Carbon storage on an elevational
gradient on the west side of the Andes
(Ecuador)
Word count: 21,681
Sebastiaan Van den Meerssche
Student number: 01404649Promotor: Prof. dr. ir. Hans Verbeeck Copromotor: dr. ir. Marijn bauters Tutor: ir. Miro Demol
A dissertation submitted to Ghent University in partial fulfilment of the requirements for the degree of Master in Bioscience Engineering
Deze pagina is niet beschikbaar omdat ze persoonsgegevens bevat.
Universiteitsbibliotheek Gent, 2020.
This page is not available because it contains personal information.
Ghent University, Library, 2020.
PREAMBULE COVID-19
In march 2019 the Belgian government implemented a lockdown to prevent the further spread of the COVID-19 virus. These measures had an impact on this thesis and on a personal level. During the fieldwork in Ecuador wood, leaf and soil samples were collected. The University of Ghent closed the labs during the lockdown, following the measures taken by the government. Consequently, it was not possible to analyze all the samples, hence the planned analyses were only partially done. Soil samples were not analyzed for N and C isotopes as planned. Leaf samples and wood samples were analyzed. This master dissertation was finished on the fieldwork and lab samples that were available.
The lockdown also had an impact on a personal level. The lockdown had an impact on courses and the further completion of this thesis. This in combination with social isolation caused stress. This preamble was drawn up in consultation between the student and the supervisor and was approved by both.
ACKNOWLEDGEMENTS
Working on your thesis is a journey with ups and downs, all of this would not have been possible thanks to a small army of people. A big gratitude goes to prof. dr. ir. Hans Verbeeck who gave me the amazing opportunity and support to go to Ecuador. I am grateful to have dr. ir. Marijn Bauters as copromotor and ir. Miro Demol as tutor, they always helped me with practical issues and feedback on my thesis, which was more than helpful. On the side of Ecuador, I would like to thank prof. Selene Baez for planning the fieldwork and helping us whenever we had questions or problems during our stay. Being invited at Selene’s home at our last night in Ecuador was a nice way to say goodbye, I hope I can ever return the favor. A big thanks to Debbie Eraly from BOS+ who helped us in the beginning with getting to Milpe, where we started the fieldwork.
My Ecuador experience would not have been the same without my thesis buddy Tine. You always helped me with getting up in the morning and communicate with the locals. Your warm personality and patience for me was inspiring, you made the adventure complete!
The fieldwork was though but also the most enjoyable part of this thesis. A lot of people were involved in the fieldwork, our master botanist German Toasa and assistant Stephanie who could name every tree and whose laugh was the most extraordinary sound in the forest. Emma, a bioscience engineering student from Ghent who was on an internship at BOS+ and who never hesitated to help us when we asked for it. The people from the Mindo Cloudforest foundation who helped us during the fieldwork, but also provided us the best after work meals. Eva who brought joy and laughter during the fieldwork.
I will always remember Stephanie, who is a biology student from Quito and helped us during the fieldwork. She invited us to stay at her home near Isla de la Plata, where she introduced us to her friends and where we had some legendary days/nights. She helped me with Tine to improve my un-dos-tres Spanish. We will always be grateful for the time when she and her father helped us to the airport, while massive protests were blocking the roads.
A special thanks to my parents for all the support they gave me during my studies at university, without their support I would not have been able to go to Ecuador.
Sebastiaan Van den Meerssche, June 2020.
ABSTRACT
In this thesis aboveground biomass (AGB) and wood productivity of undisturbed tropical forest was studied on an elevational transect on the West slope of the Andes in the provinces Imbabura and Pichincha, Ecuador. A monitoring network of 20 permanent sampling plots were measured on a transect from 400 – 3200 meters above sea level (masl), spanning from lowland tropical forest, over lower montane tropical forest to upper montane tropical forest. Field inventories of tree diameters at breast height (DBH) were made during the fieldwork in Ecuador. Leaf and wood traits were sampled from species covering 80% of the basal area. Based on DBH, estimated tree height and wood density, the AGB was calculated for the different plots using the allometric equation from Chave et al. (2014). For the first time a recensus was done over the elevational transect, this made the calculations of wood productivity as increment in aboveground biomass possible. The wood productivity was calculated based on previous fieldwork in 2015 by Bruneel and Demol (2016) and fieldwork from this study. The yearly wood productivity was determined as the increment in AGB over the time interval (2015-2019) and divided by the number of years. The aim of this study was to quantify trends of AGB and wood productivity over the elevational transect, and to identify what role abiotic variables and functional traits played in these trends.
