Revista de Biología Tropical, ISSN: 2215-2075, Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

Biogeographical analysis of the Central American clade

of Sechium (Cucurbitaceae)

Luis Ángel Barrera-Guzmán1,2; https://orcid.org/0000-0001-8057-2583

Jorge Cadena-Iñiguez2,3*; https://orcid.org/0000-0002-6427-0646

Juan Porfirio Legaria-Solano4; https://orcid.org/0000-0002-1371-9482

Víctor Manuel Cisneros-Solano1,2; https://orcid.org/0000-0001-8262-9109

Kazuo, N. Watanabe5; https://orcid.org/0000-0003-1499-9989

Daniel Alejandro Cadena-Zamudio2; https://orcid.org/0000-0002-6972-7414

1. Universidad Autónoma Chapingo, Centro Académico Regional sede Huatusco-Veracruz, México C.P. 94100,

Carretera-Federal Huatusco-Xalapa km 6.5; luisangelbg@gmail.com, vcisneross@chapingo.mx

2. Grupo Interdisciplinario de Investigación en Sechium edule en México (GISeM), Texcoco, C.P. 56 160, México; jocade-

na@gmail.com (*Correspondencia), cadenazamudio@gmail.com

3. Colegio de Postgraduados, Campus San Luis Potosí, 78 622, Salinas de Hidalgo, San Luis Potosí, México; jocadena@

gmail.com

4. Universidad Autónoma Chapingo, Carretera México-Texcoco, C.P. 56 230, Chapingo, Estado de México; legarias.

juan@yahoo.com

5. Tsukuba Plant Innovation Research Center, University of Tsukuba, 1-1-1 Tennodai, Tsukuba, Ibaraki Prefecture, 305-8

571, Japan; nabechanknw@gmail.com

Received 09-IX-2024. Corrected 04-XII-2024. Accepted 02-IX-2025.

ABSTRACT

Introduction: The genus Sechium P. Brown (Cucurbitaceae) includes 11 species, of which two are domesticated

and nine grow in the wild. The Central American clade of Sechium has six species distributed in Panama and

Costa Rica. These species have characteristics that can be transferred from wild to domesticated species.

Objective: To use three machine learning stacking algorithms and multivariate tools to describe geographic

distribution, diversity degree, and endemism, to identify major conservation areas and to promote research for

the improvement of the domesticated species.

Methods: Two hundred and nine occurrence records were retrieved from the Global Biodiversity Information

Facility. Raster values extracted from 21 bioclimatic variables were analyzed with descriptive and multivariate

statistics. The species distribution algorithms were assembled with the SSDM library from R software.

Results: Most species are distributed in type A and C climates, mainly in volcanic soils, with abundant organic

matter. These species can grow at altitudes exceeding 2 000 m and tolerate low temperatures and high humidity

levels. K-medoids established two groups and a 0.39 average silhouette coefficient, which indicates a low cluster-

ing trend. The stacked distribution models recorded good performance in areas under the curve (AUC) (> 0.75)

and true skill statistic (> 0.75).

Conclusions: The main variables that supported the models were elevation, soil types, and precipitation. The

main endemism and species diversity areas were in the Cordillera de Talamanca, the Cordillera de Guanacaste,

the Cordillera de Tilarán, and the Central Volcanic Range (Costa Rica). These species thrive under similar envi-

ronmental conditions; however, the diverse areas have significantly different precipitation and soil types.

Key words: domesticated species; machine learning; diversity; endemism; soil types.

https://doi.org/10.15517/sy4jvh88

CONSERVATION

2Revista de Biología Tropical, ISSN: 2215-2075 Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

INTRODUCTION

The genus Sechium P. Brown includes 11

species and five of these are distributed in

Mexico: S. edule (Jacq.) Sw., S. compositum

(Donn. Sm.) C. Jeffrey, S. chinantlense Lira &

F. Chiang, S. hintonii (P.G. Wilson) C. Jeffrey,

and S. mexicanum Lira & M. Nee. This study

focuses on the remaining six species, which are

mainly distributed in Costa Rica and Panama:

S. tacaco (Pittier) C. Jeffrey, S. venosum (L.D.

Gómez) Lira & F. Chiang, S. villosum (Wun-

derlin) C. Jeffrey, S. pittieri (Cogn.) C. Jeffrey,

S. panamense (Wunderlin) Lira & F. Chiang,

and S. talamancense (Wunderlin) C. Jeffrey.

According to their morphological, geographic,

and molecular characteristics, these species are

divided into two groups: Mexican and Central

American clades (Barrera-Guzmán et al., 2021;

Cross et al., 2006; Lira et al., 1997; Lira & Nee,

1994; Monge & Loría, 2017; Sebastian et al.,

2012; Wunderlin, 1976). Lira (1995) and other

authors have provided valuable morphological

information about the Central American clade

of Sechium and its possible contribution to S.

edule and S. tacaco, the only two species of this

genus that have been domesticated. Neverthe-

less, the biogeographical information about

the Central American species is scarce. Geo-

graphic Information Systems (GIS) currently

provide data on the climate and ecological

variables of a given territory. These tools, along

with multivariate analyses, are essential for

such studies about conservation, localization,

potential distribution models, and endemism

(Mateo et al., 2011).

The species distribution models (SDM) are

based on different machine learning algorithms

(Schmitt et al., 2017). One of the most used

algorithms is MaxEnt or maximum entropy

(Phillips, 2010); however, MaxEnt has some

deficiencies, because the bias in sampling or

the spatial resolution can provide arbitrary

results. Choosing the appropriate configuration

RESUMEN

Análisis biogeográfico del clado centroamericano de Sechium (Cucurbitaceae)

Introducción: El género Sechium P. Brown (Cucurbitaceae) incluye 11 especies, de las cuales dos son domestica-

das y el resto son silvestres. El clado centroamericano de Sechium incluye seis especies distribuidas en Panamá y

Costa Rica. Estas especies tienen características que pueden ser de utilidad y transferibles de especies silvestres a

especies modificadas.

