Advancing Ecosystem Recovery with Diverse Species Plantings in Tropical Forest Restoration

Abstract

Tropical forest restoration has increased in the past decades, with possible advancements given the UN declaration of the “Decade of Ecosystem Restoration”. However, robust assessments to compare ecosystem functions among restored forest stages are essential. We evaluated 13 actively restored forest stands ranging from 3 to 21 years of age and compared measures of forest biodiversity, structure, and ecosystem function to four 70+ year old “reference” stands that serve as restoration “targets” in the study region of the Premontane wet forest of Costa Rica. The restored stands were planted with an average of 13 tree species on abandoned pastures that were fallow for at least two years. Sixteen tree-stand attributes and six ecosystem function estimates were assessed, including: annual biomass (C) accumulation, N-fixation potential, threatened species conservation, and the provision of avian frugivore forage, insect habitat, and insect pollination. Using Principal Component Analysis, linear modeling, and Mahalanobis distance analyses, we learned that planting a diversity of tree species sets the stage for forest recovery at early restoration ages, with an inflection point at 15 years towards older reference forest characteristics and functions. Given that all restoration ages provided tree diversity and some level of ecosystem functions, the value of all restored stands in the landscape is notable. The assessment methods are easily employed, thereby providing an accessible tool to restoration practitioners.

1. Introduction

Tropical forests are complex ecosystems that harbor immense amounts of biomass, host high levels of biodiversity, and are considered one of the most critical zones for biogeochemical cycles [1]. Even with many calls to halt forest destruction over the past 30 years, tropical deforestation increased by 10% in 2022 [2]. According to the Food and Agriculture Organization (FAO) and United Nations Environmental Programme (UNEP) [3], 420 million hectares of forest have been lost through conversion to other land uses in the past 30 years, and another 100 million hectares are at risk. However, the call for restoration is being addressed by multiple global initiatives, especially the United Nations Environment Programme (UNEP)’s global declaration of the “Decade on Ecosystem Restoration”, which aims to prevent and reverse the degradation of ecosystems worldwide.

At the same time, active-planting restoration of tropical forests has increased over the last decade [3], including the promotion of natural regeneration and other strategies to restore these important ecosystems. Specific restoration goals include biodiversity maintenance or enhancement, habitat recuperation, water regulation, carbon sequestration, crop protection, resource extraction, and heritage site preservation, among others [4]. For example, the primary regional objectives to reforest lands in the Bellbird Biological Corridor of Costa Rica are water protection, habitat restoration for plants and animals, and fruit and timber production [5]. Carbon offset programs focus on biomass (C) accumulation, and crop protection from wind damage is another common motivation.

From a conservation viewpoint, forest restoration seeks to match the forest integrity of mature forests or regional equivalents; that is, to attain the same levels of ecosystem functions and services as their adaptive natural forest counterparts by supporting forest and ecological structure, biodiversity, species composition, and ecosystem processes [6]. Additionally, the restored forest should enhance landscape connectivity and habitat matrices while not negatively impacting the functioning of surrounding biotic and abiotic ecosystems [7]. Here, we define and evaluate the forest integrity of actively restored tropical forest stands as a measure of how well the stand matches a mature, nearby “reference” forest in terms of forest composition, structure, and ecosystem function, where composition refers to the diversity of tree species and factors that affect its overall richness and community distribution, structure refers to tree and forest stand structure (such as basal area, canopy height, etc.), and function refers to key ecosystem functions such as biomass accumulation, nitrogen (N) fixation, habitat and forage provision, threatened species conservation, and ecological functions such as pollination.

Assessments are fundamental in determining whether any given restoration effort successfully attained the desired outcome [8,9], and meta-analyses can reveal practices that lead to success [10,11]. While traditional evaluations focused on forest composition and structure, the evaluation of ecosystem functional recovery and service outputs is now being promoted as a valuable test of restoration success [12]. Some studies have incorporated such evaluations into their restoration programs, including the examination of erosion control, soil improvement, carbon and nutrient cycling, biodiversity, habitat provision, water quality, secondary forest succession, and specific taxa-driven functions [13,14,15,16]. Plant functional traits are being analyzed as surrogates for ecosystem services to assess (and direct) restoration success [11] with stand-level tree functional traits as the focus of new studies [17], although most studies are in temperate areas or in restoration actions less than five years of age [18]. In general, ecological assessments are still lacking [19,20], highly variable [4], and evaluate subsets of variables over relatively short time periods of <10 years [12]. Evaluations of the socio-economic indicators and landscape-level impacts are lacking [21].

For restoration efforts aimed at matching the forest integrity of original forests, three key questions are of interest and could be addressed in the context of the applied restoration strategy: Do reforestation stands attain the compositional, structural, and functional levels of target, older forests? If so, what is the overall trajectory over time to reestablish its integrity, and how does the applied restoration strategy influence this timeline? Finally, are there ecosystem contributions at younger restoration stages that complement the landscape mosaic of ecosystem services? To address these questions, assessments of different types of restoration actions in attaining the desired ecosystem structure and function, and at different ages and sizes, need to be done in parallel [8,9]. Ecosystem functions result from the interactions among biodiversity, forest structural components, and abiotic factors. As such, the restoration of species communities and forest stature are key restoration considerations, where multiple forest characteristics are linked and lend themselves to simultaneous evaluation. Such comprehensive studies can provide robust assessments of restoration outcomes [10,11,21,22].

Lastly, while the recovery of forest integrity is a laudable restoration goal, evaluating success can be time- and resource-intensive, posing challenges to many grassroots and conservation organizations that perform this work. The development of succinct and accessible evaluations of biodiversity, forest stand characteristics, and the estimation of ecosystem function to advance our knowledge of effective restoration practices would be more widely employed if available to the same people who restore forests.

Here, we add to the growing knowledge base of best practices for tropical restoration by assessing the forest integrity of an active, multi-species planting restoration effort in the Premontane Wet Life Zone, Monteverde, Costa Rica. We compared 13 actively restored forest stands planted 3 to 21 years prior on abandoned pastures to nearby reference stands of altered mature forests that were at least 70 years old (actual age unknown). These forests are altered with some human interference and edge effects, yet represent the best examples of pre-existing forests in the region. The restoration effort was led by the Fundación Conservacionista Costarricense (FCC) in their efforts to replicate natural forests in the region, expand the landscape matrix of forest and open lands, and provide socio-ecological benefits for the communities within the restoration areas [23]. For each restoration and reference stand, we used a suite of easily measurable tree species traits to determine the multivariate distance of forest species composition, structure, and specific ecosystem functions to older reference forests. Our framework is rooted in ecological community composition and multivariate and linear modeling analyses of forest traits, but the application utilizes basic forestry and tree species knowledge, forest/tree measurements, requires little training, and is cost-effective. Our assessment addresses the following specific objectives:

• Compare the forest composition (diversity and similarities) along age-based succession gradients of restored forests and examine the effect of actively planting a diversity of species on richness over time.
• Compare the structural characteristics of different-aged stands to the target, reference forest (>70 years).
• Determine levels of six ecosystem functions (biomass (C) accumulation, nitrogen (N) fixation, insect habitat and pollination potential, avian frugivore forage provision, and threatened species conservation) along a gradient of different-aged, restored stands, and compare those levels to the reference forests (altered mature forests >70 years).

2. Methods

2.1. Study Site and Stand Descriptions

This study strived to represent possible outcomes for small-scale plantings (landowner) reforestation efforts. Therefore, certain similarities among stands exist, such as life zone (regional climate and elevation), and prior land use history (abandoned pasture left fallow for at least 2 years). All sites were planted with species native to the region that were nursery-raised by the FCC. However, variations by stand included the planted species composition, year of planting, and seedling weeding protocols, as occurs in real-life situations and most time-based tropical forest restoration studies.

We surveyed 13 restored forest stands that were planted between 2002 and 2020 in Monteverde, Costa Rica (

Table 1. Stand descriptions of 13 restored forest stands and four mature altered primary (reference) forests. Data include the elevation, stand size, and the years fallow before planting. A summary of the year and number of species planted, the current species richness (spp_2023), and the number of new species from recruits that were not planted (new_spp) is shown. The number (#) of plots in each stand is provided. The years of maintenance (weeding) are listed since this may influence the establishment of new recruits, especially in the younger stands.