Aboveground biomass increased with elevation (from 122.53 ton ha-1 at 400 masl to 226.82 ton ha-1 at 3200 masl) whereas wood productivity decreased (0.42 ton ha-1 year-1 for 1000 m increase in altitude). A plausible explanation for this anomaly could be the difference in forest succession stages over the transect. Higher stem densities (DBH>50cm) were found in the higher stratums, suggesting these forests to be older. The increase in AGB is in conflict with previous studies performed on tropical montane cloud forest (TMCF) in the Andes (Kitayama & Aiba, 2002; Leuschner et al. 2007; Girardin et al. 2010; Malhi et al. 2017; de la Cruz-Amo, 2020). Differences in disturbances over the elevational transect could be an additional explanation for the larger number of big tree diameters in the upper strata. Plots in the upper strata are less disturbed compared to plots in the lower strata. The decrease in wood productivity over altitude could be explained by the decrease in solar radiation at higher altitude, since the occurrence of cloud immersion increases with altitude. This leads to an actual photosynthesis rate lower than the potential photosynthesis rate (van de Weg et al. 2014; Malhi et al. 2017).
Foliar N followed a decrease with altitude, which could confirm the hypothesis that a P/N limitation shift is happening over the elevational transect (Townsend et al. 2008). There was no data of P to confirm this. Foliar C increased with altitude.
SAMENVATTING
In deze studie werd de bovengrondse biomassa en hout productiviteit van een natuurlijk tropisch bos bestudeerd langs een hoogtetransect op de westelijke helling van de Andes in de provincies Imbabura en Pichincha in Ecuador. Een netwerk van 20 permanente staalname plots werd opgemeten langsheen een transect van 400 – 3200 meters boven zeeniveau (masl), variërend van laagland tropisch bos over laag montane tropische bos tot hoog montane tropisch bos. Veld inventarisaties werden gemaakt van boom diameters op borst hoogte (DBH). Blad en hout stalen werden genomen van bomen die 80% van het basale grondvlak bedekken. Op basis van DBH metingen, boomhoogtes en houtdensiteit werd de bovengrondse biomassa berekend voor de verschillende percelen door gebruik te maken van de allometrische vergelijking van Chave et al (2014). Voor het eerst werd een recensus gedaan van het hoogtetransect, waardoor de hout productiviteit kon worden berekend. De hout productiviteit werd berekend op basis van eerder veldwerk door Bruneel en Demol (2016). De jaarlijkse hout productiviteit werd berekend als de toename in bovengrondse biomassa over het tijdsinterval (2015-2019) gedeeld door het aantal jaren in het tijdsinterval. Het doel van deze studie was om trends in bovengrondse biomassa en hout productiviteit over het hoogtetransect te kwantificeren en te identificeren welke rol abiotische variabelen en functionele eigenschappen hebben in deze trends.
De bovengrondse biomassa nam toe over de hoogte (van 122.53 ton ha-1 bij 400 masl tot 226.82 ton ha-1 bij 3200 masl), terwijl de hout productiviteit daalde (0.42 ton ha-1 jaar-1 per 1000m toename in hoogte). Een mogelijke verklaring voor deze anomalie kan het verschil in bosopvolging fasen over het hoogtetransect zijn. Bovenaan het transect werden meer grote diameters gevonden (DBH>50cm), wat er op zou kunnen wijzen dat deze bossen ouder zijn. De toename in bovengrondse biomassa is in strijd met eerdere onderzoeken in tropische bossen langsheen de Andes (Kitayama & Aiba, 2002; Leuschner et al. 2007; Girardin et al. 2010; Malhi et al. 2017; de la Cruz-Amo, 2020). Verschillen in storingen langsheen het hoogtetransect kunnen een aanvullende verklaring zijn voor het groter aantal hoge boomdiameters op hogere hoogte. Percelen bovenaan het hoogtetransect zijn minder verstoord dan percelen onderaan het hoogtetransect. De afname van houtproductiviteit over het hoogtetransect kan mogelijks verklaard worden door de toename van wolkenvorming op hogere hoogte, hierdoor is er een afname van zonnestraling. Dit verklaart waarom de actuele fotosynthese niet gelijk is aan de potentiele fotosynthese (van de Weg et al. 2014; Malhi et al. 2017).