Objetivo: Ejecutar tres algoritmos de aprendizaje automático apilados y herramientas multivariadas para des-

cribir la distribución geográfica, medir el grado de diversidad y endemismos de las especies centroamericanas

de Sechium para identificar áreas de conservación y promover la investigación para el mejoramiento de especies

cultivadas del género.

Métodos: Doscientos nueve puntos de ocurrencia fueron extraídos de la Global Biodiversity Information Facility.

Los valores ráster se obtuvieron a partir de 21 variables bioclimáticas, donde se analizaron con estadística des-

criptiva y multivariada. Los modelos de distribución de especies apiladas fueron ejecutados con la librería SSDM

del software R.

Resultados: La mayoría de las especies se distribuyen en climas tipo A y C, principalmente en suelos volcánicos

con abundante materia orgánica. Estas especies prosperan en altitudes superiores a los 2 000 m y toleran bajas

temperaturas con altos índices de humedad. El análisis k-medoides estableció dos grupos con un coeficiente de la

silueta de 0.39, el cual indica una baja tendencia al agrupamiento. Los modelos de distribución apilados tuvieron

buenos rendimientos en términos de área bajo la curva y del estadístico de habilidad verdadera (> 0.75).

Conclusiones: Las principales variables que apoyaron los modelos fueron la elevación, los tipos de suelo y la pre-

cipitación. Las principales áreas de endemismo y diversidad de especies se ubican en la Cordillera de Talamanca,

la Cordillera de Guanacaste, la Cordillera de Tilarán y la Cordillera Volcánica Central (Costa Rica). Estas especies

prosperan en condiciones ambientales similares; sin embargo, las diferentes áreas tienen precipitaciones y tipos

de suelo significativamente diferentes.

Palabras clave: especies domesticadas; aprendizaje automático; diversidad; endemismo; tipos de suelo.

Revista de Biología Tropical, ISSN: 2215-2075, Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

requires advanced knowledge, hindering the use

of qualitative variables for evaluation purposes

(Baldwin, 2009). Stacked SDMs involve compil-

ing fundamental information from each algo-

rithm and subsequently merging it. Although

the output raster layer provides strong results,

they must be assessed with the area under the

curve (AUC-ROC), Cohen’s Kappa calculator,

and the True Skill Statistic (TSS).

The stacked SDMs are fundamental to

determine the biogeographical and ecologi-

cal characteristics of a species. For instance,

they provide information about the species’

prevalence, diversity, and degree of endemism

(Gelfand, 2022). They also help to establish

environmental variables influencing the taxa

distribution (Bedair et al., 2023). Other out-

standing applications include development of

climate change models for spatial biodiver-

sity predictions that support forest restoration

(Zwiener & Alves, 2023); research about the

climate change-related phenological variation

of plants (Bayliss et al., 2022); and biodiversity

patterns and assembly processes of communi-

ties (Dubuis et al., 2011). Although reviews

prefer individual models (such as MaxEnt), a

positive trend towards research and evaluations

about stacked machine learning models has

arisen (Qazi et al., 2022).

The Sechium species of the Central Ameri-

can clade are distributed in the mountains of

Panama and Costa Rica (Fig. 1); however, the

limited number of samples collected restricts

the morphologic, agronomic, and ecological

study into the species. The evaluation of SDMs

requires a certain number of matches (Schmitt

et al., 2017). Given the rarity of some Central

American Sechium species, the hypothesis is

that all the species are distributed under the

same environmental conditions and, conse-

quently, independent studies about their distri-

bution are not required. However, the stacked

SDMs provide a wider perspective about each

species. In addition, their results are more con-

sistent regarding the Sechium species commu-

nities. Therefore, the objective of this research

was to use stacked machine learning algorithms

(Random Forest (RF), Support Vector Machines

(SVM), and Classification Tree Analysis (CTA)

and multivariate tools (clustering and main

Fig. 1. Geographic distribution of the Central American Sechium (Cucurbitaceae) species in Costa Rica and Panama.

Data points from specimens retrieved from the Global Biodiversity Information Facility website.

4Revista de Biología Tropical, ISSN: 2215-2075 Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

components) to determine the geographical

distribution, diversity degree, and endemism of

the Central American Sechium species in Costa

Rica and Panama. This research also aimed to

expand databases, identify key conservation

areas, and promote the genetic improvement of

domesticated species.

MATERIALS AND METHODS

Points of occurrence: Two-hundred and

nine points of occurrence were retrieved from

the Global Biodiversity Information Facility

(GBIF, 2024) for the Sechium species of the Cen-

tral American clade: 147 of S. pittieri, 16 of S.

tacaco, 15 of S. venosum, 13 of S. talamancense,

9 of S. panamense, and 9 of S. villosum. Nine-

teen WorldClim bioclimatic variables (Fick &

Hijmans, 2017) were used in raster format, with

a 30 arcsecond (~1 km2) resolution. The raster

formats of the Köppen climate classification

(Beck, 2012), the Harmonized World Soil Data-

base v 1.2 (Fischer et al., 2002), and the eleva-

tion data (~1 km2 resolution) (Fick & Hijmans,

2017) were added as well. The raster values

were obtained with the QGIS Point Sampling

Tool version 3.20.1 (QGIS Development Team,

2020) and were exported to a spreadsheet.

Statistical and multivariate analysis: The

raster values of the 20 quantitative (19 World-

Clim variables and elevation) and qualitative

(weather and soil) variables were analyzed with

the descriptive statistics of the RStudio software

(R Core Team, 2020). The aim was to obtain

temperature and humidity referential values,

using the package ggstatsplot (Patil, 2021) and

package psych (Revelle, 2020). Given that the

quantitative variables did not comply with the

assumptions of the analysis of variance, the

Kruskal-Wallis test was used to determine the

differences between species. If significant dif-

ferences were detected, Dunn’s post hoc test

was applied. In addition, the Support Vector

Machine classification method was applied,

using the Kernel function, to classify any habi-

tat types found in the area (Meyer et al., 2019).