Stand_Id	Forest_Age	Elevation	Area_m2	YrsFallow	Year_Planted	spp_Planted	spp_2023	New_spp	# Plots	Maintenance
REFR-R20	3	1290	2755	8	2020	18	17	5	4	3
REFR-R20d	3	1280	770	8	2020	11	17	7	4	3
REFR-R15	8	1280	1204	3	2015	12	15	5	4	4
REFR-R15d	8	1280	771	3	2019	7	18	12	4	4
NACI-R11c	12	1190	3611	5	2011	15	18	9	3	5
NACI-R11i	12	1190	2340	5	2011	4	10	5	3	5
NACI-R11m	12	1220	2258	5	2011	5	9	4	3	5
REFR-R11	12	1270	680	4	2011	17	11	4	3	3
CRRE-R8	15	1420	1422	5	2008	15	36	25	3	1
NACI-R8	15	1185	2340	4	2008	23	21	8	4	0
REFR-R8	15	1240	730	4	2008	15	21	8	3	1
CALA-R2LC1	21	1210	3357	2	2002	9	22	13	4	1
CALA-R2LC3	21	1260	2392	2	2002	13	24	11	4	1
CALA-O	70	1240	292,551	--	--	--	41	41	6	0
CRRE-O	70	1440	150,000	--	--	--	41	36	4	0
KANE-O	70	1360	350,000	--	--	--	25	25	4	0
REFR-O	70	1240	76,101	--	--	--	28	28	4	0

) with 8 100 m² plots each, 1 additional plot in the 15-year stands, and 12 plots total in the 12-year stands. The stands ranged in time since planting (forest age) from 3 to 21 years at the time of this study. We also surveyed 19 plots in four corresponding altered primary forest stands (hereafter referred to as “reference” forests) that were used as our target reference to assess restoration success (Table 1).

All stands were in the Premontane Wet Life Zone on the Pacific side of the Tilarán mountain range in Costa Rica at elevations between 1200 m and 1440 m asl (10° 18′ 55.3″ N and 84° 50′ 29.8″ W; Figure 1). Average annual precipitation in this region is 2883.6 mm, based on data collections between 1981 and 2023, with a pronounced dry season from February through April (<100 m per month) and a windy, misty transition period from November to January after the five-month rainy season [24]. Temperatures ranged between lows of 15 C and highs of 26 C. Clear-sky solar irradiance at midday was estimated to be between 875 and 1085 W/m² with great variation due to cloud cover amount and types [25]. Soils were classified as Andisols, formed under wet (udic) conditions from volcanic parent material with moderate clay content [25].

The 13 restored forest stands were located close to the reference forests (average of 93 m) with a distance between stands ranging from 75 m to 7 km. The restored stands were established on pasture previously grazed by cattle and horses; three stands had been converted to coffee post-grazing. All sites were abandoned at least two years before planting (Table 1). Due to the high winds the area experiences from November to April, the individual pastures had been protected by windbreaks comprising the introduced tree species Cupressus lusitanica or Casuarina equisetifolia, or a planted mix of one coniferous species with native species of Montanoa guatemalensis, Croton niveus, and/or Zanthoxylum fagara. Two sites were quickly colonized by “guayaba” (Psidium guajava), a common weedy tree in abandoned pastures in Monteverde, after the removal of horses. Each restored stand was manually cleared (with a machete) before planting, including the cutting of Psidium guajava trees.

The four older reference forests ranged in size from 7.6 to 35+ ha and were natural compositions of nearly 100 tree species (the regional tree diversity throughout five life zones in Monteverde is 755 species). Unaltered primary forest is scarce in this study’s region of the Premontane Wet Life Zone, and human alterations from >30 years ago within the reference forests were evident with some large tree removals. Natural blowdowns and light gaps added to the natural diversity of species, and edge effects were apparent where higher solar irradiance and wind effects occur. Lauraceae was a dominant plant family in these forests, where the region holds approximately 96 species [26].

2.2. Seedling Planting and Diversity

Seedlings were planted approximately 1.75–2 m apart for 25–33 seedlings per 100 m² per plot with an average of 2900 stems ha⁻¹, similar to other tropical planting densities [27] The FCC reforestation methods focus on replacing forests with native species compositions representative of the area, and a total of 48 species of 19 tree families were planted in the restored forests (Appendix A 

Table A1. Tree species names and traits of the 98 taxa included in this study. The column spp_planted gives the total number of each species that was planted in restoration stands. Traits include the International Union for the Conservation of Nature categories of world status (1—least concern, 2—near-threatened, 3—vulnerable, 4—endangered, and 5—critically endangered), wood_density (g/cm³ according to Zanne et al. [32]), N-fixer category, the number of threatened bird species that the tree supports (ES_prov), light flower producer (binary), the number of macro-morph insect groups that visit the tree’s flowers (insect_pollinators), pollinated by bees (binary), fruit size and type (A-aril, B-berry, C-capsule, D-drupe, FL-large fleshy, NF-no fruit, P-pod, S-spike, the number of birds that feed on those fruits (bird_fruit), and the presence of fallen fruit that the tree provides (based on fruit size and fruit type, ranging from 0 to 2). Data sources include the following citations: [26,32,33,35] and [34] via the R package, BIEN. Names and authorities were verified with Catalogue of Life [36,37,38].