Blad N volgde een afname met de hoogte, wat de hypothese zou kunnen bevestigen dat er een P/N – limietverschuiving plaatsvindt over het hoogtetransect (Townsend et al. 2008). Er waren geen gegevens van P om deze stelling te bevestigen. Blad C nam toe met de hoogte
LIST OF ABREVIATIONS
AGB aboveground biomass
BGB belowground biomass
CUE carbon use efficiency
DBH diameter at breast height
GPP gross primary production
IPCC Intergovernmental Panel on Climate Change
LAI Leaf Area Index
LNC Leaf Nitrogen Content
MAP mean annual precipitation masl meters above sea level MAP mean annual precipitation
MAT mean annual temperature
NEP net ecosystem production
NPP net primary production
POM point of measurement
PSP permanent sampling plot
SLA Specific Leaf Area
TMCF tropical montane cloud forest
WD wood density
Content
1 Introduction 1
2 Literature review 2
2.1 Forest ecosystems 2
2.1.1 Tropical forests 2
2.1.2 Tropical montane cloud forest 2
2.2 The global carbon cycle 4
2.3 Carbon dynamics in the tropical forest 6
2.3.1 Net primary production 7
2.3.2 Quantification of forest biomass 8
2.3.3 Relating production with forest fluxes 10
2.3.4 Factors determining the carbon storage in tropical forests 10
2.4 Elevational transects in Tropical forests 12
2.4.1 Net primary production along an elevational transect 12
2.4.2 Tree characteristics on an elevational transect 13
2.4.3 Functional traits 13
3 Material and Methods 16
3.1 Study area 16
3.1.1 Climate 18
3.1.2 Forest description 19
3.1.3 Field protocol 21
3.1.4 Permanent Monitoring Plot establishment 21
3.1.5 Marking, measuring and counting of trees 22
3.1.6 Leaf sampling 23 3.2 Lab analysis 24 3.2.1 Leaf samples 24 3.2.2 Wood samples 25 3.3 Data Processing 25 4 Results 28 4.1 Stand structure 28
4.2 Aboveground biomass and fluxes 30
4.4 Linear mixed effect models 37
5 Discussion 38
5.1 Stand structure 38
5.2 Aboveground biomass and fluxes 39
5.3 Isotopes 43
5.4 Individual tree growth along altitude 44
5.5 Implications 45
6 Conclusion 46
7 Recommendations for future research 47
8 Bibliography 48
9 Appendix 62
1 Introduction
In the Belgian summer of 2019 large areas of the tropical forest in Brazil were on fire, emitting large amounts of CO2 in the atmosphere. This made people once more conscious about the threat these ecosystems suffer. Global warming will be the greatest challenge for humans in the nearby future and the only way to stop it, is to reduce emissions and sequester carbon from the atmosphere. Knowledge on carbon storage in tropical forests is essential to determine the role of these ecosystems in the global carbon cycle. However, climate change itself will have a profound feedback on these ecosystems, which are of today poorly understood. Tropical montane cloud forests (TMCF) could be a solution to this problem, they could be part of a bigger plan regarding forest conservation.
The importance of tropical forests in the global carbon cycle is recognized worldwide (Intergovernmental Panel on Climate Change, 2007). However, the carbon dynamics of TMCFs have only started to be explored since the last decade (Girardin, 2010). These ecosystems could provide valuable information on the influence of environmental controls (temperature, precipitation and radiation) on ecosystem productivity and carbon cycling (Malhi et al. 2010). The constant growing season and low climatic variability of TMCFs make them the ideal experimental setup to study carbon dynamics. In the scope of this thesis an elevational transect was studied on the West slope of the Andes in Ecuador, near the provinces Imbabura and Pichincha. In total, 20 permanent sampling plots (PSP) ranging from lowland tropical rainforest at 400 meter above sea level (masl) to tropical montane cloud forest at 3200 masl were inventoried.
The primary objective of this study was to evaluate productivity and aboveground biomass along an elevational transect and to make a link with functional traits and abiotic variables. Leaf and wood traits were sampled from species covering 80% of the basal area (see parallel thesis by Tine Bommarez, 2020). The main question for this thesis was: What is the trend of productivity and aboveground biomass (AGB) over the elevational transect and what are possible drivers behind these trends?
This thesis is part of a greater research project within a collaboration of the University of Ghent, the Escuela Politecnica Nacional in Quito and BOS+. The elevational transect was constructed by former students Marijn Bauters and Matthias Strubbe from the University of Ghent. Multiple students studied this transect in the scope of their thesis. This thesis was continued on their work and references are made to their research in this thesis. This study begins with a literature review, that provides some basic knowledge regarding the topic. A detailed description on the field protocol, lab analyses and data processing are provided in the section material and methods. Results are reported and discussed, and finally a conclusion was made.
2 Literature review
2.1
Forest ecosystems
Forests play a major role in the carbon cycle through photosynthesis and thereby contribute in moderating the amount of carbon released in the atmosphere by human activities. Sequestering carbon from the atmosphere mitigates the build-up of greenhouse gasses in the atmosphere (Birdsey, 1993). Forests can take up and release carbon (Houghton, 1996). However, overall the Intergovernmental Panel on climate Change (IPCC) estimates that terrestrial ecosystems have a net uptake of carbon (C) between 2.0 to 3.4 Pg C each year. When a forest is retaining more carbon then it is releasing, it is called a sink (e.g. secondary forest). While a net release turns them into a source (e.g. forest fire). Different management strategies are in place to promote the storage of carbon in forests (Barbati, 2006), e.g. the prevention of forest fires, diseases or deforestation and even planting new forests (Percy et al. 2003).
2.1.1 Tropical forests
Tropical forests contain 55% of the global terrestrial carbon stock and span an area of over 20 million km2 (Pan et al. 2011). Therefore it is important to understand these ecosystems and feedback processes of changing atmospheric conditions. Today, tropical forests globally are subjected to a wide range of anthropogenic disturbances, with large areas being deforested for agricultural land or grazing area for cattle (Wright, 2005). Coincidently the biodiversity of these forests are extremely high in comparison with other forest types, harboring nearly 50% of the world species (Lugo, 2009). Additionally, tropical forests are being subjected to climate change with increasing temperatures, frequency of extreme droughts, heatwaves and storms. The combination of climate change and anthropogenic disturbances is leading to a global decline of tropical forests (Betts et al. 2004).