The package clustertend (Yilan & Rutong, 2015)

was used to calculate the Hopkins (H) statistics,

to verify the clustering trend of the raster values

of the 20 quantitative variables. The package

clValid was used to determine the selection of

the (hierarchical and non-hierarchical) cluster-

ing algorithm (Brock et al., 2008). This proce-

dure was used to determine a non-hierarchical

k-medoids clustering, to find similar edapho-

climatic conditions, using package factoextra

(Kassambara, 2017) and package FactoMineR

(Lê et al., 2008). The NbClust package (Charrad

et al., 2014) enables a consensus of up to 30

clustering indexes and was used to calculate

the number of optimal clusters. All the above-

mentioned packages were executed with the

RStudio software (R Core Team, 2020).

Algorithms used in the stacked model-

ing: To develop the distribution models, a

Pearson correlation analysis was carried out to

eliminate the R < -0.90 and R > 0.90 variables,

to prevent collinearity, and to prevent compro-

mising the effectiveness of the models (Phillips

et al., 2010). Meanwhile, qualitative variables,

such as climate and soil, were added. The Ran-

dom Forest (RF) algorithms (Breiman, 2001)

were used to model the presence or absence of

species in areas where sampling was not carried

out, based on the occurrence records. In addi-

tion, the classification tree analysis algorithm

(CTA) was used to identify priority areas,

given its capacity to predict the presence of

endangered species or species with limited

distribution (Schmitt et al., 2017). Finally, the

support vector machine (SVM) algorithm was

chosen because of its high accuracy and capac-

ity to manage non-linear data (Vapnik, 1998).

When outliners are in place, they have a high

prediction level and provide strong results. The

three algorithms were executed and assembled

with the stacked species distribution models

(SSDM) (Schmitt et al., 2017), which can gen-

erate random data about pseudo-absences in

areas where species are absent. In addition, they

can calculate the diversity of the species and the

weighted endemism index (WEI). Twenty-five

percent of the data were used in the test, while

the remaining 75 % were used for training

Revista de Biología Tropical, ISSN: 2215-2075, Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

objectives (Hijmans & Elith, 2013). The aim

was to eliminate the spatial classification bias

studied by Lobo et al. (2007).

To evaluate the effectiveness of the mod-

els, some indicators such as the rate of change

indicator (ROC) and the total severity score

(TSS) were used. ROC values higher than

0.75 indicate a good performance, while TSS

values higher than 0.75 indicates excellent per-

formance. Meanwhile, 0.4 < TSS < 0.75 indi-

cates an optimum performance and TSS < 0.4

indicates a low performance. Cohen’s Kappa

coefficient accurately quantifies the predicted

points after the elimination of the random

match probability (RMP). In this study, its value

fluctuated between -1 and 1: values closer to 1

indicate an excellent performance of the model,

while values closer to -1 show a poor perfor-

mance. The TSS was calculated (specificity +

sensitivity) -1) to adjust for the dependence of

Cohen’s Kappa on prevalence. TSS optimizes

and corrects the problems of Kappa, consider-

ing the performance criteria. In addition, the

TSS and Kappa share the same performance

criteria (Allouche et al., 2006).

Meanwhile, the Jackknife test was used to

observe the contribution of each environmental

variable to the distribution models, along with

ROC, Cohen’s Kappa, TSS, and the percentage

of correct predictions (PCP). The assembly of

the three algorithms was calculated with the

package SSDM (Schmitt et al., 2017). The raster

format resulting from the assembly of the three

models (diversity and endemism maps) were

exported with the raster package (Hijmans,

2020), to appropriately edit and visualize them

in the QGIS Point Sampling Tool version 3.16.2

(QGIS Development Team, 2020).

RESULTS

Climate and soil diversity: The predomi-

nant climate was tropical wet (Af). This type

of climate can be found in 29.2 % of the 209

points of occurrence, followed by Cfb (27.3 %),

Am (20.6 %), Cwb (18.7 %), Aw (3.83 %), and

Cwc (0.47 %). The last climate type was only

reported for a S. pittieri specimen. Most Central

American Sechium species can be found in

tropical (A) and temperate (C) climates, with

their respective variations (Fig. 2). S. talaman-

cense, S. tacaco, and S. pittieri were found in

most occurrences’ records and, consequently,

covered different climates. The S. villosum and

S. panamense specimens can only be found in

andosols. The S. pittieri specimens can be found

in cambisols (38.1 %), andosols (25.9 %), areno-

sols (18.4 %), nitisols (7.48 %), kastanozems

(5.44 %), regosols (4.08 %), and leptosols (0.68

%) (Fig. 3).

S. panamense: This species is native to

Chiriquí, Panama, where it grows in elevations

that fluctuate between 1 500 and 3 000 m.a.s.l.

(Lira, 1995). Its botanical description is based

merely on six collected samples. Consequently,

registering more individuals and populations

is fundamental to find variations and possible

hybridizations that could drive their study and

exploitation. As a result of the elevation in

which it thrives, this species could be used in

genetic improvement programs. Regarding the

points of occurrence of this study, S. panamense

grows at a mean annual temperature of 19 ºC;

however, it can thrive in lower (10 ºC mini-

mum) and higher (31 ºC maximum) tempera-

tures and tolerates precipitations of 2 600-3 000

mm. It is typically found in Am, Af, and Cfb cli-

mate types (Fig. 2). S. panamense can perfectly

adapt to volcanic and permeable andosols, with

a light layer of organic matter (Fig. 3). The

elevation interval of this species is 1 300-2 000

m.a.s.l.; however, Lira (1995) reported that it

can expand its limits to 3 000 m.a.s.l.

