and high heterogeneity in the landscape [15, 42]. Miranda et al. (2018) found that WorldClim data resulted in the misclassification of 15 to 20% of tropical dry forests in lowland South America while the addition of soil data improved classification by 3%. Delimiting the tropical dry forest biome is complex because tropical forest historically occurred across environmental and disturbance gradients, does not always have solid boundaries as depicted in GIS, and can be a continuum from drier to wetter areas. On the dry end of the spectrum, dry forests generally grade into savannas, shrublands, woodlands within the same climatic condition. The actual extent of vegetation within this region is highly impacted by substrate, soil moisture, and fire [19, 33]. On the wetter end of the spectrum, tropical dry forest generally transitions into moister forests, a zonal riparian forests, or swamps with increased canopy heights and an increasing number of tree species that are less susceptible to seasonal drought or dry soil conditions. These gradients clearly existed in the past but have been significantly impacted by humans certainly over the last 100 years [3, 4, 55]. It should be remembered that we are currently examining a disturbed landscape in most dry forest regions. Many dry forest regions are deforested, fragmented, and degraded based on estimate of forest cover over plots at a landscape scale (1 km) and it is difficult to precisely identify and delineate their distribution using global scale data (e.g. Hawaiian Islands).

Thus, climate definitions of tropical dry forest provide only a first order hypothesis and standard and repeatable method for identifying and estimating the spatial extent of tropical dry forest biome. Then analyses based on fine-scale features, such as forests (canopy height ≥ 3 m, ≥ 40% canopy cover), woodlands (< 40% canopy cover), shrublands (stems < 3 m), and savannas (< 1 m, grasses) can be undertaken at a higher resolution such as 30 m from Landsat [27, 56] or very high resolutions < 1 m in Google Earth [17, 50].

Future research and applications

Our research focused on true positives (e.g. the percentage of dry forest plots that have been classified as tropical dry forest according to the different bioclimatic definitions) and false negatives (e.g. the percentage of dry forest plots that have not been classified as tropical dry forest according to the different bioclimatic definitions). However, future research at the biome level should include a confusion matrix with false positives (e.g. the percentage of non-dry forest plots, such as tropical rain forest, that have been classified as dry forest according to the different bioclimatic definitions), and the true negatives (e.g. the percentage of non-dry forest plots that have not been classified as dry forest according to the different bioclimatic definitions). This can be done at the biome level using high resolution locations of tropical rain forest plots.

Comparative analyses of biogeography, threats, and conservation can be undertaken using FAO (CHELSA) or Murphy and Lugo (CHELSA) boundaries as a baseline. It is well-established that biodiversity is greatly threatened by human activity [49, 57] and land cover changes such as those linked to human-induced forest loss, fragmentation, and degradation represent the largest current threat to biodiversity [57–59]. Miles et al. (2006) identified five global threat metrics for tropical dry forest including climate change (at a spatial resolution of 300 km), forest fragmentation (500 m), fire (10 km), agrosuitability (10 km), and population (10 km). Since this seminal work, there has been a significant increase in the temporal and spatial resolution of GIS and remote sensing data for these threat metrics, such as predicted future climate change (10 km), forest fragmentation (≤ 30 m), fire and burned areas (375 m), agrosuitability and grazing (1 km), and population (1 km) [26, 60–64]. Thus, there are currently a number of global threat metrics that can be analyzed with FAO and CHELSA that might significantly improve our understanding of the health of tropical dry forests.

Twenty-five of the 36 global biodiversity hotspots appear to contain tropical dry forest biome based on three climatic definitions of tropical dry forests (S4 Appendix), possibly highlighting the importance of this biome and vegetation type. Analyses of protected areas, old growth forest, and threatened and endangered species are also needed within tropical dry forest biome and vegetation type. There has been a rapid increase in the number of protected areas and clearly there is a need to identify how well different regions are protected [23]. Using time series Landsat and fire data sets since 2000, it should be possible to identify stable closed canopy tropical dry forests that have not experienced fire or forest loss since 2000 [21, 63]. These forest areas may be some of the best preserved or relictual forests and contain an increasing rare combination of dry forest species. This may be especially true for regions like Africa, Asia, and Australia where fires and grazing are common and closed canopy dry forests are isolated in natural refugia [55].

Conclusions

More precise knowledge of the extent and location of global tropical dry forest is needed to better understand forest cover dynamics, biodiversity threats and the conservation status of this vulnerable biome. We produced reliable data on tropical dry forest extent and overall forest loss spanning the last 20 years. We also demonstrate that nearly half of all tropical dry forest will be missed using the WWF ecoregion classification alone, with only 40% of our 540 tropical dry forest field plots falling within the WWF tropical and subtropical broadleaf forest ecoregion. We found that using the FAO bioclimatic definition applied to the CHELSA climate data set is a promising approach to identify potential dry forest extent based on the high agreement with field data. Closed canopy tropical dry forest currently occurs in 25 of the world’s 36 biodiversity hotspots, a majority of pantropical countries, and has been experiencing high rates of deforestation (11%) between 2000 and 2020. Identifying the extent and distribution of tropical dry forest regions are important for future research on the impacts of climate change and understanding the status of the world’s tropical dry forest for conservation purposes.

Supporting information

S1 Appendix. Tropical and subtropical dry broadleaf forest defined by World Wildlife Fund.

Name of ecoregion, extent, and forest cover.

(CSV)

Click here for additional data file. (35.5KB, csv)
S2 Appendix. Tropical dry forest field plots.

Comprehensive list of the 540 verified tropical dry forest plots, their coordinates, references, and bioclimatic characteristics.

(CSV)

Click here for additional data file. (116.5KB, csv)
S3 Appendix. Forest cover and extent statistics.