2.1.2 Tropical montane cloud forest
Tropical montane cloud forest (TMCF) is a type of tropical forest found at an elevation of 1600-2300 masl (Foster, 2001). It is a highly threatened ecosystem, as it is estimated that already 90% of this ecosystem is lost in the Andes (Sylvester, 2017). TMCFs only represent a small fraction of 2.5% of the tropical forests in general, and it is estimated that only 0.14% of Earth’s surface
represents TMCF. They can be found in the Andes, Central America, Caribbean, south and east Asia and Central Africa (Hamilton, 1995). The total potential area of TMCF is estimated to be around 380 000 km2 with most of the area in the Neotropics (Bubb et al. 2004). TMCF are known as mossy forests because of the vegetation which is covered by lichens. Lichens prefer growing on this altitude because of the high relative humidity (Hamilton, 1995).
From a meteorological viewpoint, TMCFs are characterized by a high relative humidity and development of clouds. The high relative humidity makes the TMCFs unique in their hydrological functions (Bruijnzeel, 1995). High relative humidity leads to cloud deposition through contact with vegetation surface and soil, a phenomenon commonly referred to as ‘horizontal precipitation’ (Stadtmüller, 1987), creating favorable growing conditions for epiphytes such as mosses and ferns (Bubb et al. 2004). During the dry season, when vegetation is experiencing drought stress, the horizontal precipitation surpasses the vertical precipitation (Bruijnzeel, 1990). TMCF play an essential role in the hydrological balance at regional scale and in maintaining the biodiversity that they support (Richards, 1984). Indeed, these forests provide drinking water to metropoles like Quito, hence supporting a high number of people (Bubb et al. 2004). The rainfall in TMCF can range between 600 – 8000 mm/year, with a range in average temperature from 7 to 21 °C (Holder, 2004). Soil moisture content is relatively high in TMCF due to its high precipitation. This high water content in combination with reduced solar radiation (clouds) and the low rates of decomposition lead to very acid soils in the TMCF. These acid soils will lead to a high buildup of humus in the upper soil layer (Hamilton et al. 1995).
The characteristics of the TMCF change with increasing altitude (Figure 1) with the transition between two different plant communities being called ‘an ecotone’ (Foster, 2001). Along elevation a differentiation can be made in Lowland Rainforests (LRF), Lower montane Rainforests (LMRF) and Upper Montane Rainforests (UMRF). This differentiation is due to the cloud formation in the mountains, clouds are formed in the LMRF and are more pertinent in the UMRF (Grubb et al, 1963). The elevational cloud formation shows a variation since it depends on the moisture content in the air, the wind velocity and the different cloud formation processes (Foster, 2001). Above the cloud forest the subalpine forest can be found, with generally only small trees - not higher than 9 m – and with typical small leaves (Foster, 2001).
In comparison with lowland tropical forest, trees in the TMCF are generally more crooked, lower-statured and grow slower (Bruijnzeel, 2001). These characteristics are influenced by the UV-B radiation. Vegetation like TMCF at higher altitude receive more UV-B radiation and it is known from lab experiments that plants grown under higher amounts of UV-B radiation tend to grow stunted with shorter internodes, small thicker leaves and more pigments (Flenley, 1995). Trees in the TMCF can intercept moisture out of the air, with drops being formed on leaves which are
partially diverted to the ground. This happens when small droplets are formed which coalesce to bigger drops. This process on a large scale can impact the hydrological cycle of the tree (Hamilton et al. 1995). A special form of cloud forest is called an Elfin forest or Dwarf forest. This ecosystem is located on high altitudes with poor soils, the forest is characterized by small tree species (Foster, 2001).
Figure 1: Representation of an elevational gradient in a tropical mountain forest, tree stature and leaf size decrease from the Lowland Rainforest to the more Upper Montane Rainforest. The Cloud Forest occurs under influence of the presence of clouds in the Lower Montane Rainforest or the more Upper Montane Rainforest. Figure adopted from Foster (2001).
2.2
The global carbon cycle
The carbon cycle encompasses the ensemble of processes in which carbon is exchanged between biosphere, geosphere, hydrosphere and the atmosphere of Earth. There are three major pools of carbon; the atmosphere, the ocean and the terrestrial biosphere (Falkowski, 2000) (Table 1). From the total terrestrial biosphere, approximately 45 % is stored in forests (Bonan, 2008). The current carbon stored in the forests is estimated at 861 Pg C with 383 Pg C stored in the soil to a depth of 1 m, 363 Pg C in living organisms, 73 Pg C in deceased wood and 43 Pg C in debris/litter on the topsoil. The three mayor forest types are tropical forest, boreal forest and temperate forest (Table 2). Tropical forest is the biggest carbon sink of these three. The share of carbon stored in soil and biomass of tropical forest shifts for boreal and temperate forest.
The delicate equilibrium between the three major pools is in danger because of the release of carbon by human activities (Schimel et al. 2001). Carbon dioxide is the most important greenhouse gas emitted in the atmosphere. Methane is a second carbon-based greenhouse gas, which absorbs 30 times more energy than carbon dioxide, but has a lower life time than carbon dioxide, making it a less important greenhouse gas than carbon dioxide (Forster, 2007).