S. pittieri: This species is rarely found in

Nicaragua; however, it prevails in Costa Rica,

where it is consumed as a vegetable. As a result

of its phenotypic and genetic plasticity, it can

adapt to different environments, with a wide

range of elevations (up to 2 500 m.a.s.l.). This

phenomenon could indicate that S. pittieri has

a wide genetic diversity and could be used

to improve domesticated species, just like S.

talamancense and S. panamense (Lira, 1995).

According to this research, this species grows

6Revista de Biología Tropical, ISSN: 2215-2075 Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

and develops at a mean annual temperature

of 19 ºC; however, it can thrive at lower (2 ºC

minimum) and higher (32 ºC maximum) tem-

peratures. In addition, it can adapt to areas

with average precipitations of 2 600-3 200 mm,

although one specimen was found in an area

with 4 000 mm annual precipitation. Its cli-

matic adaptability range allows its establish-

ment in the six climate types (Af, Am, Aw, Cfb,

Cwb, and Cwc) (Fig. 2) and seven soil types

(Fig. 3) registered in this study. Its elevation

interval is 900-2 900 m.a.s.l., which matches

Fig. 2. Climate type percentages of the points of occurrence of Sechium (Cucurbitaceae) species from Costa Rica and Panama..

Cwc = Temperate, dry winter, and cold summer; Cwb = Temperate, dry winter, and warm summer; Cfb = Temperate, without

dry season, and warm summer; Aw = Tropical savannah; Am = Tropical monsoon; Af = Tropical rainforest. Climate Group

A: tropical, no month with average temperatures below 18 °C, precipitation is still higher than evaporation. Climate Group C:

the average temperature of the coldest month is between -3 °C and 18 °C and that of the warmest month exceeds 10 °C, these

climates are in temperate forests.

Fig. 3. Soil type ratio in the points of occurrence of Sechium (Cucurbitaceae) species from Costa Rica and Panama.

Revista de Biología Tropical, ISSN: 2215-2075, Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

the findings of Lira (1995). However, S. pittieri

was also found at an elevation of 16 m.a.s.l.

The adaptability of this species suggests a wide

genetic variability that has not been explored

and that should be studied with morphologic

and molecular markers.

S. tacaco: This species can be found in the

Cordillera de Talamanca, Costa Rica, in eleva-

tions of 1 000-1 700 m.a.s.l. It is locally known

as tacaco (Wunderlin, 1976). Domesticated and

semi-domesticated populations of this species

can be found in San José, Costa Rica, where its

fruits are used in regional dishes (Lira, 1995).

Morales (1994) studied the plant and repro-

ductive morphology of S. tacaco and found

similarities with the anatomical structure of the

organs, multicellular trichomes, floral nectary

structures, and anomocytic stomata in the plant

and reproductive organs of S. edule. Lira (1995)

emphasized the low morphological diversity

of the S. tacaco fruits, except for the presence/

absence of thorns and the amount of fiber.

However, Monge and Loría (2017) described

the morphology of five S. tacaco populations

from distant localities with a similar aver-

age elevation (1 100 m.a.s.l.) and observed

diversity and significant differences regarding

the weight, length, and width of its fruit. The

most important findings of Monge and Loría

(2017) were fruits with 6-7 complete longitu-

dinal sutures and 2-5 incomplete longitudinal

sutures; this information might help to clarify

the evolution processes of the species and the

genus Sechium. According to the ecological

data of this study, the mean annual temperature

requirement of this species is 18 ºC; however, it

can tolerate lower (9 ºC minimum) and higher

(30 ºC maximum) temperatures. It grows in

areas with 2 200-2 800 mm precipitations, and

its elevation interval is 1 400-2 000 m.a.s.l. The

occurrence records match tropical (Aw, Af, and

Am) and warm (Cwb and Cfb) areas (Fig. 2).

Although it can adapt to a wide range of soils, it

thrives under crop conditions in andosols and

nitisols, because of its high structural stability,

depth, and low base saturation (Fig. 3).

S. talamancense: Also known as chayotillo

and tacaquillo, this species (just like S. tacaco)

is endemic to the Cordillera de Talamanca,

Costa Rica. It can be found in the cloud forest,

at an elevation of 2 400-3 200 m.a.s.l. Because

of the low temperatures of the area where it

thrives, S. talamancense could be used for gene

transfer and to improve the resistance to frost

of the domesticated species (Lira, 1995). Few

specimens have been kept in herbaria. The

species requires a mean annual temperature

of 20 ºC; nevertheless, it can tolerate lower (9

ºC minimum) and higher (30 ºC maximum)

temperatures. This species grows under humid

conditions, with a precipitation of 3 300-3 800

mm and at an elevation interval of 500-2 200

m.a.s.l. However, it has been found at an eleva-

tion of up to 3 200 m.a.s.l. (Lira, 1995). S. tala-

mancense can be found in Af, Am, Aw, Cfb, and

Cwb climates. As a result of its water require-

ments, it prefers arenosols, but it can also devel-

op in cambisols and andosols (Fig. 2, Fig. 3).

S. venosum: This species is endemic to the

Caribbean Coast of Costa Rica and its pendu-

lar inflorescence is very similar to S. hintonii.

It adapts to high-humidity conditions (Lira,

1995), suggesting that its genetic structure is

resistant to phytopathogens, which represent a

significant issue for domesticated Sechium spe-

cies (Olguín-Hernández et al., 2013). Conse-

quently, it is one of the main foci of the genetic

improvement of the agriculturally important

crops (Newstrom, 1990). Like the rest of the

Central American species, there are few S.

venosum specimens in herbaria. It requires a

mean annual temperature of 17 ºC but can tol-

erate lower (6 ºC minimum) and higher (30 ºC

maximum) temperatures. Like S. talamancense,

it can thrive in high-humidity environments,

with annual precipitations of 2 600-3 700 mm.

Its elevation interval fluctuates between 1 100

and 2 500 m.a.s.l. This species can adapt to

Cfb, Am, and Af climates and prefers andosols,

arenosols, and cambisols (Fig. 2, Fig. 3).