Taxa_Id	Taxa_Name	Catalogue of Life_Id	Family	Name_Common	spp_Planted	IUCN	Wood_Density	N_Fixer	ES_Prov	Flowers_Light	Insect_Pollin-Ator	Bees	Fruit_Size	Fruit_Type	Bird_Fruit	Fallen Fruit
AIBR	Aiouea brenesii (Standl.) R. Rohde	BCMX	Lauraceae	Aguacatillo	475	2	0.6	0	4	1	4	1	1	D	9	1
AICO *	Aiouea costaricana (Mez & Pittier) Lorea-Hern.	48HKB	Laura-ceae	Aguacatillo	564	1	0.6	0	4	1	4	1	1	D	9	1
AITO	Aiouea tonduzii (Mez) Kosterm.	BCQV	Lauraceae	Aguacatillo	20	1	0.6	0	0	1	4	1	2	B	9	2
ARRE	Ardisia revoluta Kunth.	5W44P	Myrsinaceae	Tucuico	126	1	0.6	0	0	1	1	1	1	B	7	1
BEAL	Beilschmiedia alloiophylla (Rusby) Kosterm.	68FJ3	Lauraceae	Aguacatillo	7	1	0.5	0	4	1	4	1	3	D	3	2
BEBR	Beilschmiedia brenesii	L93V	Lauraceae	Aguacatillo	79	1	0.466	0	4	1	4	1	3	D	3	2
BEIM	Beilschmiedia immersinervis	L96V	Lauraceae	Aguacatillo	44	4	0.466	0	4	1	4	1	3	D	3	2
BIRO	Billia rosea (Planch. & Linden) C.Ulloa & P.M.Jørg.	68KXJ	Sapindaceae	Cucaracho	7	1	0.69	0	0	1	1	1	5.5	C	0	2
BUVE	Bunchosia veluticarpa W.R. Andersen	68RYL	Rutaceae	Bunchosia	4	1	0.6	0	0	1	1	1	2	FL	2	2
CAED	Casimiroa edulis La Llave	RFK7	Rutaceae	Matasano	147	1	0.6	0	0	1	2	1	6.5	FL	0	2
CAEL	Cassipourea elliptica (Sw.) Poir.	5XD7C	Rhizophoraceae	Cassipourea	0	1	0.82	0	0	1	1	1	1	B	5	1
CAGU	Cassipourea guianensis Aubl.	RNB5	Rhizophoraceae	Cassipourea	0	1	0.82	0	0	1	1	1	1	B	5	1
CEOB	Cecropia obtusifolia Bertol.	RZSK	Cecropiaceae	Guarumo	0	1	0.308	0	5	0	0	0	1	SP	10	1
CEOD	Cedrela odorata L.	RZZT	Meliaceae	Cedro amargo	0	3	0.4	0	0	1	1	0	0	NF	0	0
CETO	Cedrela tonduzii C.DC	S22J	Meliaceae	Cedro dulce	0	2	0.36	0	0	1	1	0	0	NF	0	0
CICO	Citharexylum costaricense Moldenke	VLZ6	Verbenaceae	Dama	752	1	0.66	0	0	1	2	1	1	B	23	1
COAR	Coffea arabica L.	WVWV	Rubiaceae	Café	0	1	0.62	0	0	1	1	1	1	B	3	1
COCO	Cojoba costaricensis Britton & Rose	5ZK4D	Fabaceae	Cabello de ángel	1	2	0.65	1	0	1	1	0	0	P	0	0
COER	Cordia eriostigma Pittier	YBC7	Boraginaceae	Laurel muñeco	0	1	0.6	0	0	1	2	1	1	B	2	1
COXA	Conestegia xalapensis (Bonpl.) D.Don ex DC.	6RG6V	Melastomaceae	Maria	0	1	0.6	0	0	1	1	1	1	B	16	1
CRNI	Croton niveus Griseb.	ZQPM	Euphorbiaceae	Colpachi	85	1	0.645	0	0	1	3	0	0	NF	0	0
CUGL	Cupania glabra Sw.	6C3PL	Sapindaceae	Cupania	0	1	0.611	0	0	1	2	1	1	C,A	5	0
DAAM	Daphnopsis americana (Mill.) J.R. Johnst.	345N6	Thymelaeaceae	Mastate	0	1	0.6	0	0	1	3	1	1	B	5	0
DASA	Damburneya salicina (C.K. Allen) Trofimov & Rohwer	6CB4G	Lauraceae	Canelo	417	2	0.565	0	4	1	4	1	3	D	5	2
DEAR	Dendropanax arboreus (L.) Decne. & Planch.	34R42	Arialaceae	Dendropanex	0	1	0.42	0	0	1	4	1	1	B	4	1
DIAM	Diphysa americana (Mill.) M.Sousa	36G8X	Fabaceae	Guachepelin	3	1	1	1	0	1	1	0	0	P	0	0
ERLA	Erythrina lanceolata Standl.	6GRZH	Fabaceae	Poro	27	1	0.32	1	0	0	0	0	5	P	0	0
ERMA	Erythroxylum macrophyllum Cav.	3BFXL	Erythroxylaceae	Coca	0	1	1.07	0	0	1	2	1	1	B	2	0
EUMO	Eugenia monteverdensis Barrie	3C5HJ	Myrtaceae	Muerta blanca	42	2	0.829	0	5	1	1	1	2.5	NF	1	0
EXPA	Exothea paniculata Radlk.	3DQF2	Sapindaceae	Dantisco	58	1	0.73	0	0	1	1	1	1.5	B	1	0
FIPE	Ficus pertusa L.f.	6HYFY	Moraceae	Higueron	0	1	0.6	0	0	1	1	0	5	B	8	2
GAIN	Garcinia intermedia (Pittier) Hammel	3F9WY	Clusiaceae	Jorco	11	1	0.625	0	0	1	4	1	1	B	2	1
GUGL	Guarea glabra Vahl	8S4YD	Meliaceae	Guarea	0	1	0.45	0	5	1	1	0	1	D	5	1
HAAP	Hampea appendiculata (Donn. Sm.) Standl.	YLBTJ	Malvaceae	Burio	57	1	0.254	0	5	1	2	1	3	C,A	22	2
HAPA	Hamelia patens Jacq.	3JHJ8	Rubiaceae	Hamelia	0	1	0.37	0	0	0	0	0	1	B	2	0
HAFL	Hasseltia floribunda Kunth	3JVNZ	Flacourtaceae	Layo	0	1	0.53	0	5	1	1	1	1	B	14	2
HEAM	Heliocarpus americanus L.	6LLBX	Tiliaceae	Heliocarpus	0	1	0.18	0	0	1	2	1	1	C	0	0
INOE	Inga oerstediana Kunth.	3PPTZ	Fabaceae	Guava	2	1	0.425	1	0	1	1	0	5	P,A	0	2
INPU	Inga punctata Willd.	3PPWK	Fabaceae	Guava	701	1	0.56	1	0	1	1	0	7	P,A	0	2
IOAR	Iochroma arborescens (L.) J.M.H.Shaw	3PTQG	Solanaceae	Guitite	200	1	0.44	0	1	1	5	1	1	B	43	1
KOHY	Koanophyllon pittieri (Klatt) R.M. King & H. Rob.	6NNC5	Asteraceae	Aster	0	1	0.4	0	0	1	1	0	0	NF	0	0
KOPI	Koanophyllon hylonomum (B.L.Rob.) R.M.King & H.Rob.	6NLLC	Asteraceae	Aster	0	1	0.4	0	0	1	1	0	0	NF	0	0
LAFR	Lasianthaea fruticosa (L.) K.M.Becker	6NYVP	Asteraceae	Aster	0	1	0.4	0	0	1	0	0	0	NF	0	0
LAUR	Lauraceae spp	BTB	Lauraceae	Aguacatillo	0	1	0.4	0	0	1	4	1	3	D	4	2
LOOL	Lonchocarpus oliganthus F.J.Herm.	3VXX4	Fabaceae	unknown	0	1	0.64	1	0	1	1	1	2	P	0	0
MAHE	Mauria heterophylla Benth.	3YFX4	Anacardiaceae	Cirri amarillo	602	1	0.311	0	0	1	1	0	1	B	5	2
MAOP	Matayba oppositifolia (A.Rich.) Britton	6R78W	Sapindaceae	Aranillo	4	1	0.6	0	0	1	1	1	2	C,A	6	2
MATE **	Matayba “teton”		Sapindaceae	Teton	0	1	0.6	0	0	1	1	1	2	C,A	6	0
MEVE	Meliosma vernicosa (Liebm.) Griseb.	3ZJ5W	Sabiaceae	Espavel de altura	0	1	0.6	0	2	1	1	1	2	NF	3	0
MIOE	Miconia spp.	63B9N	Melastomaceae	Unknown	0	1	0.6	0	0	1	1	1	1	B	6	2
MOGU	Montanoa guatemalensis B.L.Rob. & Greenm.	7426W	Asteraceae	Tubu	0	1	0.6	0	0	1	0	0	0	NF	0	0
MYBF	Myrcianthes “black fruit”	5WH7	Myrtaceae	Myrcianthes	302	2	0.7	0	5	1	1	1	1	B	1	2
MYCO	Myrsine coriacea (Sw.) R. Br. ex Roem. & Schult.	KZBYL	Myrsinaceae	Ratoncillo	29	1	0.65	0	0	1	0	0	1	B	10	0
MYFR	Myrcianthes fragrans (Sw.) McVaugh	457YB	Myrtaceae	Albajaquillo, Calico Tree	0	2	0.7	0	0	1	1	1	1	B	1	2
MYSP	Myrcia splendens (Sw.) DC. *	73S4K	Myrtaceae	Myrcia	0	1	0.8	0	0	0	1	1	1	B	5	2
NEME	Nectandra membranacea (Sw.) Griseb.	462GP	Lauraceae	Aguacatillo	333	1	0.4	0	10	1	4	1	4	D	7	2
OCFL	Ocotea floribunda (Sw.)Mez	48HGQ	Lauraceae	Quizarra qunia	329	1	0.388	0	10	1	4	1	2	D	18	2
OCLL	Ocotea “losllanos”		Lauraceae	Quizarra blanca	275	2	0.395	0	0	1	4	1	2	D	6	2
OCMO	Ocotea monteverdensis W.C. Burger	74FZF	Lauraceae	Aguacatillo	383	5	0.523	0	10	1	4	1	3	D	6	2
OCSI	Ocotea sinuata (Mez) Rohwer	48HWJ	Lauraceae	Aguacatillo	67	2	0.395	0	0	1	4	1	2	D	6	0
OCTE	Ocotea tenera Mez & Donn. Sm.	48HXZ	Lauraceae	Aguacatillo	130	1	0.395	0	5	1	4	1	3	D	6	2
OCWH	Ocotea whitei Woodson	48J22	Lauraceae	Ira rosa	376	1	0.523	0	5	1	4	1	5	D	6	2
ORCR	Ormosia cruenta Rudd.	74V52	Fabaceae	unknown	0	1	0.6	1	0	1	1	0	6	P	0	0
ORXA	Oreopanax xalapensis (Kunth) Decne. & Planch.	75,696	Arialaceae	Oreopanex	0	1	0.505	0	0	1	3	1	1	B	13	2
PACO	Panopsis costarricensis Standl.	75MD9	Proteaceae	Papa	0	1	0.51	0	0	1	1	0	5	NF	0	0
PEAM	Persea americana Mill	4F97T	Lauraceae	Aguacatillo	160	1	0.6	0	0	1	2	0	5	D	1	0
PIPER	Piperaceae spp.	625NL	Piperaceae	Piper	0	1	0.3	0	0	1	0	0	1	SP	2	0
POEX	Pouteria exfoliata T.D. Penn.	4M7HQ	Sapotaceae	Tempisque	60	1	0.63	0	0	0	1	0	4	D	0	0
PRBR	Prunus brachybotrya Zucc.	4N8PP	Rosaeae	Duraznillo, wild Cherry	25	1	0.6	0	0	1	1	1	5	D	2	2
PRRH	Prunus rhamnoides Koehne	4N96G	Rosaeae	Duraznillo, wild cherry	33	1	0.6	0	0	1	1	1	2	D	2	2
PSGU	Psidium guajava Koehne	4PFV7	Myrtaceae	Guayaba	0	1	0.652	0	0	1	1	1	3	FL	0	2
PSMO	Pseudolmedia mollis Standl.	78FRN	Moraceae	Pseudolmedia	0	1	0.75	0	0	1	1	0	3	P	0	0
PSYC	Psychotria spp.	8VSFT	Rubiaceae	Psychotria	0	1	0.37	0	0	1	2	1	1	B	2	0
QUCO	Quercus cortesii Liebm.	4R4P5	Fagaceae	Roble	4	2	0.61	0	0	1	0	0	2	NF	0	0
QUIN	Quercus insignis M. Martens & Galeotti	4R58L	Fagaceae	Roble	40	5	0.61	0	0	1	0	0	4	NF	0	0
RAAM	Randia matudae Lorence & Dwyer	79SZM	Rubiaceae	Randia	0	1	0.6	0	0	1	3	0	1	B	2	0
ROGL	Roupala glaberrima Pittier	4TH2V	Proteaceae	Danto	38	2	0.8	0	0	1	0	0	0	NF	0	0
RUBI	Rubiaceae spp	SGJN5	Rubiaceae	Rubiaceae	0	1	0.62	0	0	1	0	0	1	B	2	0
SAGL	Sapium glandulosum (L.) Morong	6XLD6	Euphorbiaceae	Yos	0	1	0.415	0	0	0	1	0	1	D	22	0
SALA	Sapium laurifolium (A. Rich.) Griseb.	6XLG2	Sapindaceae	Yos	0	1	0.415	0	0	1	1	0	3	D	22	0
SIGR	Siparuna grandiflora (Kunth) Perkins	4XN62	Monimiaceae	Siparuna	0	1	0.6	0	0	0	0	0	1.5	B	1	1
SIPO	Sideroxylon portoricense-Urb.	4X9JR	Sapotaceae	Tempisque	54	2	0.914	0	0	1	1	1	2	D	0	0
SOAP	Solanum aphyodendron S. Knapp	4XZJR	Solanaceae	Solanum	0	1	0.545	0	0	1	1	1	1.5	B	5	0
SOTR	Sorocea trophoides W.C. Burger	4YCPW	Moraceae	Sorocea	41	1	0.6	0	0	1	0	0	1	B	8	1
STAR	Styrax argenteus C.Pensl	7B3TL	Styracaceae	Recino	0	1	0.6	0	0	1	1	1	2	D	5	1
STMO	Styphnolobium monteviridis M. Sousa & Rudd	9DN7J	Fabaceae	Fijolillo	0	4	0.371	1	0	0	2	1	4	P,A	0	2
STPE	Stauranthus perforatus Liebl.	4ZLXW	Rutaceae	Stauranthus	115	2	0.5	0	0	1	2	1	4	D	5	2
SYGL	Symphonia globulifera L. f.	53PKY	Clusiaceae	Cerillo	0	1	0.6	0	0	1	0	0	2.5	D	5	0
SYLI	Symplocos limoncillo Bonpl.	53RT3	Symplocaceae	Pava	33	1	0.8	0	10	1	2	1	1.5	D	7	1
TAME	Tapirira mexicana Marchand	54TFS	Anacardiaceae	Duraznillo	0	1	0.45	0	0	1	1	0	2	B	5	2
TRHA	Trichlia havanensis Jacq.	588RG	Meliceae	Uruca	0	1	0.46	0	0	1	2	1	1	NF	5	0
TRMI	Trema micranthum (L.) Blume	582Z4	Cannabaceae	Capulin	0	1	0.55	0	0	1	1	0	1	B	16	0
TRRA	Trophis racemosa (L.) Urb.	7D7B6	Moraceae	Ficus	48	1	0.638	0	0	1	0	0	1	B	8	1
UNKN	Unknown	P	Unknown	Unknown	0	1	0.6	0	0	0	1	0	0	NF	0	0
UNKNLL	Unknown-longleaf	P	Unknown	unknown	0	1	0.6	0	0	0	0	0	0	NF	0	0
URCA	Urera caracasana (Jacq.) Gaudich. ex Griseb.	7DST5	Urticaceae	Ortiga	0	1	0.6	0	0	0	0	0	1	B	13	0
VICO	Viburnum costaricanum (Oerst.) Hemsl.	7G2R7	Viburnaceae	Paraviento	0	1	0.6	0	0	1	1	1	1	B	8	1
XYIN	Xylosma intermedia (Seem.) Triana & Planch.	5CMPS	Salicaceae	Xylosma	20	1	0.623	0	0	1	1	0	1	B	5	1
YUGU	Yucca guatemalensis Baker	5CT75	Aparagaceae	Itabo	0	1	0.3	0	0	1	1	1	4	FL	0	0
ZAFA	Zanthoxylum fagara G. Don	7GD77	Rutaceae	Limoncillo	166	1	0.633	0	0	1	1	0	1	C	2	0
ZAJU	Zanthoxylum juniperinum Poepp.	5LTTS	Rutaceae	Lagartillo	0	1	0.6	0	0	1	1	0	6	C,A	2	0
ZYPA	Zygia palmana (Standl.) L. Rico	7GFTP	Fabaceae	Zygia	0	1	0.553	1	0	1	1	0	6	P	0	0