Carbon dioxide is used by plants in photosynthesis, a process in which organic molecules are assimilated out of carbon dioxide. This way, carbon dioxide is entering the terrestrial and oceanic biosphere. Additionally, CO2 dissolves in the ocean, where it contributes substantially to the acidification of the ocean (Ito, 2000). Due to anthropogenic influences, the balance of exchange in carbon between the atmosphere and the ocean is disturbed (Schimel et al. 2001). Oceans are getting warmer, which leads to a decrease in CO2 solubility, in order to remain balance the sequestration of phytoplankton needs to be higher (Falkowski, 2000).
Tropical forests play an important role as carbon sink (Phillips, 1998). This carbon sink is reducing its strength because of disturbances such as deforestation, which is predicted to emit 1.5 Mg C each year for Amazonia. The tropical forest is not capable in reclaiming this amount of carbon by regrowth (Hubau et al. 2020). The importance of tropical forests in the carbon cycle is becoming more indisputable for governments around the world and new local initiatives are helping to protect tropical forest against deforestation by making people conscious about the importance of these forests.
Table 1: Distribution of carbon (C) in different pools. Table composed from Falkowski, (2000).
Different pools Quantity C (Gt) % of total non-lithosphere
Oceans 38400 81.3 Fossil fuels 4130 8.7 Atmosphere 720 1.5 Aquatic biosphere 1.5 0.003 Terrestrial biosphere 2000 4.2 Living biomass 600-1000 1.8 Dead biomass 1200 2.5 Total non-lithosphere 47250 100 Lithosphere 75000000 1587.3
Table 2: Representation of the three mayor forest types and the carbon stored in their ecosystem, relative Carbon in soil and biomass are also represented. Table composed from Pan et al. (2011)
Carbon stock (Pg) Relative share (%) Carbon in soil (%) Carbon in Biomass (%) Tropical forest 471+/-93 55 32 56 Boreal forest 272+/-23 32 60 20 Temperate forest 119+/-6 14 69 31
2.3
Carbon dynamics in the tropical forest
Carbon dynamics can be described by the Gross Primary Production (GPP) (g C m−2 yr−1) which is defined by the photosynthesis and is the total rate of carbon sequestration in an ecosystem. The Net Primary Production (NPP) (g C m−2 yr−1) equals the difference between the GPP and the autotrophic respiration/metabolism (Ra) (g C m−2 yr−1) and hence its lower value which reflects the amount of carbon stored as biomass (Clark et al. 2001).
NPP = GPP − Ra (1) Measuring NPP is challenging because of diverse measurements involved in the calculations, GPP cannot be measured directly and defining plant respiration on a community level is difficult because of the large size of forest dominants and the complexity in forest communities (Lavigne et al. 1997).
The Net Ecosystem Production (NEP) (g C m−2 yr−1) also takes into account the heterotrophic respiration Rh (g C m−2 yr−1). This is the loss of carbon by decomposition of organic material.
NEP = NPP − Rh (2) NEP is determined by the balance of NPP and the heterotrophic respiration (Ohtsuka, 2007). This requires measurements of carbon stored in a community over time. NEP is then calculated as the net amount of carbon stored in the vegetation over time. To use this approach quantification of biomass is necessary, such as aboveground biomass (leaves, branches, stems, new and dead trees) and belowground biomass (roots, litterfall, leaching nutrients) increments (Clark et al. 2001).
2.3.1 Net primary production
NPP is described as the new organic material assembled in a certain time interval. A distinction is made for NPPAG (aboveground productivity) containing stem, branches, leaves and fruits and NPPBG (belowground productivity) containing fine and coarse roots. Aboveground productivity, NPPAG can be written as the following equation:
NPPAG= NPPStem+ NPPCanopy+ NPPVOC+ NPPBranch (3) With NPPStem, Branch respectively the change in biomass of tree stems and branches, NPPCanopy the change of biomass in canopy containing leaves, fruits and flowers (Girardin, 2010). Aragao et al. (2009) estimated a ratio of 0.4 from values in lowland Amazonian plots to estimate NPPBranch based on NPPStem (Aragao et al. 2009). NPPVOC are the emissions from volatile organic compound production into the atmosphere (e.g. toluene, benzene, xylene,…), this is only a minor part of the total NPPAG and can be neglected for tropical forest (Malhi et al. 2009). NPPBG can be written as the following equation:
NPPBG = NPPFine roots+ NPPCoarse roots+ NPPExudates (4) With NPPFine roots, Coarse roots, exudates respectively the change in biomass of fine roots, coarse roots and exudates. Estimating the different parts of NPPBGB is difficult, NPPFine roots can be estimated using rhizotron cores, these work as an observation chamber where monthly growth can be observed (Girardin, 2010). Malhi et al. (2009) proposed the following equation to estimate NPPCoarse roots based on NPPStem:
NPPCoarse roots = 0.21(±0.03) × NPPStem (5) This equation is based on values observed in lowland tropical forest, using this equation for TMCF could result in underestimations, since trees in the TMCF at steep slopes invest more in coarse roots (Girardin, 2010). Belowground components are often ignored since measurements of coarse roots, fine roots and exudates are difficult, hence the introduction of aboveground NPP (ANPP) (Gower et al. 2001).