S. villosum: This species in endemic to

Costa Rica and thrives in disturbed tropical or

8Revista de Biología Tropical, ISSN: 2215-2075 Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

cloud forest environments, at an elevation of

1 500-2 000 m.a.s.l. Like S. venosum, it devel-

ops in humid environments, which could be

a source of gene resistance against fungal dis-

eases (Lira, 1995). S. villosum requires a tem-

perature of 19 ºC; however, it tolerates lower

(6 ºC minimum) and higher (29 ºC maximum)

temperatures. Like all the other species of

the clade, it develops in humid environments,

with precipitations of 2 900-3 500 mm, and its

elevation interval fluctuates between 800 and

900 m.a.s.l.; however, it can be found above

2 700 m.a.s.l. Lira (1995) reported elevation

intervals of 1 500-2 000 m.a.s.l. for this species.

The occurrence records match the Cfb and

Af climates, with a predominance of andosols

(Fig. 2, Fig. 3).

Multivariate analysis: Two clusters

(k = 2) were created using the non-hierarchical

k-medoids and the 209 points of occurrence.

The NbClust package was used to calculate the

optimum number. The H coefficient was 0.04,

which is lower than the threshold (0.5); conse-

quently, the data can be subjected to a cluster-

ing analysis, despite their low trend. The two

first dimensions accounted for 59.6 % of the

total variation (Fig. 4A). Dimension 1 provided

35.1 % of the total variation and was integrated

by the elevation variables bio1, bio15, bio4, and

bio2 (Fig. 4C).

Stacked modeling: The RF, CTA, and

SVM algorithms recorded good performances:

AUC > 0.85, Kappa > 0.75, and TSS > 0.75

(Fig. 5A). The Jackknife test showed that the

soil variable made the greatest contribution to

the Sechium species and is responsible for its

distribution (Fig. 5B). The three algorithms

showed a significant Pearson correlation (aver-

age: 0.85). This result indicated that the indi-

vidual models had similar percentages (86 %)

of correct predictions.

The assembly of the RC, CTA, and SVM

algorithms —used to quantify the diversity

of the species— indicated that the greatest

Fig. 4. Multivariate analysis for Sechium species: A. K-medoids cluster plot. B. Percentages of the points of occurrences of

Sechium species in the clusters. C. Contribution of the variables to the components 1 and 2. D. Cluster silhouette plot (average

silhouette width: 0.39).

Revista de Biología Tropical, ISSN: 2215-2075, Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

diversity of Sechium can mainly be found in the

central mountain region of Costa Rica (Limón,

Cartago, Puntarenas, San José Heredia, and

Alajuela) and Panama (Bocas del Río, Chiriquí,

Bocas de Toro). This phenomenon could be

associated with factors including elevation,

temperature, and humidity (Fig. 6). Although

the red areas on the maps show the greatest

Fig. 6. Distribution of species diversity of Sechium (Cucurbitaceae) in Central America (Costa Rica and Panama).

Fig. 5. A. Algorithms evaluated with AUC (Y axis) (> 0.75) and subsequently weighed with the previous metric means and

B. Jackknife test for variable relative contribution to the algorithms. bio1 = mean annual temperature; bio2 = mean diurnal

range (monthly mean (max temperature - min temperature)); bio3 = isothermality (bio2/bio7) (×100); bio4 = temperature

seasonality (standard deviation ×100); bio12 = annual precipitation; bio13 = wettest month precipitation rate; bio15 =

precipitation seasonality (coefficient of variation); bio16 = wettest quarter precipitation rate; bio18 = warmest quarter

precipitation rate; bio19 = coldest quarter precipitation rate; clms = climates; elvt= elevation.

10 Revista de Biología Tropical, ISSN: 2215-2075 Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

diversity, the number of species is closer to four,

which indicates low diversity levels. The areas

with less diversity are the coasts and most of the

Panamanian territory.

Fig. 7 shows the endemism percentages,

which are very similar to the diversity results

recorded in Fig. 6. Endemism is measured as

a proportion of the total species within a given

area. It shows the percentage of the local spe-

cies that can be endangered if their habitats are

destroyed or threatened (Shipley & McGuire,

2022). The greatest concentration of endemism

in Costa Rica can be found in the central and

mountain area, while in Panama they are more

scattered and less concentrated. Therefore, the

Cordillera de Talamanca is a notable region due

to its high concentration of endemism.

DISCUSSION

The predominant climates of the occur-

rence records were type A and C, this is because

Sechium species have a wide phenotypic plas-

ticity and can adapt to different environments;

on the other hand, mountainous areas favor

a rich variety of climates where species can

diversify in morphological, genetic and physi-

ological aspects. Andosol is the predominant

soil for the Central American Sechium species.

This type of soil has a high organic matter

content, a volcanic origin, and a high-water

retention and cation exchange capacity. The

SVM classification analysis and the Kruskal-

Wallis test failed to divide the species into

groups or significantly different means. In par-

ticular, the Kruskal-Wallis test was only able to

differentiate the Sechium species into precip-

itation related variables (bio13: wettest precipi-

tation month; bio15: precipitation seasonality;

and bio6: wettest quarter precipitation). Most

occurrence records showed low consistency

levels, suggesting that all of them can cluster in

a single group; however, the NbClust package

can only calculate the parameters based on a

minimum of two clusters. There are no well-

defined groups of the species. Their distribu-

tion is limited to Costa Rica and Panama and,

consequently, their edaphoclimatic conditions

are very similar and they probably interact in

the same ecological niches.

Fig. 7. Map of the endemism of Central American Sechium species in Costa Rica and Panama. The caption shows endemism

percentages.