Forest cover statistics for 540 tropical dry forest plots using resampled Global Forest Watch tree canopy cover (1 km2) data for the year 2000 and a 1 km buffer around each plot. Total forest cover estimates and gross forest cover loss are also calculated for open canopy (≥ 10%) and closed canopy (≥ 40% and ≥ 60%) for FAO CHELSA from 2000 to 2020.

(DOCX)

Click here for additional data file. (1.6MB, docx)
S4 Appendix. Biodiversity hotspots for mesoscale analyses of tropical dry forest.

Extent and agreement of tropical dry forest area.

(CSV)

Click here for additional data file. (2.1KB, csv)
S5 Appendix. Tropical dry forest extent at country level.

Summaries of tropical dry forest extent based on climatic definitions and datasets at the country level.

(CSV)

Click here for additional data file. (6.5KB, csv)
S6 Appendix. Computer code for global analysis of tropical dry forest.

Code used to accomplish a global analysis of tropical dry forest extent and cover based on a suite of remote sensing and modeled climate and forest data, as well as GIS layers of political and conservation boundaries.

(PDF)

Click here for additional data file. (315.5KB, pdf)

Acknowledgments

We gratefully acknowledge DryFlor.org (Karina Banda, Kyle Dexter, Danilo Neves, and Toby Pennington) for sharing advice and insights on data collection and neotropical dry forest inventories. We also would like to thank Kyle Cavanaugh for comments and suggestions.

Data Availability

All relevant data are within the manuscript and its Supporting Information files.

Funding Statement

JPO received the 2019 Geospatial @ UCLA Summer Fellowship which helped fund this research. URL:https://gis.ucla.edu/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