2.3.2 Quantification of forest biomass
Assumptions are necessary to determine NPP, since direct measurements are not possible. AGB and BGB are a solution for this problem. Determining AGB and BGB by destructive measurements such as deforestation and measuring dry weight is time consuming and not cost effective. Allometric equations were introduced to overcome this problem (Brown, 1997). These are simple equations used to calculate AGB based on tree parameters such as tree height, diameter at breast height (DBH) and wood density. Forest inventories are necessary to go from individual tree level to plot level to ecosystem level (Clark et al. 2001). The increment of AGB in a certain time interval is defined as wood productivity. Tall trees contribute significantly to the AGB and wood productivity at forest stand level (Bastin, 2018). Brown (1997) estimated that on average less than 3% of the AGB of an ecosystem in a tropical forest comes from understory vegetation. The amount of sunlight that can reach understory vegetation is very low, making the contribution of understory vegetation in tropical forests negligible (Brown, 1997). In boreal and temperate forests, the understory vegetation cannot be neglected since the overstory vegetation is more open, making the understory vegetation an important part of the AGB (Gower et al. 2001). In tropical forests, however, the AGB and wood productivity is typically estimated based on trees with a diameter > 10 cm alone, since the AGB and wood productivity on a community level is determined for over 90 % by these trees (Brown, 1997).
Wood productivity is calculated as the increment of AGB over a certain time interval (Figure 2). Two methods exist to calculate the wood productivity. In the first method trees that die are assumed to have no wood productivity and are thus ignored in the calculations. Increment in AGB is determined for living trees. When a new tree is added, its increment is calculated as the difference between its AGB and the minimum AGB of a tree to be measured e.g. trees with a diameter of 10 cm. In the second method, the biomass of trees that died are estimated and added to the stand increment. New recruits are incorporated by multiplying the number of new trees by their minimum tree size and subtracting this in the stand increment (Figure 2). Yearly measurements are recommended if the time interval is bigger than two years since the biomass of dead trees are estimated from their initial diameters, if not, an underestimation of the wood productivity will happen since the last measurement for trees that died could be more than 2 years old. (Clark et al. 2001). The second method is used for large field campaigns and/or when the time interval is large (Schulze et al. 1999). When dead trees are ignored in the calculations, the underestimation of increment in AGB will be around 1-3% of initial AGB each year of the interval (Clark et al. 1994). New recruits that die in this time interval can enlarge this underestimation. Missing trees can also lead to an underestimation of wood productivity, especially in smaller study areas where large trees went missing (Clark et al. 2001).
Figure 2: Representation of two methods to calculate the wood productivity as the addition of aboveground biomass (AGB) over a certain time interval at the start t1 and the end t2. Approach 1 is based on individual tree
measurements. Approach 2 is based on measuring the stand stock at t1 and t2, and involves measuring biomass
from trees that died and from new recruits. New recruits are trees that outgrow the minimum diameter size of measurement. Figure adopted from Clark et al. (2001).
Large trees (DBH > 50 cm) can lead to an under/overestimation of the wood productivity (Gonzalez et al. 2018). In tropical forests, large trees from this size are not frequent, but they can be responsible for 25-50% of the AGB of a small study area (Brown and Lugo, 1992; Brown et al. 1995; Clark and Clark 1996). Wrong estimates for these trees are obtained if the allometric equation used to estimate its AGB was not built for this size of trees (Brown and Lugo, 1992; Brown et al. 1995). Trees can lose biomass (branches) by huge storms and hurricanes, which could lead to an overestimation of AGB and wood productivity. This can be included in allometric equations by making these relationships for a representative area where some trees already suffered damages. If allometric equations include such elements, an evaluation of the AGB increment is recommended by making a mass balance (Clark et al 2001). Wrong measurements of DBH can lead to wrong estimates of AGB as well e.g. buttressed trees should be measured above the point of measurement (POM), otherwise an overestimation of the tree AGB will be obtained. The allometric equations calculate AGB as a cone based on DBH and tree height, when diameters are measured at buttress trees, these cones are much wider than in reality, leading to an overestimation of the AGB (Brown, 1997; Clark et al. 2001). Wood density is a good predictor for AGB in allometric equations next to DBH and tree height, therefore it should not be neglected (Baker et al. 2004).
Belowground biomass increment is still poorly understood. These measurements contain coarse and fine root increments, mortality, losses to consumers and exudates. Similar as AGB, allometric equations can be used for relating tree diameters with BGB, but the uncertainty on these equations is high (Kuyah, 2012).