Revista de Biología Tropical, ISSN: 2215-2075, Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

Most of the occurrence records of the spe-

cies are distributed in different clusters —i.e.,

one cluster can include several points of occur-

rence of distinct species. They occupy similar

ecological niches, due to their limited distribu-

tion in Costa Rica and Panama. For instance,

65 % and 35 % of the 147 occurrence records of

S. pittieri were included in cluster 1 and cluster

2, respectively (Fig. 4B). The silhouette coef-

ficients of the two clusters were 0.42 and 0.38

(Fig. 4D), respectively, which is lower than the

threshold (0.5), suggesting that all the occur-

rence records could be grouped in a single

cluster. The high elevation and precipitation

fluctuations resulted in a multivariate sensitiv-

ity analysis. In addition, soil and climate types

play a major role in the distribution of the spe-

cies (Fig. 5B). The high diversity levels in the

mountains could be influenced by the variety

of microclimates and the available habitats.

This situation is different in low diversity areas,

where factors such as urbanization or a greater

agricultural activity could reduce the possibili-

ties of finding Sechium species.

On the one hand, the mountains of Costa

Rica, such as the Cordillera de Talamanca, Cor-

dillera de Guanacaste, Cordillera de Tilarán,

and the Central Volcanic Range, usually have

microclimates and specific temperature, light,

soil diversity, and humidity conditions that

favor the development of Sechium and other

species. On the other hand, the strong geo-

graphical isolation is also a major endemism

factor in this area (Noroozi et al., 2018; Peñas

et al., 2005). Mountain ranges are also sensitive

to climate change (La Sorte & Jetz, 2010), which

impacts several major bioclimatic variables,

such as mean daily temperature, seasonality,

and precipitation (Chang et al., 2024). This

situation can trigger phenological and apti-

tude changes that damage the diversity of the

species (Munson & Sher, 2015). In addition,

factors such as agricultural expansion, new

crops, deculturization, urban developments,

recreation, and tourism impact endemism areas

(Wani et al., 2023).

Wild populations of S. edule can be found

in Mexico, particularly in Veracruz. They are

distributed in the cloud forest, where the spe-

cific humidity and shade conditions allow their

development. The fruits of these species have

different morphological characteristics (Vil-

lanueva-Jiménez, 2012). Likewise, the cloud

forest of the Costa Rican Mountain ranges is

the characteristic habitat of Sechium and other

species. Their association with other plants

enables most of these species to climb tall trees.

Regarding endemism and diversity, in some

cases, there are few points of occurrence; this is

a key indicator that limits the perception of the

actual distribution of the species under study.

The multivariate analysis (k-medoids and

principal components) showed a slight cluster-

ing trend of the species, suggesting that Central

American Sechium is mainly distributed under

the same bioclimatic conditions. However, this

was not the case regarding the soil type vari-

able. The analysis of variance did not show

significant differences for most of the studied

variables, unlike the precipitation variables.

The references included in this study sug-

gest that the high humidity and low tempera-

ture conditions required by these species, the

Central American Sechium species should be

included in genetic improvement programs.

They can be used to improve not only the

domesticated species, but also to obtain cucur-

bitacin and pharmaceuticals. Within Mexico,

successful interspecific hybridization (S. edule

× S. compositum) has been carried out. In addi-

tion, several edible domesticated species and

their wild and bitter ancestor were subjected

to intraspecific hybridization, obtaining geno-

types of outstanding hybrid vigor regarding the

number of secondary metabolites (Aguiñiga-

Sánchez et al., 2015; Avendaño-Arrazate et al.,

2014; Cadena-Iñiguez et al., 2008). Compatibil-

ity tests between species of interest should be

carried out for this type of research. In addition,

studies about the morphological, molecular,

and biochemical characteristics of the popula-

tions of this genus should be conducted.

Central American Sechium species thrive

under similar bioclimatic conditions; howev-

er, significant precipitation differences were

recorded. These species develop in volcanic

12 Revista de Biología Tropical, ISSN: 2215-2075 Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

soils, high humidity conditions, and highly fluc-

tuating elevation ranges, but mainly in C type

climates. These characteristics are ideal for the

improvement of Sechium domesticated species.

They are also interesting from a bioprospecting

point of view, mainly in the research about their

biochemical components (e.g., cucurbitacins).

The Costa Rican Mountain ranges are the main

habitat of Sechium species.

Ethical statement: The authors declare

that they all agree with this publication and

made significant contributions; that there is no

conflict of interest of any kind; and that we fol-

lowed all pertinent ethical and legal procedures

and requirements. All financial sources are fully

and clearly stated in the acknowledgments sec-

tion. A signed document has been filed in the

journal archives.

REFERENCES

Aguiñiga-Sánchez, I., Soto-Hernández, M., Cadena-Iñi-

guez, J., Ruíz-Posadas, L. del M., Cadena-Zamudio, J.

D., González-Ugarte, A. K., Steider, B. W., & Santiago-

Osorio, E. (2015). Fruit extract from a Sechium edule

hybrid induce apoptosis in leukaemic cell lines but

not in normal cells. Nutrition and Cancer, 67(2), 250–

257. https://doi.org/10.1080/01635581.2015.989370

Allouche, O., Tsoar, A., & Kadmon, R. (2006). Assessing

the accuracy of species distribution models: Preva-

lence, kappa and the true skill statistic (TSS). Journal

of Applied Ecology, 43(6), 1223–1232. https://doi.

org/10.1111/j.1365-2664.2006.01214.x

Avendaño-Arrazate, C. H., Cadena-Iñiguez, J., Arévalo-

Galarza, M. L. C., Cisneros-Solano, V. M., Mora-

les-Flores, F. J., & Ruiz-Posadas, L. M. (2014).

Mejoramiento genético participativo en chayote.

AgroProductividad, 7, 30–39.