  • 1.Murphy PG, Lugo AE. Ecology of tropical dry forest. Annual Review of Ecology and Systematics. 1986. November;17(1):67–88. [Google Scholar]
  • 2.Janzen DH. Tropical dry forests. Biodiversity. 1988. January 15:130–7. [Google Scholar]
  • 3.Riggio J, Baillie JE, Brumby S, Ellis E, Kennedy CM, Oakleaf JR, et al. Global human influence maps reveal clear opportunities in conserving Earth’s remaining intact terrestrial ecosystems. Global Change Biology. 2020. June 5. 10.1111/gcb.15109 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Blackie R, Baldauf C, Gautier D, Sunderland T. Tropical Dry Forests: The State of Global Knowledge and Recommendations for Future Research, vol. 2. CIFOR, Indonesia. 2014. [Google Scholar]
  • 5.Siyum ZG. Tropical dry forest dynamics in the context of climate change: syntheses of drivers, gaps, and management perspectives. Ecological Processes. 2020. December;9(1):1–6. [Google Scholar]
  • 6.Bruegmann MM. Hawai’i’s dry forests. Endangered Species Bulletin. 1996. 11:26–27. [Google Scholar]
  • 7.Cabin RJ, Weller SG, Lorence DH, Flynn TW, Sakai AK, Sandquist D, et al. Effects of long‐term ungulate exclusion and recent alien species control on the preservation and restoration of a Hawaiian tropical dry forest. Conservation Biology. 2000. April;14(2):439–53. [Google Scholar]
  • 8.Banda K, Delgado-Salinas A, Dexter KG, Linares-Palomino R, Oliveira-Filho A, Prado D, et al. Plant diversity patterns in neotropical dry forests and their conservation implications. Science. 2016. September 23;353(6306):1383–7. 10.1126/science.aaf5080 [DOI] [PubMed] [Google Scholar]
  • 9.Von Humboldt A. Aspects of Nature, in Different Lands and Different Climates: With Scientific Elucidations. in 2 Vols. Lea and Blanchard; 1849. [Google Scholar]
  • 10.Schimper AF. Plant-geography Upon a Physiological Basis. Clarendon Press; 1903. [Google Scholar]
  • 11.Box EE. Macroclimate and plant forms: an introduction to predictive modeling in phytogeography. Springer Science & Business Media; 2012. December 6. [Google Scholar]
  • 12.Meentemeyer V, Box EO. Scale effects in landscape studies. In Landscape heterogeneity and disturbance 1987. (pp. 15–34). Springer, New York, NY. [Google Scholar]
  • 13.Ratnam J, Bond WJ, Fensham RJ, Hoffmann WA, Archibald S, Lehmann CE, et al. When is a ‘forest’ a savanna, and why does it matter?. Global Ecology and Biogeography. 2011. September; 20(5):653–60. [Google Scholar]
  • 14.Pausas JG, Bond WJ. Humboldt and the reinvention of nature. Journal of Ecology. 2019. May;107(3):1031–7. [Google Scholar]
  • 15.Fick SE, Hijmans RJ. WorldClim 2: new 1‐km spatial resolution climate surfaces for global land areas. International Journal of Climatology. 2017. October;37(12):4302–15. [Google Scholar]
  • 16.Karger DN, Conrad O, Böhner J, Kawohl T, Kreft H, Soria-Auza RW, et al. Climatologies at high resolution for the earth’s land surface areas. Scientific Data. 2017. September 5;4:170122. 10.1038/sdata.2017.122 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Bastin JF, Berrahmouni N, Grainger A, Maniatis D, Mollicone D, Moore R, et al. The extent of forest in dryland biomes. Science. 2017. May 12;356(6338):635–8. 10.1126/science.aam6527 [DOI] [PubMed] [Google Scholar]
  • 18.Trabucco A, Zomer RJ. Global aridity index and potential evapotranspiration (ET0) climate database v2. CGIAR Consort Spat Inf. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Huang H, Xue Y, Li F, Liu Y. Modeling long-term fire impact on ecosystem characteristics and surface energy using a process-based vegetation–fire model SSiB4/TRIFFID-Fire v1.0. Geoscientific Model Development. 2020. December;13(12):6029–6050. [Google Scholar]
  • 20.Staver AC, Archibald S, Levin SA. The global extent and determinants of savanna and forest as alternative biome states. Science. 2011. October 14;334(6053):230–2. 10.1126/science.1210465 [DOI] [PubMed] [Google Scholar]
  • 21.Dexter KG, Pennington RT, Oliveira-Filho AT, Bueno ML, Silva de Miranda PL, Neves DM. Inserting tropical dry forests into the discussion on biome transitions in the tropics. Frontiers in Ecology and Evolution. 2018. July 24;6:104. [Google Scholar]
  • 22.Silva AC, Souza AF (2018) Aridity drives plant biogeographical sub regions in the Caatinga, the largest tropical dry forest and woodland block in South America. PLoS ONE 13(4): e0196130. 10.1371/journal.pone.0196130 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Miles L, Newton AC, DeFries RS, Ravilious C, May I, Blyth S, et al. A global overview of the conservation status of tropical dry forests. Journal of Biogeography. 2006. March;33(3):491–505. [Google Scholar]
  • 24.Fraser RH, Olthof I, Pouliot D. Monitoring land cover change and ecological integrity in Canada’s national parks. Remote Sensing of Environment. 2009. July 15;113(7):1397–409. [Google Scholar]
  • 25.Portillo-Quintero CA, Sánchez-Azofeifa GA. Extent and conservation of tropical dry forests in the Americas. Biological Conservation. 2010. January 1;143(1):144–55. [Google Scholar]
  • 26.Secades C, O’Connor B, Brown C, Walpole M. Earth observation for biodiversity monitoring: a review of current approaches and future opportunities for tracking progress towards the Aichi Biodiversity Targets. CBD technical series. 2014(72). [Google Scholar]
  • 27.Hansen MC, Potapov PV, Moore R, Hancher M, Turubanova SA, Tyukavina A, et al. High-resolution global maps of 21st-century forest cover change. Science. 2013. November 15;342(6160):850–3. 10.1126/science.1244693 [DOI] [PubMed] [Google Scholar]
  • 28.Hansen AJ, Neilson RP, Dale VH, Flather CH, Iverson LR, Currie DJ, et al. Global change in forests: responses of species, communities, and biomes: interactions between climate change and land use are projected to cause large shifts in biodiversity. BioScience. 2001. September; 51(9):765–779. [Google Scholar]
  • 29.Schmitt CB, Burgess ND, Coad L, Belokurov A, Besançon C, Boisrobert L, et al. Global analysis of the protection status of the world’s forests. Biological Conservation. 2009. October 1;142(10):2122–30. [Google Scholar]
  • 30.Sunderland T, Apgaua D, Baldauf C, Blackie R, Colfer C, Cunningham AB, et al. Global dry forests: a prologue. International Forestry Review. 2015. August 1;17(2):1–9. [Google Scholar]
  • 31.Crowther TW, Glick HB, Covey KR, Bettigole C, Maynard DS, Thomas SM, et al. Mapping tree density at a global scale. Nature. 2015. September;525(7568):201–5. 10.1038/nature14967 [DOI] [PubMed] [Google Scholar]
  • 32.Newbold T, Hudson LN, Arnell AP, Contu S, De Palma A, Ferrier S, et al. Has land use pushed terrestrial biodiversity beyond the planetary boundary? A global assessment. Science. 2016. July 15;353(6296):288–91. 10.1126/science.aaf2201 [DOI] [PubMed] [Google Scholar]
  • 33.Silva de Miranda PL, Oliveira‐Filho AT, Pennington RT, Neves DM, Baker TR, Dexter KG. Using tree species inventories to map biomes and assess their climatic overlaps in lowland tropical South America. Global Ecology and Biogeography. 2018. August;27(8):899–912. [Google Scholar]