2.3.3 Relating production with forest fluxes
So-called eddy covariance can be used to estimate carbon exchange of a forest with the atmosphere (Schulze et al. 1999). This is a technique to measure vertical turbulent fluxes of CO2 between a forest ecosystem and the atmosphere. By measuring windspeed, direction and analysing the CO2 concentration at high frequency, an instantaneous CO2 flux can be calculated (Barr, 2006). In addition to measuring eddy covariance at a tower to estimate net ecosystem exchange (NEE), biomass measurements of the forest at a ground level can provide additional valuable information. A relationship exists between NEE (g C m−2 yr−1) of CO
2 with the atmosphere and NPP (g C m−2 yr−1) and GPP (g C m−2 yr−1) (Saigusa, 2002) :
NEE = NPP − Rh (6) NEE = GPP − Rec (7) With Rh the heterotrophic transpiration and Rec the ecosystem respiration (sum of AGB respiration and CO2 released by the soil) (Saigusa, 2002). A positive NEE means that more carbon is released in the atmosphere than incorporated by the forest. NEE is a small difference in these fluxes, nevertheless, it can be an indication for NPP. A negative NEE should lead to an increment in AGB and/or BGB. If this is not the case, then a negative NEE could be explained by an increase in soil carbon (Clark et al. 2001).
2.3.4 Factors determining the carbon storage in tropical forests
AGB is the largest carbon pool in the tropical forest (Gibbs et al. 2007). AGB and NPP can have variations due to changes in abiotic factors (temperature, precipitation, radiation) (Quesada et al. 2009). Understanding how the forest ecosystem reacts to these changes can help improve carbon storage, by improvements in forest management. Tropical forests along an elevational transect are the perfect setup to study the influence of these changes on ecosystem functions, because of a constant growing season and the absence of seasonal variation (Malhi et al. 2010; Sundqvist et al. 2013). A distinction can be made for environmental controls (radiation, temperature, precipitation and soil) and biotic controls (forest structure and composition) (Figure 3).
Environmental conditions have a direct control on the productivity (GPP and NPP) (Fyllas et al. 2017).
Figure 3: Relationship of environmental control and biotic control on productivity. Radiation, temperature and precipitation can have a direct impact on the GPP and NPP (black arrow) but also an indirect effect by influencing the forest structure and the composition (grey arrow). Forest structure and composition can also have a direct effect on GPP and NPP. Forest structure and composition are not only influenced by environmental conditions, the biogeography of a region has also a big influence. Only composition (traits) and radiation (full black line) explain the elevational gradient in GPP and NPP in a study performed by Fyllas et al. (2017), where the productivity was investigated along an elevational gradient in the Amazon-Andes. Figure adopted from Fyllas et al. (2017).
2.3.4.1 The influence of climate on tropical forests
Temperature has an influence on the photosynthesis and respiration of plants (Berry and Bjorkman, 1980), but also on the microbial activity in soil and the decomposition of litter (Swift et al. 1979). In tropical forests, temperature remains more or less stable throughout the year (Houghton et al. 2001). Nevertheless, small changes in temperature can have major influences on the ecosystem (Swift et al. 1979). In a study along an elevational transect in Mauna Loa, Hawaii, the decomposition of litter increased with temperature. In plots located at high altitude the soil organic matter was higher whereas temperature and aboveground net primary production was lower (Raich et al. 1997). In another study performed by Raich et al. (2006) a meta-analysis was done to evaluate the influence of mean annual temperature on carbon fluxes. Litter production, tree growth and belowground carbon allocation increased with mean annual temperature (Raich et al. 2006).
High rainfall (>3000mm/year) can lead to a reduction in NPP. The excess of moisture induces a decrease in soil oxygen which results in a decrease of mineralisation and excess of nutrients leaching (Clark et al. 2001; Schuur et al. 2001). Tropical rainforests with a lower rainfall (<3000mm/year) are less covered by clouds and experience more solar radiation, hence they experience an improvement in photosynthesis (Linger et al. 2020). The lower rainfall also improves
soil aerobic conditions, which promotes nutrient uptake by plants and thus improves the wood productivity (Schuur and Matson, 2001).
2.3.4.2 The influence of fertility on tropical forests
The effect of soil nutrients on the tropical forest carbon cycle is less consistent since N and P concentrations are reported to give negative as well as positive effects on the aboveground biomass (Quesada et al. 2010). A positive effect is the rapid increase of NPP, but high turnover rates are a negative effect since it is leading to less carbon storage (Baraloto et al. 2011).
2.4
Elevational transects in Tropical forests
Elevational transects of tropical forests provide a setup to study change of ecosystems in terms of temperature and climate variables (Malhi et al. 2010). Temperature decreases approximately 0.6°C per 100 m increase in altitude (Bush et al. 2004). Beside temperature other variables change with altitude: atmospheric pressure, solar radiation and cloudiness. Other variables such as wind and topography are less dependent on height above sea level (Körner, 2007; van de Weg, 2010). Elevational transects in tropical forests are getting more interest because they are susceptible to global warming. Species living in the tropics have a smaller thermal niche compared to species in temperate forests (Janzen, 1967), this makes them more sensitive to global warming (Deutsch et al. 2008). When temperature rises, species with a small thermal niche will become unfit and will migrate to higher altitudes where their thermal niche fits the local temperature. Species from the lowland tropics will move upward to the montane tropics (Malhi et al. 2010). This is already noticeable. Species who were prominent in the lowland tropics are now seen as montane species (Liu & Colinvaux, 1985; Bush et al. 1990).