Baldwin, R. A. (2009). Use of maximum entropy modeling

in wildlife research. Entropy, 11(4), 854–866 https://

doi.org/10.3390/e11040854

Barrera-Guzmán, L. A., Legaria-Solano, J. P., Cadena-Iñi-

guez, J., & Sahagún-Castellanos, J. (2021). Phylogene-

tic relationships among Mexican species of the genus

Sechium (Cucurbitaceae). Turkish Journal of Botany,

45(4), 302–314. https://doi.org/10.3906/bot-2007-18

Bayliss, S. L. J., Mueller, L. O., Ware, I. M., Schweitzer, J. A.,

& Bailey, J. K. (2022). Stacked distribution models pre-

dict climate-driven loss of variation in leaf phenology

at continental scales. Communications Biology, 5(1),

1213. https://doi.org/10.1038/s42003-022-04131-z

Beck, J. (2012). Predicting climate change effects on agri-

culture from ecological niche modeling: Who profits,

who loses? Climatic Change, 116(1–2), 177–189.

Bedair, H., Shaltout, K., & Halmy, M. W. A. (2023). Stac-

ked machine learning models for predicting species

richness and endemism for Mediterranean endemic

plants in the Mareotis subsector in Egypt. Plant

Ecology, 224(12), 1113–1126. https://doi.org/10.1007/

s11258-023-01366-6

Breiman, L. (2001). Random forests. Machine Learning,

45(1), 5–32. https://doi.org/10.1023/A:1010933404324

Brock, G., Pihur, V., Datta, S., & Datta, S. (2008). clValid: An

R package for cluster validation. Journal of Statistical

Software, 25(4), 1–22. https://doi.org/10.18637/jss.

v025.i04

Cadena-Iñiguez, J., Avendaño-Arrazate, C. H., Soto-Her-

nández, M., Ruiz-Posadas, L. M., Aguirre-Medina,

J. F., & Arévalo-Galarza, L. (2008). Infraspecific

variation of Sechium edule (Jacq.) Sw. In the state

of Veracruz, Mexico. Genetic Resources and Crop

Evolution, 55(6), 835–847. https://doi.org/10.1007/

s10722-007-9288-4

Chang, A., Wu, T., Li, B., Jiao, D., Wang, Y., He, D., Jiang,

Z., & Fan, Z. (2024). Distribution pattern of spe-

cies richness of endemic genera in mountainous

areas of Southwest China and its influencing factors.

Sustainability, 16(9), 3750. https://doi.org/10.3390/

su16093750

Charrad, M., Ghazzali, N., Boiteau, V., & Niknafs, A.

(2014). NbClust: An R Package for determining the

relevant number of clusters in a data set. Jour-

nal of Statistical Software, 61(6), 1–36. https://doi.

org/10.18637/jss.v061.i06

Cross, H., Lira, S. R., & Motley, T. J. (2006). Origin and

diversification of chayote. In T. J. Motley, N. Zerega,

& H. Cross (Eds.), Darwin’s harvest: New approaches

to the origins, evolution, and conservation of crops (pp.

171–194). Columbia University Press.

Dubuis, A., Pottier, J., Rion, V., Pellissier, L., Theurillat, J.

P., & Guisan, A. (2011). Predicting spatial patterns of

plant species richness: A comparison of direct macro-

ecological and species stacking modelling approaches.

Diversity and Distributions, 17(6), 1122–1131. https://

doi.org/10.1111/j.1472-4642.2011.00792.x

Fick, S. E., & Hijmans, R. J. (2017). WorldClim 2: New 1-km

spatial resolution climate surfaces for global land

areas. International Journal of Climatology, 37(12),

4302–4315. https://doi.org/10.1002/joc.5086

Fischer, G., van Velthuizen, H., & Shah, M. (2002). Glo-

bal agro-ecological assessment for agriculture in the

21st century: Methodology and results. International

Revista de Biología Tropical, ISSN: 2215-2075, Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

Institute for Applied Systems Analysis (IIASA) &

Food and Agriculture Organization of the United

Nations (FAO).

Global Biodiversity Information Facility. (2024). GBIF

Occurrence Download. https://www.gbif.org/es/

Gelfand, A. E. (2022). Spatial modeling for the distribution

of species in plant communities. Spatial Statistics, 50,

100582. https://doi.org/10.1016/j.spasta.2021.100582

Hijmans, R. J. (2020). raster: Geographic data analysis and

modeling. R package (Version 3.3-13) [Software].

https://CRAN.R-project.org/package=raster

Hijmans, R. J., & Elith, J. (2013). Species distribution mode-

ling with R. R CRAN Project.

Kassambara, A. (2017). Multivariate analysis I. Practical

guide to cluster analysis in R. Unsupervised Machine

Learning (1st ed.). STHDA.

La Sorte, F. A., & Jetz, W. (2010). Projected range contrac-

tions of montane biodiversity under global warming.

Proceedings of the Royal Society B: Biological Scien-

ces, 277(1699), 3401–3410. https://doi.org/10.1098/

rspb.2010.0612

Lê, S., Josse, J., & Husson, F. (2008). FactoMineR: An R

package for multivariate analysis. Journal of Statistical

Software, 25(1), 1–18. https://doi.org/10.18637/jss.

v025.i01

Lira, R., Caballero, J., & Dávila, P. (1997). A contribution

to the generic delimitation of Sechium (Cucurbita-

ceae, Sicyinae). Taxon, 46(2), 269–282. https://doi.

org/10.2307/1224097

Lira, R., & Nee, M. (1994). A new species of Sechium

sect. Frantzia (Cucurbitaceae, Sicyeae, Sicyinae)

from México. Brittonia, 51(2), 204–209. https://doi.

org/10.2307/2666628

Lira, S. R. (1995). Estudios taxonómicos en el género Sechium

P. Br. Cucurbitaceae [Doctoral thesis, Universidad

Nacional Autónoma de México]. UNAM Repository.

https://repositorio.unam.mx/contenidos/82785

Lobo, J. M., Jiménez-Valverde, A., & Real, R. (2007).