* Aiouea costaricana is an accepted name according to Tropicos.org of the Missouri Botanical Garden under the authorities of (Mez & Pittier) Lorea-Hern. Synonyms are Aiouea pittieri, but this is not conclusive in the Monteverde zone. Other synonyms include Cinnamomum costaricanum and Phoebe costaricana. ** Matayba “teton” and Ocotea “losllanos” are undescribed species found in the Monteverde zone.

). Each restored forest stand was planted with combinations of multiple species (average of 13 per stand) in a random nurse cropping method [28] where one-third of the species were fast-growing colonizing “nurse” species. Aside from growth rates, these taxa also varied in seedling survival [29] and traits at maturity, including fruit size and type. Of the 7494 individual trees planted in the stands, the most abundantly planted species were Inga punctata (Fabaceae), Mauria heterophylla (Anacardiaceae), Citharexylum costarricensis (Verbenaceae), Myrcianthes “black fruit” (Myrtaceae), and multiple species of Lauraceae (examples: Ocotea monteverdensis, floribunda, tenera, whitei, and sinuata, Aiouea brenesii and costaricana, Damburneya salicina, and Nectandra membranacea) (Appendix A Table A1; Supplementary Table S1). Most stands were maintained (weeded) within 1 m around seedlings for a minimum of two years.

2.3. Tree/Stand Metrics, Traits, and Calculations

A total of 64 plots of 100 m² in the 17 stands were sampled (Table 1). Each restored forest stand age had between 8 and 12 plots each. Reference forests, to which the restored forests were compared, were more heavily sampled (22 total) than restored stands in order to provide a more accurate representation of older forests, resulting in a more conservative comparison between restored and reference stands. Plots were randomly selected in ordinal directions and paces from the center of the stand. If the selection fell within a previously established plot used in prior bird studies [30], it was used in this study. Tree identification and measurements of the trees, understory vegetation, and stand characteristics were gathered in a two-year period of 2023–2024 from stands that ranged from 3, 8, 12, 15, to 21 years since planting (with each “age” class a 1-year possible difference if measured after June of 2024) and the 70-year reference forests.

Forest structure. In each plot, we collected data to assess the forest composition and structure and key variables related to selected ecosystem functions (

Table 2. Description of variables assessed in this study by research objective, identified by category (species richness, tree measurements, tree ecosystem function traits, stand measurements, and derived ecosystem traits.