  • 34.Holdridge LR. Life zone ecology. Life zone ecology. 1967(rev. ed.)). [Google Scholar]
  • 35.Griffith DM, Lehmann CE, Strömberg CA, Parr CL, Pennington RT, Sankaran M, et al. Comment on “The extent of forest in dryland biomes”. Science. 2017. November 17;358(6365). [DOI] [PubMed] [Google Scholar]
  • 36.Särkinen T, Iganci JR, Linares-Palomino R, Simon MF, Prado DE. Forgotten forests-issues and prospects in biome mapping using Seasonally Dry Tropical Forests as a case study. BMC ecology. 2011. December 1;11(1):27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Ostertag R, Inman-Narahari F, Cordell S, Giardina CP, Sack L. Forest structure in low-diversity tropical forests: a study of Hawaiian wet and dry forests. PloS one. 2014. August 27;9(8):e103268. 10.1371/journal.pone.0103268 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Slik JF, Franklin J, Arroyo-Rodríguez V, Field R, Aguilar S, Aguirre N, et al. Phylogenetic classification of the world’s tropical forests. Proceedings of the National Academy of Sciences. 2018. February 20;115(8):1837–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Title PO, Bemmels JB. ENVIREM: an expanded set of bioclimatic and topographic variables increases flexibility and improves performance of ecological niche modeling. Ecography. 2018. February;41(2):291–307. [Google Scholar]
  • 40.Olson DM, Dinerstein E, Wikramanayake ED, Burgess ND, Powell GV, Underwood EC, et al. Terrestrial Ecoregions of the World: A New Map of Life on EarthA new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. BioScience. 2001. November 1;51(11):933–8. [Google Scholar]
  • 41.Franklin J, Andrade R, Daniels ML, Fairbairn P, Fandino MC, Gillespie TW, et al. Geographical ecology of dry forest tree communities in the West Indies. Journal of Biogeography. 2018. May;45(5):1168–81. [Google Scholar]
  • 42.Ibanez T, Keppel G, Baider C, Birkinshaw C, Culmsee H, Cordell S, et al. Regional forcing explains local species diversity and turnover on tropical islands. Global Ecology and Biogeography. 2018. April;27(4):474–86. [Google Scholar]
  • 43.Franklin J, Serra‐Diaz JM, Syphard AD, Regan HM. Big data for forecasting the impacts of global change on plant communities. Global Ecology and Biogeography. 2017. January;26(1):6–17. [Google Scholar]
  • 44.Prado‐Junior JA, Schiavini I, Vale VS, Arantes CS, van der Sande MT, Lohbeck M, et al. Conservative species drive biomass productivity in tropical dry forests. Journal of Ecology. 2016. May;104(3):817–27. [Google Scholar]
  • 45.Salas‐Morales SH, González EJ, Meave JA. Canopy height variation and environmental heterogeneity in the tropical dry forests of coastal Oaxaca, Mexico. Biotropica. 2018. January;50(1):26–38. [Google Scholar]
  • 46.Suazo‐Ortuño I, Urbina‐Cardona JN, Lara‐Uribe N, Marroquín‐Páramo J, Soto‐Sandoval Y, Rangel‐Orozco J, et al. Impact of a hurricane on the herpetofaunal assemblages of a successional chronosequence in a tropical dry forest. Biotropica. 2018. July;50(4):649–63. [Google Scholar]
  • 47.Myers N, Mittermeier RA, Mittermeier CG, Da Fonseca GA, Kent J. Biodiversity hotspots for conservation priorities. Nature. 2000. February;403(6772):853–8. 10.1038/35002501 [DOI] [PubMed] [Google Scholar]
  • 48.Hoffman M, Koenig K, Bunting G, Costanza J, Williams K. Biodiversity Hotspots; 2016. (Version 2016.1) [Data set]. Available from: 10.5281/zenodo.3261807. [DOI] [Google Scholar]
  • 49.Earth Natural. 1:10m Cultural Vectors, Admin 0 –Countries; 2019. [Data set]. Available at: https://www.naturalearthdata.com/about/terms-of-use/. [Google Scholar]
  • 50.Gorelick N, Hancher M, Dixon M, Ilyushchenko S, Thau D, Moore R. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote sensing of Environment. 2017. December 1;202:18–27. [Google Scholar]
  • 51.Murphy BP, Bowman DM. What controls the distribution of tropical forest and savanna?. Ecology Letters. 2012. March; 15(7):748–758. 10.1111/j.1461-0248.2012.01771.x [DOI] [PubMed] [Google Scholar]
  • 52.Ratnam J, Tomlinson KW, Rasquinha DN, Sankaran M. Savannahs of Asia: antiquity, biogeography, and an uncertain future. Philosophical Transactions of the Royal Society B: Biological Sciences. 2016. September 19; 371(1703):20150305. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Gillespie TW, Lipkin B, Sullivan L, Benowitz DR, Pau S, Keppel G. The rarest and least protected forests in biodiversity hotspots. Biodiversity and Conservation. 2012. December 1;21(14):3597–611. [Google Scholar]
  • 54.Gamisch A. Oscillayers: A dataset for the study of climatic oscillations over Plio‐Pleistocene time‐scales at high spatial‐temporal resolution. Global Ecology and Biogeography. 2019. November;28(11):1552–60. 10.1111/geb.12979 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Lock JM. The seasonally dry vegetation of Africa: parallels and comparisons with the Neotropics. InNeotropical Savannas and Seasonally Dry Forests 2006. May 25 (pp. 449–467). CRC Press. [Google Scholar]
  • 56.Alencar A, Shimbo JZ, Lenti F, Balzani Marques C, Zimbres B, Rosa M, et al. Mapping three decades of changes in the brazilian savanna native vegetation using landsat data processed in the google earth engine platform. Remote Sensing 2020. March;12(6):924. [Google Scholar]
  • 57.Gaston KJ. Biodiversity and extinction: species and people. Progress in Physical Geography. 2005. June;29(2):239–47. [Google Scholar]
  • 58.Chapin Iii FS, Zavaleta ES, Eviner VT, Naylor RL, Vitousek PM, Reynolds HL, et al. Consequences of changing biodiversity. Nature. 2000. May;405(6783):234–42. 10.1038/35012241 [DOI] [PubMed] [Google Scholar]
  • 59.Menon S, Pontius RG Jr, Rose J, Khan ML, Bawa KS. Identifying conservation‐priority areas in the tropics: a land‐use change modeling approach. Conservation Biology. 2001. April 8;15(2):501–12. [Google Scholar]
  • 60.Elmqvist T, Pyykönen M, Tengö M, Rakotondrasoa F, Rabakonandrianina E, et al. (2007) Patterns of Loss and Regeneration of Tropical Dry Forest in Madagascar: The Social Institutional Context. PLoS ONE 2(5): e402. 10.1371/journal.pone.0000402 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Gallardo-Cruz JA, Meave JA, González EJ, Lebrija-Trejos EE, Romero-Romero MA, et al. (2012) Predicting Tropical Dry Forest Successional Attributes from Space: Is the Key Hidden in Image Texture? PLoS ONE 7(2): e30506. 10.1371/journal.pone.0030506 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Laurance WF, Useche DC, Rendeiro J, Kalka M, Bradshaw CJ, Sloan SP, et al. Averting biodiversity collapse in tropical forest protected areas. Nature. 2012. September;489(7415):290–4. 10.1038/nature11318 [DOI] [PubMed] [Google Scholar]
  • 63.Giglio L, Boschetti L, Roy DP, Humber ML, Justice CO. The Collection 6 MODIS burned area mapping algorithm and product. Remote Sensing of Environment. 2018. November 1;217:72–85. 10.1016/j.rse.2018.08.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64.Sun Q, Miao C, Duan Q, Ashouri H, Sorooshian S, Hsu KL. A review of global precipitation data sets: Data sources, estimation, and intercomparisons. Reviews of Geophysics. 2018. March;56(1):79–10. [Google Scholar]