AUC: A misleading measure of the performan-

ce of predictive distribution models. Global Eco-

logy and Biogeography, 17, 145–151. https://doi.

org/10.1111/j.1466-8238.2007.00358.x

Mateo, R. G., Felicísimo, A. M., & Muñoz, J. (2011). Species

distributions models: A synthetic revision. Revista

Chilena de Historia Natural, 84, 217–240. http://

dx.doi.org/10.4067/S0716-078X2011000200008

Meyer, D., Dimitriadou, E., Hornik, K., Weingessel, A.,

& Leisch, F. (2019). E1071: Misc functions of the

department of statistics (E1071), probability theory

group (formerly: E1071) TU Wien. R package

(Version 1.7) [Software]. https://CRAN.R-project.

org/package=e1071

Monge, J. E., & Loría, M. (2017). Caracterización de frutos

de cinco genotipos de tacaco [Sechium tacaco (Pittier)

C. Jeffrey] en Costa Rica. Tecnología en Marcha, 30(3),

71–84. https://doi.org/10.18845/tm.v30i3.3274

Morales, A. J. (1994). Morfología general del tacaco,

Sechium tacaco (Cucurbitaceae). Revista de Biología

Tropical, 42(1-2), 59-71. https://archivo.revistas.ucr.

ac.cr/index.php/rbt/article/view/22462

Munson, S. M., & Sher, A. A. (2015). Long-term shifts in

the phenology of rare and endemic Rocky Mountain

plants. American Journal of Botany, 102(8), 1268–

1276. https://doi.org/10.3732/ajb.1500156

Newstrom, L. E. (1990). Origin and evolution of chayote,

Sechium edule. In C. Jeffrey (Ed.), Biology and utili-

zation of the Cucurbitaceae (pp. 141–149). Cornell

University Press.

Noroozi, J., Talebi, A., Doostmohammadi, M., Rumpf,

S. B., Linder, H. P., & Schneeweiss, G. M. (2018).

Hotspots within a global biodiversity hotspot—Areas

of endemism are associated with high mountain

ranges. Scientific Reports, 8(1), 10345. https://doi.

org/10.1038/s41598-018-28504-9

Olguín-Hernández, G., Valdovinos-Ponce, G., Cadena-

Iñiguez, J., & Arévalo-Galarza, M. L. C. (2013). Etio-

logía de la marchitez de plantas de chayote (Sechium

edule) en el Estado de Veracruz. Revista Mexicana de

Fitopatología, 31(2), 161–169.

Patil, I. (2021). Visualizations with statistical details: The

‘ggstatsplot’ approach. The Journal of Open Sour-

ce Software, 6(61), 3167. https://doi.org/10.21105/

joss.03167

Peñas, J., Pérez-García, F. J., & Mota, J. F. (2005). Patterns

of endemic plants and biogeography of the Baetic

high mountains (south Spain). Acta Botanica Gallica,

152(3), 347–360. https://doi.org/10.1080/12538078.2

005.10515494

Phillips, S. J. (2010). A brief tutorial on Maxent. Lessons in

Conservation, 3, 108–135.

Qazi, A. W., Saqib, Z., & Zaman-ul-Haq, M. (2022).

Trends in species distribution modelling in context

of rare and endemic plants: A systematic review. Eco-

logical Processes, 11(1), 40. https://doi.org/10.1186/

s13717-022-00384-y

QGIS Development Team. (2020). QGIS Geographic Infor-

mation System. Open-Source Geospatial Foundation

Project (Version 3.16.2) [Software]. https://www.qgis.

org/en/site/

R Core Team. (2020). R: A language and environment for

statistical computing (Version 1.3.1093) [Software]. R

14 Revista de Biología Tropical, ISSN: 2215-2075 Vol. 73: e20251987, enero-diciembre 2025 (Publicado Set. 18, 2025)

Foundation for Statistical Computing. https://www.R-

project.org/

Revelle, W. (2020). psych: Procedures for psychological,

psychometric, and personality research. R package

(Version 2.0.9) [Software]. Northwestern University.

https://CRAN.R-project.org/package=psych

Schmitt, S., Pouteau, R., Justeau, D., de Boissieu, F., & Bir-

nbaum, P. (2017). SSDM: An R package to predict dis-

tribution of species richness and composition based

on stacked species distribution models. Methods in

Ecology and Evolution, 8(12), 1795–1803. https://doi.

org/10.1111/2041-210X.12841

Sebastian, P., Schaefer, H., Lira, R., Telford, I. R. H., & Ren-

ner, S. S. (2012). Radiation following long-distance

dispersal: The contributions of time, opportunity

and diaspore morphology in Sicyos (Cucurbitaceae).

Journal of Biogeography, 39(8), 1427–1438. https://

doi.org/10.1111/j.1365-2699.2012.02695.x

Shipley, B. R., & McGuire, J. L. (2022). Interpreting and

integrating multiple endemism metrics to identi-

fy hotspots for conservation priorities. Biological

Conservation, 265, 109403. https://doi.org/10.1016/j.

biocon.2021.109403

Vapnik, V. N. (1998). Statistical Learning Theory. Wiley.

Villanueva-Jiménez, J. A. (2012). Las variedades del chayote

(Sechium edule (Jacq.) Sw) y su comercio mundial.

Agricultura, Sociedad y Desarrollo, 9(4), 481–482.

Wani, Z. A., Akhter, F., Ridwan, Q., Rawat, Y. S., Ahmad, Z.,

& Pant, S. (2023). A bibliometric analysis of studies

on plant endemism during the period of 1991–2022.

Journal of Zoological and Botanical Gardens, 4(4).

692–710. https://doi.org/10.3390/jzbg4040049

Wunderlin, R. P. (1976). Two new species and a new com-

bination in Frantzia (Cucurbitaceae). Brittonia, 28(2),

239–244. https://doi.org/10.2307/2805833

Yilan, L., & Rutong, Z. (2015). clustertend: Check the clus-

tering tendency. R package (Version 1.4.) [Software].

https://CRAN.R-project.org/package=clustertend

Zwiener, V. P., & Alves, V. A. (2023). Community-level

predictions in a megadiverse hotspot: Comparison of

stacked species distribution models to forest inven-

tory data. Journal of Plant Ecology, 16(3), 099. https://

doi.org/10.1093/jpe/rtac099