Variable	Obj.	Units	Description
Species richness
spp_planted	1	count	Number of species planted
spp_2023	1	count	Species richness at time of assessment (2023–2024)
Tree Measurements
basal.area_m2	2,3	m2	Basal area per m2 is the space covered by the tree trunk in a m2; then summed by plot and averaged by stand
branching_depth	2	m	Average measurement (m) between the tree height and the lowest branch height
dbh1	1,2,3	cm	Diameter (cm) at breast height (1.3 m from ground); aka basal area or DAP
dbh2, _dbh3, _dbh4	3	cm	For trees with multiple boles, all dbh > 2 cm were measured; dbh1 = greatest, dbh2 = 2nd greatest, etc.
lowest_branch_ht	2	m	Distance (m) from the ground to the lowest horizontal branch on the tree
tree_ht	2	m	Height (m) from the forest floor to the highest (apical) growing point
Tree ecosystem function traits
bees	3	yes/no	Affirmative value of 1 if the tree’s flowers are visited by bees, 0 if it does not.
bird_fruit	3	count	Number of bird species recorded to feed on the tree species
ES_prov	3	count	Number of threatened avian species known to be supported by the tree species
fallen_fruit	3	0–2	0 for non-fleshy, pod, or capsule fruits, 1 for fruits < 2 cm in diameter, 2 for fruits> 1.9 cm
flower_light	2,3	yes/no	Affirmative value of 1 if the tree produces light-colored flowers, 0 if it does not.
fruit_size	3	mm	Sources [26,31]
fruit_type	3	category	A-arillate seed, B-berry, C-capsule, D-drupe, FL-larger fleshy fruit, NF-dry fruit, P-pod, SP-spike
insect_pollinators	3	yes/no	Number of macro-taxa groups that visit the tree’s flowers: Hymenoptera (w-wasps, b-bees, f-flies, a-ants), Lepidoptera (bfl-butterflies, m-moths), Coleoptera (c-beetles), and other (i-other)
IUCN	3	category	Category defined by the International Union for the Conservation of Nature (IUCN) as 1—least concern, 2—near-threatened, 3—vulnerable, 4—endangered, 5—critically endangered
N_fixer	3	yes/no	Affirmative value of 1 if a leguminous N-fixing symbiont, 0 if it does not.
wood_density	2,3	g/cm3	Values provided by [32] (see notes in text)
Stand measurements
canopy_cvr	2	%	% of canopy covered measured from 1.5 m using % Cover IOS application: 4 measurements by plot, averaged by stand and forest age
canopy_ht	2	m	Sample measurements (by drone/clinometer) of the closed canopy, average of 4 measurements by plot, averaged by stand and forest age
grass_cvr	2,3	%	Estimated % of forest floor covered by grass—average of 4 subplots of 1 m2 per plot
groundveg_cvr	2,3	%	Estimated % of forest floor covered by all types of vegetation—average of 4 subplots of 1 m2 per plot
palms	2,3	count	Average number of palms in the plot; palms are late to colonize restoration areas
psyc	2,3	count	Average number of Psychotria (Rubiaceae) plants in the plot; Psychotria are one of the most abundant understory fruiting plants
ratio	2	ratio	Proportion of larger (>10 cm dbh) trees to smaller (3–9.9 cm dbh) trees as a measure of structural complexity of habitat
snags_logs	2,3	count	Average number of snags and logs (>10 cm diameter) in the plot
strata	2,3	average	# of closed foliage heights at different levels of the forest
veg_ht	2,3	m	Average height (m) of the tallest understory plants
vines_lianas	2,3	count	Average number of trees with vines or lianas in the plot
Derived Eco-Function Estimates
Biomass(C)	3	calculated	Biomass(C) (kg) = EXP (3.41222 + 2.61148 x LN(dbh cm)) [31]
N_fixation_potential	3	calculated	Count of each n-fixing bacterial host tree in the plot x inverse relationship with age
Theatened_spp_protection	3	calculated	Sum per plot of the IUCN category ratings for trees > 10 cm dbh1
Avian_frugivore_forage	3	calculated	Sum of the number of bird species that feed on each tree
Insect_habitat_potential	3	calculated	The number of light-colored flower-producing trees, the fallen fruit estimate, and the number of snags/logs > 10 cm dbh.
Insect_pollinator_potential	3	calculated	The number of morpho taxa that visit the tree species’ flowers, plus 1 additional point if bees are one of the morpho groups

). For each tree > 150 cm in height, we identified the tree taxa and measured tree height (tree_ht), diameter at breast height (dbh; up to 4 boles for multi-trunk trees: dbh1–dbh4), and height from the ground to the lowest branch (lowest_branch_ht). We estimated the branching depth (branching_depth) of each tree as the height of the tree minus the lowest branch height. From these data, we calculated two plot level datum: basal area per m² (basal_area_m²) and ratio of large (>10 cm) to small (<10 cm dbh) trees in the plot (ratio). The sum of the tree basal area represents the area covered by tree trunks and was used in place of stem density.

Other plot characteristics of note included average canopy height (canopy_ht; measured with clinometer and drone), number of strata including the lower shrub layer (strata), percent canopy cover from 1.5 m above the ground (canopy_cvr; using Canopy % iOS app by Public Interest Enterprises), number of snags and dead logs > 10 cm dbh (snags_logs), the count of key understory plants (palms and Psychotria species, labeled palms and psyc, respectively), and the number of trees hosting vines or lianas (vines_lianas) (Table 2).

We measured understory traits in four 1 m² sub-plots within each 10 m × 10 m plot at 1 m distances diagonal within the four corners (Table 2). The plants at each corner of the 1 m sub-plot were measured for height (veg_understory_ht) and identified by taxa for a total of 16 plants per plot, and also used to calculate ground cover of plots (ground_veg_cover). The percentage covered by grass species (grass_covr) was visually estimated in each sub-plot. Of note, due to high levels of precipitation during this study, we were not able to accurately measure leaf litter, yet we recognize its ecosystem importance.

Tree species traits. In addition to the direct measurements of tree characteristics, we compiled tree species traits applicable to this study from the literature [26,32,33,34]: wood density (wood_density), light flower color (flower_light; binary), fruit characteristics, and insect pollination characteristics (Table 2; Appendix A Table A1). The wood density metric was obtained through a global tree species database for tropical trees [32]. As all species are not yet listed, the wood density was estimated in the following order: (1) the species’ wood density as sampled from Central American (CA) trees, (2) the species’ wood density from a South American (SA) sample, (3) if no species, the average of the wood density from the same genera in CA (1st choice) or SA, or (4) if no congeneric, the standard wood density of trees at 0.60 g cm³ as presented in Chave et al. [3] was used. In terms of fruit characteristics, we compiled fruit size (fruit_size), fruit type (fruit_type: berry, capsule, drupe, arillate seed, pod), fruit accumulation on the ground (fallen_fruit), and records of avian frugivory on each tree taxa. The variable “fallen fruit” applied to trees with a diameter > 10 cm as a designation of maturity. The assigned value was 0 if the tree did not produce fleshy fruit. Exceptions included trees with arillate seeds inside of capsules or pods. Trees with fruits < 1 cm in diameter received a rating of “1”, above that size received a rating of “2”. The variables bird_fruit and ES_prov provided the total number of avian species and the number of threatened species recorded feeding on a tree taxon, respectively [35]. In terms of insect pollination potential for each tree taxa, we tallied the macro-morphic insect groups that are known to visit the flowers of the tree species (pollinators), including Hymenoptera (w-wasps, b-bees, f-flies, a-ants), Lepidoptera (bfl-butterflies, m-moths), Coleoptera (c-beetles), and other (i-other) in cases where unspecified insects were cited as potential pollinators, according to Haber et al. [26]. In addition to these variables, we obtained the International Union for the Conservation of Nature (IUCN) Red List status of each tree taxon (https://www.iucnredlist.org). Because not all tropical trees have been fully assessed, we assigned a value of “near threatened (2)” for locally identified species of concern. Taxonomic names and authorities were verified with the Catalogue of Life [36], Tropicos [37], and World Flora Online [38].

2.4. Ecosystem Functions

We aimed to evaluate the integrity of the restoration actions in this study by determining if the levels of ecosystem functions in the restored forests are advancing toward the levels of the reference forests. To avoid circularity and reduce confounding with measured forest characteristics, we estimated ecosystem functions as independently as possible from our collected metric data, which constrained the set of functions we could estimate. Specifically, we based function estimates on published ecological relationships, such as the number of frugivorous bird species feeding on each tree species [35], documented use of tree species by pollinating insects [26], counts of trees with distinct functional traits (e.g., nitrogen-fixing symbionts or light-colored flowers), and IUCN Red List categories for identifying threatened species [39]. In the case of biomass(C) accumulation, we selected a formula, adapted to the country of study, which used only one measurement (tree dbh), rather than algorithms that utilize up to three traits [40]. All sites were measured with the same formula, creating a comparable estimation among forest stands; estimates are presented at the 100 m² plot level. We describe the estimation methods of each ecosystem function as follows.

Average annual biomass(C) accumulation was calculated using an algorithm developed in Costa Rica by Fonseca et al. [31] that incorporates calculated coefficients and tree dbh (cm), of which we included up to the largest three boles for multi-trunk trees in the calculation of the biomass (C).

$${\mathrm{Biomass}}_{\mathrm{C}}\hspace{0.33em}\left(\mathrm{k}\mathrm{g}\right)=\mathrm{e}\mathrm{x}\mathrm{p}\left(3.41222+2.61148\hspace{0.17em}\mathrm{l}\mathrm{n}\left(\mathrm{dbh}\right)\right)$$

(1)

The tree’s calculated biomass(C) was converted to Mg⁻¹ ha⁻¹, summed for each 100 m² plot, and divided by the number of years of growth to calculate the average annual plot level of biomass(C) Mg⁻¹ ha⁻¹ yr⁻¹

For nitrogen fixation (N_fixation_potential), we assigned a yes/no value of “1” to each species that is a nitrogen-fixing symbiont and “0” to each species that did not (Appendix A Table A1). We included the tree age as an inverse value of N-fixation potential, as a previous study of Inga punctata in these plots found the mean percent detection of nitrogen-fixing DNA decreased steadily with tree age at an approximate magnitude of 60% higher for 4-year old trees than mature I. punctata trees, 40% higher for 8-year trees, and 20% higher for 11-year trees [41]. Therefore, the presence value of each N-fixing symbiont tree was 1 and then multiplied by 1 if mature (>70-years), 1.2 if 12 to 21 years, 1.4 if 8 years, or 1.6 if 3 years. The N fixation potential estimation was a sum of these scores by tree plot, then averaged by stand.