Decision Letter 0

RunGuo Zang

19 Mar 2021

PONE-D-20-37820

Global tropical dry forest cover and extent: A comparative study of bioclimatic definitions

PLOS ONE

Dear Dr. Ocón,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================

ACADEMIC EDITOR: Please pay attention to the definition and range of tropical dry forests,and the methodsof data utilization.

Please ensure that your decision is justified on PLOS ONE’s publication criteria and not, for example, on novelty or perceived impact.

For Lab, Study and Registered Report Protocols: These article types are not expected to include results but may include pilot data. 

==============================

Please submit your revised manuscript by May 03 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

  • A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

  • A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

  • An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

RunGuo Zang

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Review of Ocon et al. Definition of Tropical Dry Forests

Overall comments:

There have been endless debates about what constitutes a tropical dry forest (especially as opposed to potentially similar vegetation such as savannas). The problem is partly compounded by the fact that tropical dry forests and savannas are highly dynamic and lie at the boundary of sharp vegetation transitions in some regions. The understorey may also change from predominantly grasses to woody plants within short time spans. Therefore, more objective criteria on defining tropical dry forests are certainly welcome. This study is a straightforward effort to try to delineate dry forests based on climatic criteria alone, which has its uses, particularly in large-scale analyses. I agree with the main conclusions of the study, namely, that the FAO definition of tropical dry forests seems to be the most appropriate. The much-cited Murphy and Lugo (1986) paper defines dry forests over an extreme range (250-2000 mm annually) of rainfall conditions.

However, I have several concerns regarding the methods.

Firstly, the authors do not account for the possibility that some of their validation plots may have been misclassified as dry forests. This is particularly true of some savanna regions (a lot has been written about this: see, for example, Ratnam et al., Proc. Roy. Soc. B, 2016). The authors should either account for this (for example, by checking sources for mention of % canopy cover and C4 grass cover) or discuss the limitation in detail. The authors must also discuss the possible circularity of some of the validation plot studies having adopted much-used definition such as Murphy & Lugo (1986) to classify their vegetation as dry forest. On a related note, the 40% canopy-cover cutoff to define a closed-canopy forest (L88, L219) needs a clearer justification or needs to be contrasted with results that use an alternate cutoff. Several studies treat < 60% as savanna and >= 60% as closed-canopy forest (e.g. Murphy & Bowman, Ecol. Lett., 2012).

Secondly, any study like this must assess results using a confusion matrix (proportions of false positives, true positives, false negatives, and true negatives). By only trying to maximize true positives and minimize false negatives, their criteria become overly liberal, potentially incorrectly including other vegetation types such as moist forest, savanna, shrublands, etc.

Finally, as they correctly note, the validation plot data set is highly biased in its geographical representation. While partly reflective of differences in research effort, it is also partly because their search criteria are inadequate as they miss several studies from South Asia and Indo-China (which host large extents of tropical dry forest) that should have been included.

Specific comments:

Line 52: should be “complemented” (not “complimented”)

Line 55-59: Seems out of place in the first paragraph of the Introduction. These do not provide any insight into the research questions but merely state the type of climate data sets available. I suggest these are shifted to the Methods section or at least reduced to a single sentence. In fact, I would go further and suggest that lines 72-108 are brought up front in the Introduction (after line 54) in order to clearly articulate the purpose of this study. The authors can then mention the availability of global data sets on climate, vegetation, land-cover and so on which could be used to resolve the questions.

Line 299: word missing in the sentence. In fact, there is a need for careful copyediting of the manuscript to correct minor errors.

Line 411-412: These categories are not mutually exclusive: C4 grasses can exist under a range of canopy covers.

Reviewer #2: The manuscript is an interesting address of the potential extent of the tropical dry forest biome based on bioclimatic definitions and climatic data sets to improve global estimates of distribution, cover, and change over time.

The data support the conclusions and answer the research question.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: ADEL H YOUKHANA

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Attachment

Submitted filename: Reviewer-1 comments.pdf

Click here for additional data file. (99.2KB, pdf)
PLoS One. 2021 May 20;16(5):e0252063. doi: 10.1371/journal.pone.0252063.r002

Author response to Decision Letter 0

3 May 2021

Response to reviewers in blue

All authors would like to thank the reviewers for their time and comments and suggestions that significantly improve the manuscript.

Reviewer #2

Review of the manuscript PONE-D-20-37820

The manuscript is an interesting address of the potential extent of the tropical dry forest biome based on bioclimatic definitions and climatic data sets to improve global estimates of distribution, cover, and change over time.

However, I have certain concern about:

-The authors try to do many comparisons in this manuscript but statistically there was no test used to do the real comparisons.

The reviewer is correct. The agreement between bioclimatic definitions and ground survey was measured using only the percentage of True Positives, i.e., the percentage of dry forest plots that have been classified as dry forest according to the different bioclimatic definitions. We also have the False Negatives, i.e., the percentage of dry forest plots that have not been classified as dry forest according to the different bioclimatic definitions. However, we don't have the False Positive, i.e., the percentage of non-dry forest plots that have been classified as dry forest according to the different bioclimatic definitions, and the True Negative, i.e., the percentage of non-dry forest plots that have not been classified as dry forest according to the different bioclimatic definitions.

So we have two solutions (1) we add other plots (tropical rain forests) to compute False Positives and True Negatives, (2) we keep focusing on True Positives and look at the True Positive/Extent ratios and we better acknowledge the limitation of our manuscript.

Collecting high resolution location data for tropical rain forests at a global scale currently seems a bit outside the scope of this study. Thus we examine the True Positive/Extent ratios to identify if the largest extents of different bioclimatic definitions were associates with True Positives. We use a bootstrap procedure to calculate confident intervals around the percentage of True Positives and then test whether or not estimated percentage of True Positive is statistically greater than with the other bioclimatic definition.

We now include these statistics and the reviewer’s point in the Discussion section as limitations and future research. We are open to identifying tropical rain forest plots using a similar search criteria as our tropical dry forest plots (e.g. plot locations within 1 km) if absolutely needed. However, it does not seem possible to do a confusion metric for vegetation types such as savannas, woodlands, and shrublands plots because all three vegetation types can occur in the tropical dry forest biome.

-Validation issue is important to be highlighted and there is lack of validation especially the ground validation.

This is an excellent point. The most important issue was the x, y of plot locations and we only selected plots for this study that had high resolution x, y data. However, there is a lack of ground validation for the FAO CHELSA biome and vegetation map and we now highlight this in the Discussion section. In theory we could use ICESAT or GEDI lidar data (covers 2% of the earth) to examine canopy heights to validate forests and compare results with FAO CHELSA forest cover maps or select a number of random points and undertake an assessment of forest cover and seasonality using very high resolution imagery in a similar fashion as Bastin et al 2017. However, this would take a bit more time.

-The authors need to explain with more details, why they used WorldClim and CHELSA Climatologies.

The reviewer is correct that this needs to be highlighted up front in the Introduction. We used WorldClim because it the most widely used climatic dataset in biogeography. We used CHELSA because it is a downscaled global circulation model and a novel climatic dataset. However, CHELSA is based on global circulation models thus they are easier to mechanistically explain the results and patterns and use for modeling future climate change than WorldClim. We now highlight this in the Introduction, Methods, and Discussion.

-More information about vegetation and plant cover is needed within study area.

Both reviewers raised this important point. We now provide data on estimates of percent canopy cover for all plots (mean, minimum, maximum forest cover at 1 km from Hansen’s 2000). This provides estimates of percent forest cover across the landscape. We also calculate > 10% and > 60% canopy cover for FAO CHELSA definition following Bastin et al. (2017) and Ratnam et al. (2016) as a comparison and provide a new AppendixS3 and further discussion.

-Variability in selected plots with different sites and validation is problematic, the authors failed to consider and state the variability between the study sites, plots and locations, so the authors need to show that statistically (i.e, residual plots).

We now provide variability in the percent canopy cover over plots (x,y) at landscape scale of 1 km in 2000. This provides estimates of landscape variability over plots.

-For the comparison within WWF Ecoregions and field plots, the authors need to show the accuracy level of the field plot location and within what range?

The accuracy of the WWF Ecoregions and field plots were assessed using the presence and absence of field plots within WWF Biome and Ecoregions polygons. We now include this information in Appendix S2 along with the WWF biome and ecoregion name. We are not sure how to estimate within what range and are open to ideas. Is this the distance from the TDF plots to the nearest WWF Tropical and Subtropical Broadleaf Forest?

-Table 5. is showing the best methods for estimating closed canopy (>40% close), what about open canopy?

In this manuscript, we chose to mainly focus on closed canopy forests (> 40%) based on a standard definition following the FAO, Miles et al. 2006, and Bastin et al. 2017. Open canopy was not used as a criteria for tropical dry forest plot selection. However, we calculated open forest for the reviewer, update results to 2020, and include this as a new Appendix S3.

-There is lack of information about tree species and density.

The reviewer is correct. We need to provide more information about tree species and density from the field plots. In the revisions of Appendix S2, we identify plot methods, area, and if they have standardize tree species and density data and include summaries in the Results section.

- Abbreviations in the text are confusing and need to be simple.

The reviewer is correct. We have kept abbreviations in wide use in the literature (FAO, TDF, WWF, CHELSA, PET) and reduced abbreviations (e.g. GDAL) that are not in common usage. We also added the full names over each Table in the revised manuscript. Abbreviations used less than three times were removed.

- The objectives are not clear and need to be more specific with some hypothesis and expectation for each objective.

The reviewer is correct. We now include specific expected results in the Problem Statement section.

- Discussion part needs to be more specific related to the results.

The reviewer is correct. On re-examination, the Discussion paragraph order were 1) conclusion, 2) Tropical dry forest cover and changes, 3) WWF ecoregions, 4) Bioclimatic definitions of tropical dry forest, 5) Climatic data sets, 6) Limitations, 7) Future research and applications, and 8) conclusions again. We have reorganized the Discussion to follow the presented research objectives and Results and now it is organized as 1) WWF ecoregions, 2) Bioclimatic definitions of tropical dry forest, 3) Climatic data sets, 4) Tropical dry forest cover and changes, 5) Limitations, 6) Future research and applications, and 7) Conclusions.

Other Comments:

Title: Add “using two climatic data set” :

Global tropical dry forest cover and extent: A comparative study of bioclimatic definitions using two climatic data sets.

Thank you this. The title has now been changed to more accurately reflect the research undertaken.

Introduction:

Line 63-65: need more details

Originally we stated. “Thus, climatic definitions allow delimiting the potential extent of biomes, but not detailed mapping of biome boundaries.” This was a mistake on our part.

We now say “Thus, climatic definitions allow delimiting the potential extent of biomes, but not detailed mapping of vegetation boundaries”.

Line 69-71: need citation.

We now provide two new citations.

Hansen AJ, Neilson RP, Dale VH, Flather CH, Iverson LR, Currie DJ, Shafer S, Cook R, Bartlein PJ. Global change in forests: responses of species, communities, and biomes: interactions between climate change and land use are projected to cause large shifts in biodiversity. BioScience. 2001. Sept; 51(9):765-779.