To quantify insect pollination potential in each plot (insect_pollinator_potential), we tallied the number of macro-groups of insect visitors reported for each tree and treelet species and summed that number for each tree with a dbh ≥ 10 cm (combined total of 3 ramets in the case of multiple boles) or treelet. Since bees are of particular importance to the region due to honey and coffee production, an additional point was given to each tree that is bee-pollinated.

A basic insect habitat provision (insect_habitat potential) was calculated for each plot using the sum of trees having light colored flowers (as potential nectar and pollen sources), the sum of the fallen fleshy fruit variable (as an insect forage and reproductive source), and the number of snags/logs as reproductive areas for insects. Leaf litter, an important forest component for insect habitat, was not included due to difficulty in data collection during this time period.

We estimated the avian frugivore forage potential (avian_frugivore_forage) in each plot as the number of birds per fruiting tree species that was recorded by Wheelwright et al. [35] in Monteverde. For example, the study recorded 16 species of birds eating Trema micrantha and 10 species eating Cecropia obtusifolia. For 8 of the 98 species with fleshy fruit, no data were found, so we applied the average of bird species using trees in this study to those taxa. The bird frugivory estimate was summed for the corresponding tree species if it had a combined multiple ramet dbh ≥ 10 cm. Understory fruiting species were always included (e.g., Psychotria, Erythroxylum, Solanum, Hamelia spp).

Finally, we estimated each plot’s threatened species protection (threatened_spp_protection) by categorizing each tree species on a scale of 1 to 5 (1 = least concern, 2 = near threatened, 3 = vulnerable, 4 = endangered, 5 = critically endangered). We added the number of threatened bird species that the tree supports through forage provision using information provided by Wheelwright et al. [35].

2.5. Data Analysis

All statistical analyses were performed in the R programming language [42], and the analyses were selected, coded, and verified by the authors. To assist in identifying coding errors in R, OpenAI ChatGPT (version 5.3) was consulted, and the output was verified.

• Objective 1. Compare the Forest Composition (Diversity and similarities) Along Succession Gradients of Restored Stands

We examined forest composition in two ways. First, we used linear mixed-effect modeling approaches, based on likelihood estimates, to estimate species richness (total species per plot) as a function of forest age. As each of the 13 restored stands had different planting regimes (with different numbers of individuals and taxa planted), we included the number of species planted (spp_planted) as an interactive term in the model, specifically testing if the effects of forest age on species richness depended on the initial number of species planted. We used the package, lme4 [43], to fit the model: spp_richness ~ forest_age + spp_planted + forest_age × spp_planted + (1|stand_id), where spp_richness was the species richness at time of measurement (2023–24) and 1|stand_id was a random factor to account for dependencies of plots that were nested within stands and unmeasured stand-level factors. We used the R package DHARMA [44] to evaluate the fit of the linear mixed model, including QQ plots of residuals, tests of uniformity, and tests of over- and under-dispersion.

Second, we used the R package, vegan [45], to evaluate Bray–Curtis [46] community dissimilarities among stands, based on the maximum count of each tree species observed across plots within a stand. Then, we tested the homogeneity of multivariate dispersion between forest age classes using the betadisper function in the package vegan. Finally, we used the adonis2 function in vegan to conduct an analysis of variance of the Bray dissimilarity distance matrix, testing if the means varied by forest age.

We then used NMDS (non-metric multidimensional scaling) ordination analysis, an iterative, rank-based technique to visualize non-linear similarities, to plot Bray–Curtis community dissimilarity among stands, which focused on the rank order of dissimilarities.

• Objective 2. Compare Stand Structural Characteristics Measured at Different Ages as Compared to a Target Goal of Similarity to Altered Primary Forests (>70 Years)

To identify the overall patterns (variation) in the structural characteristics of the forest stands and ages, we performed a Principal Component Analysis (PCA) with the package FactoMineR [47] on 16 metrics related to tree size, stand characteristics, and understory vegetation (Table 2). We used the FactoMineR::fviz_contrib function to identify those metrics that made above-average contributions to the first two principal components.

Of key interest was how each stand (at the time of its last measurement, 2023) compared to the reference forest stand. We computed PC scores of each forest age stand. To create a global “reference forest”, we calculated the averages of each PC axis with eigenvalues > 1 (n = 3) for reference forests only. We then computed the Mahalanobis distance between each of the 17 stands and the average reference forest [48]. We used linear model approaches to model the Mahalanobis distance as a function of stand age. We used Spearman’s rank correlation to understand how each forest metric was associated with Mahalanobis distance to the reference forest, where metrics with negative correlations indicated that as the variable value increased in value, the Mahalanobis distance to the reference forest decreased (e.g., more similar to the reference forest), while metrics with positive correlations indicated that as the variable increased in value, distance to the Mahalanobis distance increased (i.e., less similar than the reference forest).

• Objective 3. Determine Levels of Six Ecosystem Services Along a Succession Gradient of Restored Forests, and Compare Those to the Levels Occurring in Altered Primary (Reference) Forests (>70 Years)

Each of the six ecosystem functions was estimated at the plot level to analyze the specific ecosystem function and is reported as estimates per 100 m². We repeated the same series of analyses as employed in Objective 2 to evaluate how ecosystem functions varied among stands of different ages. That is, we used PCA to identify patterns of ecosystem function estimates by forest age, followed by estimates of Mahalanobis distances between each age class and the reference forest.

3. Results

3.1. Species Richness and Composition

A total of 1591 trees were measured in the 64,100 m² plots within the 17 forest stands, of which 1538 were identifiable species. Overall, 98 tree species of 40 families were recorded. The naturally dominant Lauraceae family was the most diverse with 15 species represented and was the most abundant in all forest ages (18.9% in the 70-year reference forests and 26.8% in restored stands, as is included in the FCC restoration program). Fabaceae was the second most diverse tree family with nine species; Inga punctata was the most abundant Fabaceae species in the restored stands, while also abundant in reference forests. Sapindaceae was the third most represented family with six species. The rest of the 37 families comprised one to five species in each.

Species richness. The results of the linear mixed model analysis (LMER: spp_richness ~ forest_age × spp_planted + (1|stand_id)) showed that a greater tree species diversity is achieved by planting a diverse set of species at the outset (Figure 2A; see Appendix B 

Table A2. Beta coefficients and significance for the linear mixed model. Fixed effects include forest age and the number of species planted per plot, and their interaction. Random effects included the reforestation stand in which plots were situated.

Variable	Estimate	SE	t	2.5%	97.5%
(Intercept)	13.236	4.534	2.919	4.457	21.779
forest_age	−0.559	0.354	−1.576	−1.226	0.130
spp_planted	−0.650	0.501	−1.297	−1.599	0.346
forest_age:spp_planted	0.082	0.038	2.168	0.006	0.153

for model coefficients). The model adequately fit the data, and there was no evidence of assumption violations as indicated by non-significant tests, including the KS test of QQ plot residuals, outlier test, dispersion test, and test of uniformity (all p > 0.05). The model results indicated that species richness increased both with the age of the forest and with the number of species planted (p < 0.01). However, the interaction term forest age × spp_planted was marginally significant (p = 0.082), indicating a loss of species in the first 3–8 years followed by a continued increase in species richness, attaining reference forest levels in 15 years (if initially planted with 15 species) and 21 years if planted with 10 species. Planting with a few species (5 or fewer) showed a declining species richness over time, yet maintained more species than the original planting due to natural recruitment. However, uncertainty was greatest with both the youngest (3 years) and the oldest (21 years) restored forest.

Species composition. Twenty-one (21) of the 74 species found in the reference forests were exclusive to the reference forest. The restored stands had 78 species (47 surviving species that were planted and 31 new recruits) across all ages, of which 25 were exclusive to the restored stands. The recruited species included typical colonizing edge species and treelets (genera of Conostegia/Miconia (Melastomaceae), Iochroma (Solanaceae), Psidium (Myrtaceae), Trema (Ulmaceae), Cecropia (Cecropiaceae), and multiple treelets of the Asteraceae, Piperaceae, and Rubiaceae families).

The analysis of variance of the Bray dissimilarity distance matrix found that the means varied by forest age (adonis, df = 16, F = 4.4032, p < 0.01). The test of dispersion of the Bray–Curtis dissimilarity between forest age classes was not significant (betadisp, p = 0.30).

A non-metric multidimensional scaling (NMDS) analysis of ranked Bray–Curtis dissimilarity showed that stands separated by the two NMDS axes, where the reference stands (age 70+) had negative NMDS1 values and younger stands had positive NMDS1 values (Figure 2B). The stress was 0.14, indicating a fair fit to the ranked Bray–Curtis data (a linear analysis of the observed similarity and the ordination distance had an R² value of 0.98). The NMDS analysis indicated that the 15- and 21-year forest stands were more similar to the reference forests than the younger forests (Figure 2B). This appears to be associated with the recruitment of 13 shade-tolerant species found in the 15–21-year restoration plots that are typical of older forests of the following families/genera: Arialaceae (Dendropanax, Oreopanax), Bignoniaceae (Trichlia), Boraginaceae (Cordia), Euphorbiaceae (Sapium), Flacourtaceae (Hasseltia), Moraceae (Sorocea), Myrtaceae (Myrcianthes), Sapindaceae (Billia, Cupania, Matayba), and treelets of Clusiaceae (Symphonia) and Rutaceae (Zanthoxylum). Only 4 of these 13 recruited mature forest species were found in forests less than 15 years of age.