Schmitt CB, Burgess ND, Coad L, Belokurov A, Besançon C, Boisrobert L, Campbell A, Fish L, Gliddon D, Humphries K, Kapos V. Global analysis of the protection status of the world’s forests. Biological Conservation 2009. Oct; 142(10):2122-2130.

Materials and Methods:

- Data collection needs more clarification

The reviewer is correct. We really spent a great deal of time identifying publications and datasets that contained 1) high resolution locations data, defined as TDF by authors, and contain floristic composition and structural data (canopy cover). We now expand on this in the Methods section and provide further data in the appendix on plot method used (FIA, Gentry), plot size, DBH and if there is species composition and density data.

Reviewer #2: Review of Ocon et al. Definition of Tropical Dry Forests

Overall comments:

There have been endless debates about what constitutes a tropical dry forest (especially as opposed to potentially similar vegetation such as savannas). The problem is partly compounded by the fact that tropical dry forests and savannas are highly dynamic and lie at the boundary of sharp vegetation transitions in some regions. The understory may also change from predominantly grasses to woody plants within short time spans. Therefore, more objective criteria on defining tropical dry forests are certainly welcome. This study is a straightforward effort to try to delineate dry forests based on climatic criteria alone, which has its uses, particularly in large-scale analyses. I agree with the main conclusions of the study, namely, that the FAO definition of tropical dry forests seems to be the most appropriate. The much-cited Murphy and Lugo (1986) paper defines dry forests over an extreme range (250-2000 mm annually) of rainfall conditions.

However, I have several concerns regarding the methods.

Firstly, the authors do not account for the possibility that some of their validation plots may have been misclassified as dry forests. This is particularly true of some savanna regions (a lot has been written about this: see, for example, Ratnam et al., Proc. Roy. Soc. B, 2016). The authors should either account for this (for example, by checking sources for mention of % canopy cover and C4 grass cover) or discuss the limitation in detail.

The reviewer is correct. There is very little comparative data on C4 grass cover for all TDF plot sites however all plots contain estimates of tree canopy cover (e.g. closed canopy). There are a number of way to calculate % canopy cover to insure plots were not misclassified as dry forests and compare results among sites.

We now provide data on estimates of percent canopy cover over all plots (mean, minimum, maximum forest cover at 1 km from Hansen’s 2000). We also examined > 60 canopy cover for FAO CHELSA definition following Ratnam et al. (2016) as a comparison in an appendix (S3) and provide further discussion of the limitations.

Ratnam J, Tomlinson KW, Rasquinha DN, Sankaran M. Savannahs of Asia: antiquity, biogeography, and an uncertain future. Philosophical Transactions of the Royal Society B: Biological Sciences. 2016. Sept 19; 371(1703):20150305.

The authors must also discuss the possible circularity of some of the validation plot studies having adopted much-used definition such as Murphy & Lugo (1986) to classify their vegetation as dry forest.

This is an interesting point.  Indeed, the last author has collected a number of field plots in the Pacific loosely based on Murphy & Lugo (1986) climatic definition and an earlier version of the WorldClim data set.  However, all sites were selected based on local botanists identification of tropical dry forest.  However, we see your point.  Maybe other authors defined dry forest based on Murphy & Lugo (1986) definition in the past.  We now provide data on the number of plots that fell with both Murphy and Lugo and FAO definitions, only FAO, only Murphy and Lugo and plots that were outside both bioclimatic definitions.  There were 358 field plots (66%) that overlapped with both the FAO and Murphy and Lugo bioclimatic definitions (80 plots FAO only, 47 plots Murphy and Lugo only) and 54 field plots (10%), primarily in regions with long dry seasons (dry season > 9 months) or regions with over 2000 mm of mean annual precipitation such as Pacific islands, that did not overlap with either definition. 

On a related note, the 40% canopy-cover cutoff to define a closed-canopy forest (L88, L219) needs a clearer justification or needs to be contrasted with results that use an alternate cutoff. Several studies treat < 60% as savanna and >= 60% as closed-canopy forest (e.g. Murphy & Bowman, Ecol. Lett., 2012).

Both reviewers brought up this point. We used 40% canopy cover based on Miles et al. (2006) and Bastin et al. (2016) for comparison and we have now included this in the revisions.

Miles L, Newton AC, DeFries RS, Ravilious C, May I, Blyth S, Kapos V, Gordon JE. A global overview of the conservation status of tropical dry forests. Journal of Biogeography. 2006 Mar;33(3):491-505.

Bastin JF, Berrahmouni N, Grainger A, Maniatis D, Mollicone D, Moore R, Patriarca C, Picard N, Sparrow B, Abraham EM, Aloui K. The extent of forest in dryland biomes. Science. 2017 May 12;356(6338):635-8.

We also include data on > 60 canopy cover for FAO CHELSA definition following Murphy & Bowman (2012) and Ratnam et al. (2016) as a comparison in a new Appendix and update the data on forest cover to 2020 in the revised manuscript.

Murphy BP, Bowman DM. What controls the distribution of tropical forest and savanna?. Ecology Letters. 2012. March; 15(7):748-758.

Ratnam J, Tomlinson KW, Rasquinha DN, Sankaran M. Savannahs of Asia: antiquity, biogeography, and an uncertain future. Philosophical Transactions of the Royal Society B: Biological Sciences. 2016. Sept 19; 371(1703):20150305.

Secondly, any study like this must assess results using a confusion matrix (proportions of false positives, true positives, false negatives, and true negatives). By only trying to maximize true positives and minimize false negatives, their criteria become overly liberal, potentially incorrectly including other vegetation types such as moist forest, savanna, shrublands, etc.

Both reviewers brought up this point. However, we do not have absence data for non-TDF plots (e.g. tropical rain forest). For instance, we do not have comparative data on high resolution locations of savannas or moist or mesic forests and this is especially true at a global spatial scale. We examined a number of global studies such as Slik et al. and Ibanez et al. but the resolution of their location data was generally 1 degree latitude which is very coarse. Furthermore, it is possible to have savanna, woodlands, and shrublands within 1 km pixel resolution of tropical dry forest.

Thus we try to maximize true positives. We examine the True Positive/Extent ratios to identify if the largest extents of different bioclimatic definitions were associates with True Positives. We use a bootstrap procedure to calculate confident intervals around the percentage of True Positives and then test whether or not estimated percentage of True Positive is statistically greater than with the other bioclimatic definition.

We are not opposed to identifying about 500 globally distributed tropical rain forest plots with 1 km from Dryad or BIEN but this would take more time.

Finally, as they correctly note, the validation plot data set is highly biased in its geographical representation. While partly reflective of differences in research effort, it is also partly because their search criteria are inadequate as they miss several studies from South Asia and Indo-China (which host large extents of tropical dry forest) that should have been included.

All methods for identifying published plots defined as “tropical dry forest” with high resolution x, y locations were undertaken in a standard and repeatable manner during the literature search. We did identify a number of vegetation surveys and plot data in South Asia and Indo-China but we could not use them because they lack x,y locations to within 1 km and many lacked closed canopy. Please let us know of any references we might of missed in South Asia and Indo-China. However, we now state that some plots in South Asia and Indo-China could be missed because they may be defined as ‘monsoon forest’. We now mention this in the discussion.

Specific comments:

Line 52: should be “complemented” (not “complimented”)

This has been corrected in the revised manuscript.

Line 55-59: Seems out of place in the first paragraph of the Introduction. These do not provide any insight into the research questions but merely state the type of climate data sets available. I suggest these are shifted to the Methods section or at least reduced to a single sentence. In fact, I would go further and suggest that lines 72-108 are brought up front in the Introduction (after line 54) in order to clearly articulate the purpose of this study. The authors can then mention the availability of global data sets on climate, vegetation, land-cover and so on which could be used to resolve the questions.

The reviewer is correct and we have reorganized the Introduction. Thank you.

Line 299: word missing in the sentence. In fact, there is a need for careful copyediting of the manuscript to correct minor errors.

This has been correct in the revised manuscript along with a number of minor errors which we now highlight in Blue text.

Line 411-412: These categories are not mutually exclusive: C4 grasses can exist under a range of canopy covers.

The reviewer is correct. We now identify grasses and structure as the structure (< 1 m) as the best way to identify the savanna vegetation type and provide a citation.

Alencar A, Shimbo JZ, Lenti F, Balzani Marques C, Zimbres B, Rosa M, Arruda V, Castro I, Fernandes Marcico Ribeiro JP, Varela V, Alencar I. Mapping three decades of changes in the brazilian savanna native vegetation using landsat data processed in the google earth engine platform. Remote Sensing 2020 March;12(6):924.

Attachment

Submitted filename: Response to Reviewers .docx

Click here for additional data file. (18.7KB, docx)

Decision Letter 1

RunGuo Zang

10 May 2021

Global tropical dry forest extent and cover: A comparative study of bioclimatic definitions using two climatic data sets

PONE-D-20-37820R1

Dear Dr. Ocón,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

RunGuo Zang

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Acceptance letter

RunGuo Zang

12 May 2021

PONE-D-20-37820R1

Global tropical dry forest extent and cover: A comparative study of bioclimatic definitions using two climatic data sets

Dear Dr. Ocón:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Professor RunGuo Zang

Academic Editor

PLOS ONE

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    S1 Appendix. Tropical and subtropical dry broadleaf forest defined by World Wildlife Fund.

    Name of ecoregion, extent, and forest cover.

    (CSV)

    Click here for additional data file. (35.5KB, csv)
    S2 Appendix. Tropical dry forest field plots.

    Comprehensive list of the 540 verified tropical dry forest plots, their coordinates, references, and bioclimatic characteristics.

    (CSV)

    Click here for additional data file. (116.5KB, csv)
    S3 Appendix. Forest cover and extent statistics.

    Forest cover statistics for 540 tropical dry forest plots using resampled Global Forest Watch tree canopy cover (1 km2) data for the year 2000 and a 1 km buffer around each plot. Total forest cover estimates and gross forest cover loss are also calculated for open canopy (≥ 10%) and closed canopy (≥ 40% and ≥ 60%) for FAO CHELSA from 2000 to 2020.

    (DOCX)

    Click here for additional data file. (1.6MB, docx)
    S4 Appendix. Biodiversity hotspots for mesoscale analyses of tropical dry forest.

    Extent and agreement of tropical dry forest area.

    (CSV)

    Click here for additional data file. (2.1KB, csv)
    S5 Appendix. Tropical dry forest extent at country level.

    Summaries of tropical dry forest extent based on climatic definitions and datasets at the country level.

    (CSV)

    Click here for additional data file. (6.5KB, csv)
    S6 Appendix. Computer code for global analysis of tropical dry forest.

    Code used to accomplish a global analysis of tropical dry forest extent and cover based on a suite of remote sensing and modeled climate and forest data, as well as GIS layers of political and conservation boundaries.

    (PDF)

    Click here for additional data file. (315.5KB, pdf)
    Attachment

    Submitted filename: Reviewer-1 comments.pdf

    Click here for additional data file. (99.2KB, pdf)
    Attachment

    Submitted filename: Response to Reviewers .docx

    Click here for additional data file. (18.7KB, docx)

    Data Availability Statement

    All relevant data are within the manuscript and its Supporting Information files.

    Articles from PLoS ONE are provided here courtesy of PLOS

    RESOURCES