3.2. Forest Structure

We measured 16 forest characteristics among stands, and 10 of 16 had the greatest values in the reference forest (Figure 3). Of note, the reference forest had the highest average tree height, dbh, lowest branch height, basal area m², canopy height, strata, and the number of Psychotria plants. Apart from the latter, there was an evident shift toward reference forest levels in these variables once the forest stand reached 15 years of age (Figure 3).

The PCA indicated that PC1 and PC2 explained a large proportion (71.8%) of the overall variation in the 16 forest structure metrics for each of the assessed stands Figure 4 (A; see Appendix B 

Table A3. Principle Component Analysis (PCA) on structural metrics of forest stands. The top section includes eigenvalues and cumulative variation explained for the first 5 components. The middle section provides the principal coordinates, while the bottom section provides standard component loadings.

X	Dim.1	Dim.2	Dim.3	Dim.4	Dim.5
eigenvalue	9.13	2.35	1.57	0.87	0.60
variance.percent	57.05	14.70	9.82	5.44	3.75
cumulative.variance.percent	57.05	71.75	81.58	87.02	90.77
tree_ht	0.94	0.24	−0.18	0.05	0.08
ba_sum	0.90	−0.32	0.11	−0.01	0.08
lowest_branch	0.94	−0.20	−0.04	0.03	0.02
branching_depth	0.66	0.66	−0.27	0.06	0.10
dbh1	0.95	0.14	0.14	0.08	0.02
canopy_ht	0.87	−0.31	−0.24	−0.07	0.18
strata	0.93	−0.07	0.01	−0.01	0.21
groundveg_cvr	−0.69	0.53	0.21	0.23	0.28
veg_ht	0.32	−0.13	0.86	0.12	0.14
canopy_cvr	0.75	0.11	−0.07	−0.48	−0.14
ratio	0.58	0.60	0.32	0.22	−0.25
grass_cvr	−0.78	0.45	0.04	−0.01	0.23
palms	0.71	0.18	0.05	0.34	−0.43
psyc	0.73	−0.40	0.11	0.30	0.23
vines	0.35	0.36	0.58	−0.54	0.01
snags_logs	0.58	0.63	−0.36	−0.03	0.18
tree_ht	0.31	0.16	−0.14	0.05	0.10
ba_sum	0.30	−0.21	0.09	−0.01	0.10
lowest_branch	0.31	−0.13	−0.03	0.03	0.02
branching_depth	0.22	0.43	−0.22	0.07	0.13
dbh1	0.31	0.09	0.11	0.08	0.02
canopy_ht	0.29	−0.20	−0.19	−0.08	0.24
strata	0.31	−0.04	0.01	−0.01	0.27
groundveg_cvr	−0.23	0.35	0.17	0.25	0.36
veg_ht	0.11	−0.08	0.69	0.12	0.19
canopy_cvr	0.25	0.07	−0.06	−0.51	−0.18
ratio	0.19	0.39	0.26	0.24	−0.32
grass_cvr	−0.26	0.29	0.03	−0.01	0.30
palms	0.24	0.12	0.04	0.37	−0.55
psyc	0.24	−0.26	0.09	0.32	0.30
vines	0.12	0.23	0.46	−0.58	0.01
snags_logs	0.19	0.41	−0.28	−0.04	0.23

for PCA loadings and eigenvectors). Across both components, 10 variables contributed more than the expected average contribution. Nearly all variables were significantly correlated with PC1 (p < 0.05), with lowest_branch, dbh, tree height, basal area, canopy height and cover, psyc, and the presence of palms with positive correlations > 0.71, while ground cover and grass cover had negative correlations < −0.68. Four variables were significantly and positively correlated with PC2, including branching depth, ratio, snags and logs, and ground cover (R-square > 0.53).

The reference forests had high values of PC1 and low values of PC2, and the youngest forests had low values of PC1 and slightly negative PC2 values; intermediate ages had high values of PC2 and slightly positive values of PC1 (Figure 4B). Generally speaking, the younger forests scored low on PC1 and were characterized by higher values of ground vegetation and grass cover, as expected. The reference forests were characterized by high levels of stratification, canopy height, total basal area, Psychotria plant abundance, and the average height of the lowest branch on trees. The intermediate forest stands lost understory vegetation and grass cover and were characterized by specific attributes such as larger trees (both height and dbh) with deeper branching depths. The ratio of tree size changed to larger than smaller trees, snags and logs appeared, and palms were recolonizing the stands.

The Mahalanobis distance calculation included PC1, PC2, and PC3 (accounting for over 95% of the total variation in the data) and provided a measure of the multivariate distance of each stand to the average reference forest, with distances decreasing as a general function of forest age (Figure 4C) and where the 15-year and 21-year old stands begin to approximate the reference forest. The linear model fit to the distances was significant (y = 280.31 − −683 · forest age + 349 · forest age²; Adj R² = 0.59; 2, 14 DF, F = 12.8, p < 0.001). A pairwise Spearman rank correlation between Mahalanobis distance (MD) to the average reference forest and each forest metric identified those traits most closely aligned with MD (Figure 4D), with the canopy height, height of the lowest branch, tree height, stratification layers, and the presence of palms as the most indicative of closeness to the reference forest (as these variables increase, MD decreases).

3.3. Forest Function

We measured 6 forest ecosystem functions among stands, which show a general transition toward the reference forest levels, yet with variation (Figure 5). The estimates of four of the five ecosystem functions (average annual biomass(C) accumulation, avian frugivore forage, insect habitat potential, and N-fixation) reached or exceeded the reference forest by 15 years. Three functions, frugivore forage potential, threatened species protection, and N-fixation were notable in the young stand ages and most likely influenced by species that were planted in the reforestation stands. This was also true for biomass(C) accumulation in the youngest-aged forests due to fast-growing trees, both planted and natural recruits. The 21-year forest estimates showed less variation due to a smaller sample size in that age group.

The PCA showed that PC1 and PC2 explained a large proportion (73.1%) of the overall variation in the 6 forest ecosystem metrics for each of the assessed stands (Figure 6A; see Appendix B Table A3 for PCA loadings and eigenvectors). Four functions contributed more than the expected average contribution in PC1: frugivore forage potential, annual biomass(C) accumulation, insect pollination, and insect habitat potential. Five functions were positively correlated with PC1, with R² values > 0.60 (p < 0.05), except for nitrogen fixation, which was null (neither positive nor negative). Two functions contributed more than average to PC2: nitrogen fixation and insect pollination potential, yet only nitrogen fixation potential was significantly correlated with PC2 (R² = 0.88).

The reference and 15-year forests had high values of PC1, while the youngest forests had low values of PC1; PC2 scores did not differentiate by age as clearly as PC1, where the intermediate stage forests tended to have the higher scores (Figure 6B). Generally speaking, the youngest forests were characterized by low values of the ecosystem functions. Intermediate forests were characterized by high levels of nitrogen fixation potential and began to take on characteristics of the reference forests. The reference forests were characterized by high levels of all ecosystem functions except for nitrogen fixation potential and, to a lesser extent, threatened species protection.

The Mahalanobis distance calculation included PC1, PC2, and PC3 (accounting for over 95% of the total variation in the data), and provided a measure of the multivariate distance of each stand to the average reference forest, with distances decreasing as a general function of forest age (Figure 6C) and where the 15-year and 21-year old stands begin to approximate the reference forest. The linear model fit to the distances was significant (y = 18.5 − 24.9 × forest age + 17.0 × forest age²; Adj R² = 0.295; 2, 14 DF, F = 4.35, p < 0.005). A pairwise Spearman rank correlation between Mahalanobis distance (MD) to the average reference forest and each forest metric identified those traits most closely aligned with MD (Figure 6D), with the pollinator potential, annual biomass (C), and insect habitat potential as the most indicative of closeness to the reference forest (as these variables increase, MD decreases).

4. Discussion and Conclusions

In response to the growing call for rigorous evaluations of tropical forest restoration actions, we employed a comprehensive assessment strategy to evaluate the “forest integrity” of a restoration effort in terms of similarity of its forest composition, structure, and ecosystem function to nearby mature “reference” forests. We evaluated a diverse-species, active planting strategy on 13 small-scale restored stands (<3600 m²) of five distinct ages (3–21 years) in Monteverde, Costa Rica. Between 4 and 24 native tree species were planted to restore abandoned pastures, where natural regeneration of herbaceous and tree growth was mostly retained except within a 50 cm radius around the seedlings for two to five years after planting. Active planting restoration efforts can promote natural succession by limiting maintenance, thereby allowing other species and forest features, such as vines, lianas, palms, shrubs, etc., to recolonize the planted stands [29]. Natural recruits are especially important given that certain tree species do not tolerate transplanting or are difficult to germinate and require ecological seed dispersal to colonize areas. Using this collaboration with nature, the restored stands showed encouraging results in their successful transitions into forests with varying levels of diversity, structure, and ecosystem functionality.

Our first general result indicates that planting a greater diversity of tree species begets greater diversity over time. This may seem obvious, but there was an initial loss of species after planting due to seedling mortality and weeding [29], as well as nutrient and light limitations by planted seedlings that may reduce colonization by natural recruits [49]. However, our results showed a quick reversal of this trend with reforested stands reaching a richness level nearly equivalent to the reference forest at 12–15 years of age, depending on the number of species planted (15 or 10 species, respectively). On the other hand, if a plot is planted with only five species, the full recovery is not reached by 20 years. Our observed data from this study align with this finding, as restored stands that were planted with an average of 11 species attained 85.1% of the reference forest species richness by 15 years. On the other hand, several plots with minimal diversity plantings (<5 species) were dissimilar to the reference forests after 12 years. These results contrast with multiple other studies of planted and natural regeneration that included small trees (with <10 cm dbh) in their analyses, which indicated a longer time period of 30–50 years for species recuperation [49,50,51]. While our results show planting a large diversity of trees as beneficial (averaging 13 species per site in our study), it is important to note that many restoration efforts are limited in their ability to do so [52], particularly because the availability of a selection of native tree seedlings at no or low cost is a widespread issue even in countries with well-established nursery systems [53].

Secondly, our analysis of the structural complexity of the different restored stands indicated a distinct shift towards reference forest conditions at the 15-year mark (Figure 3). At this tipping point, over 50% of the 16 observed structural components shifted to the same range as the reference forests: tree height, dbh, canopy cover, branching depth, understory vegetation height, and the presence of palms, vines/lianas, dead logs, and snags. Grass and ground cover decreased, and older forest understory species, such as Psychotria and palm plants, appeared. Other studies have modeled results showing greater stand and individual tree basal area [54], tree density, and canopy cover [49] over time when planted with a diversity of species rather than natural regeneration or secondary succession. Direct seeding strategies with multiple species can result in canopy closure within eight years [55].

Third, regarding ecosystem function recovery, the 15-year stage was associated with distinct increases in five of the six estimated functions, aligning with the results of recent functional tree trait studies that report a recovery at 20 years when assisted or active planting strategies are employed [17,56]. Our estimates of biomass (C) accumulation, avian frugivore forage, insect pollination, insect habitat provision, and N-fixation showed an increasing convergence level to the reference forests at this stage. Nitrogen fixing by symbionts slows with age [57]; therefore, a decrease in this activity aligns more closely with reference forests. Biomass (C) studies show almost 50% recovery of above-ground biomass(C) in less than 15 years [58] or improvement at 15-years [29,59].

The ecosystem contributions of young forest stands should be noted and may depend largely on the number and composition of the seedlings planted. First, planting tree species with specific effective traits, such as planting N-fixing symbionts (I. punctata and other leguminous species), or by planting species with response-traits, such as planting threatened species (e.g., O. monteverdensis, Q. insignis), led to successful outcomes of these two functions. As we identify specific tree and forest traits that are the key drivers of other ecosystem functions, as suggested by Carlucci et al. [11] and Loureino et al. [15], we will also be able to intentionally plant for those results, keeping in mind that landscape functional diversity is key and a diversity of species plantings is intuitively beneficial to promote the collaborative recovery by non-anthropogenic sources.

Additionally, young forests substantially contributed to avian frugivore forage, and are heavily utilized by neotropical migratory birds from November to March [29]. At eight years, the forest was providing insect pollination sources similar to 21-year-old forests, and the N-fixation and threatened species protection capabilities were in place at early stages. While our restoration target is typically to reach older forest levels, which is sensible given the loss of mature forest, forest mosaics of multiple ages may provide a more comprehensive landscape-level functionality and provide an array of ecosystem services necessary for biotic communities. We are encouraged to see the contribution that restoration is making to the landscape.

While our results pertain to active planting restoration efforts only, we fully recognize the debate over active planting versus natural regeneration [60]. Several studies indicate that active planting favors carbon pool and density accumulation faster than natural regeneration [17,61,62]. In addition to planting, a greater biomass accumulation may be helped by complementary resource requirements, method of acquisition, and use-efficiency among the assembly of planted species and their individual trait differences [63]. However, others have shown greater recovery with natural regeneration when time, annual precipitation, landscape scale, and previous land use were controlled [64]. Additional studies show near equivalent recovery of soil carbon stock and microbial communities, including complex carbon degraders, in both natural regeneration and active planting with a diversity of species [41]. However, it is important to note that diverse species plantings are uncommon in active planting restoration studies [52], limiting the ability to determine how this effect plays out in active versus natural regeneration in terms of achieving forest integrity.

Several limitations of the study are worth consideration. First, our sampling was limited to replicates of small plots where multiple variables were held constant (previous land use, soil conditions, seed sources, nursery conditions, and climatic variables). Recognizing that small sample sizes can limit inferences [65] (although some variables appear valid with small samples, such as canopy cover and vegetation height [66]), we compared our stands to reference forests using the same sampling techniques and employed modeling tools useful with smaller and unequal sample sizes. The reference forests were sampled more than the restored forests, as we deemed greater sampling would provide a more valid representation of the older forest systems, to which the younger restored forests were compared, thus providing a more conservative comparison of the two. Secondly, the reference forests were altered primary forests, as there are no remaining primary forests in the study area, with the exception of these fragments that were purposefully left on farms as wind protection and for resource extraction. Thus, the restoration efforts of this region would be considered successful in reaching the characteristics of these altered primary forests.

Third, we used a space-for-time approach in evaluating the effects of age and number of species planted on forest integrity due to the lack of available continuous plot measurements, yet we recognize that greater depth would be attained with a time series repeated measures. Fourth, site-specific caveats are known to influence restoration [51], and this study’s location in the Premontane Wet–Humid Life Zone transition region of Monteverde, Costa Rica, has unique climatic and abiotic conditions. For example, tropical forest succession may be heavily influenced by macroclimatic characteristics [13], and the abundant rainfall, short dry season, and reduced temperature fluctuations are factors specific to this study versus other regions.

A final caveat is both a strength and a weakness. Our study results pertain to a suite of forest metrics and functions that were available to measure across stands, and results may have varied if we had selected a different suite of metrics or functions. In spite of that, our approach of assessing forest structure relies on easily obtainable forest metrics that most practitioners routinely collect. In this light, a recent study is promoting the use of traditional survey methods, which validate a low-cost, manual method of gathering forest measurements [60]. Further, our approach of assessing ecosystem structure rests on published ecological relationships and knowledge of tree traits that pertain to specific functions, which were obtained from independent, literature-based sources and were not duplicated in the forest structure analyses. This avoided circularity with our measured forest metrics but did not produce an absolute value of ecosystem function (e.g., amount of fruit forage or insect habitat use). However, by using the same estimation method for each forest-aged stand, we posit that we can provide a valid comparison across ages. Our location, Monteverde, has a 50-year, rich research history; we have used regional studies, published literature, and open-access databases. These last criteria might be difficult in some regions where multi-taxa and cross-disciplinary studies are not readily available.

Reestablishing a complex forest ecosystem from grass pastures to functioning forests seems daunting, yet our results indicate that active restoration can be successful and we may be able to recover a substantial portion of species diversity, forest structure, and function in relatively short time frames (while recognizing that the attainment of old forest stature, biodiversity, and biomass accumulation truly takes time). The interesting shift in tree structure from an “adolescent” status towards a more mature state at the 15-year mark, and thereby increasing biodiversity, stand characteristics, and ecosystem functions, shows a quick initial recovery towards mature tropical forests when planted with a diversity of species. At the same time, we learned that each forest stage contributes to the overall landscape ecosystem capability, where each forest age had some level of contribution to each of the six studied ecosystem functions, with some functions quite significant for young forests considering their smaller stature. As landscape heterogeneity has been shown to favor the maintenance and enhancement of biodiversity and ecosystem functions [67,68], we show that even when restoration is not rapidly approaching mature forest capabilities, restoration actions at all ages are providing important services to the overall ecosystem well-being.

Lastly, we were able to measure restoration outcomes with standard approaches, thereby promoting participation with on-site restoration practitioners. With a growing database of studies, we will further our ability to either intentionally plant to achieve specific ecosystem functions, restore forests to assist the recovery of biodiversity, or simply to set a solid initial stage for natural, hands-off forest recovery, all of which requires collaboration with nature.