2.4.1 Net primary production along an elevational transect
Different researchers have studied the AGB on elevational transects in TMCFs in Ecuador, Peru, Venezuela, Panama and Hawaii. All these studies showed a decline of AGB along the elevational transect, making this a typical aspect of TMCF (Kitayama & Mueller Dombois, 1994; Delaney et al. 1997; Leuschner et al. 2007). Less studies have been performed on BGB along an elevational transect (Girardin et al. 2010). Roderstein et al. (2005) investigated BGB along an elevational transect in the TMCF in South Ecuador, but did not compare his results with AGB, the study concluded that BGB increased along the elevational transect (Roderstein et al. 2005).
Girardin et al. (2010) did a study on an elevational transect in the Peruvian Andes. A fast decline of AGB was seen until the cloud immersion zone (1500 m). In the cloud immersion zone was no trend. BGB showed a decrease with the lowest value in the cloud immersion zone. The partitioning of AGB and BGB showed no clear trend in elevation, but a high variety between plots was seen. Girardin et al. (2010) concluded that AGB was the most important carbon stock over the elevational transect.
2.4.2 Tree characteristics on an elevational transect
Tree height is one of the first noticeable changes on an elevational transect. In the lowland tropics trees reach a height of 40 m (Sullivan, 2018), whereas trees in the highlands (4000 masl) reach a height of 2 m (Grubb, 1977) and grow slower (van de Weg et al. 2009). At higher altitude trees have thicker leaves, are more stunted and have lower internode distances (Jose et al. 1996). Trends in leaf traits and nutrient ratios are less visible, but provide valuable information on ecosystem functioning and biogeochemistry (Wright et al. 2004).
2.4.3 Functional traits
Functional traits are used as an environmental proxy. They describe the response of plants to the environment and provide valuable information on ecosystem functioning and biogeochemistry. Canopy traits are easy accessible and show rapid variations with the environment (Asner et al. 2015).
2.4.3.1 Leaf nitrogen content and specific leaf area
The specific leaf area (SLA) is calculated as the ratio of one sided leaf area to the leaf dry weight. SLA is an indication for the leaf area a plant is able to grow with a specific amount of leaf biomass e.g. thick leaves have a small SLA (Poorter, 1999). The leaf nitrogen concentration (LNC) in the foliage is an indication for the photosynthesis of the plant (Westoby, 2006). These traits are linked with the biogeochemistry of the location (Asner et al. 2015) and thus the canopy chemistry along an elevational transect is a response to nutrient availability (Bauters et al. 2017). Species with a high LNC and SLA are plants with a high photosynthesis capacity (Poorter at al. 2009), but these species have a faster leaf turnover (Bauters et al. 2017).
2.4.3.2 Nitrogen isotopes
Around 0,37% of the total N is in the form of 15N. The ratio 15N:14N can help understand processes which are difficult to measure. The N isotope composition is calculated by the following formula and is written in δ notation and expressed in per mil (%°) (Coplen, 2011).
δ15C = (15N14N)sample
(15N14N)standard − 1 ‰ (8)
with (15N14N)sample the isotopic ratio of the sample and (15N14N)standard the N isotopic composition of a standard. The standard sample is atmospheric N, which is equal to 0 %° (Craine at al. 2015). N is a key element for plants, it is the most limiting resource used in proteins, DNA and photosynthesis (Thomas et al. 2013). Due to multiple molecular states of N in ecosystems (ammonia, organic acids cyanides,…) and the complex interactions with other biogeochemical processes such as demineralisation because of weathering, knowledge of the nitrogen cycle is key to understand TMCF ecosystem functioning, notable plant productivity, C sequestration and fluxes (Galloway et al. 2008; Craine et al. 2015). The ratio of N isotopes in soil is more stable than in plants, where more short-term variation is expected. This is the reason why plants are a good indicator for short term variations in the N cycling of an ecosystem (Craine et al. 2015). Leaves are a good representation for δ15N of plants. 95% of plant leaves are in the range of −7.8 ‰ to 8.7 ‰ (Craine et al. 2009, 2012). The variation of δ15N for plant tissue such as roots, leaves and stems are low (Kolb and Evans, 2002). Dijkstra et al. (2003) reported a study where the difference in δ15N for root and plant tissue for forests in North America was less than 1‰ (Dijkstra et al. 2003).
Variation in foliar δ15N at different locations is influenced by different processes: N can be disposed by rainfall or geological factors. Rainfall can dispose N directly on leaves leading to an increase in δ15N (Houlton and Bai, 2009). Geological N can come available by weathering of rocks. Approximately 99.9% of the fixed N on earth is present in rocks (Capone et al. 2006). The N available in rocks is organic N, NH4+ and in smaller amount NO3- (Craine et al. 2015). Other factors influencing foliar δ15N variation are mycorrhizal fungi living in symbioses with plants and N-fixation (Craine et al 2015). Plants who rely on N-fixation have usually a δ15N value of 0 ‰, reflecting the atmospheric isotopic nitrogen (Handley, 1992). Mycorrhizal fungi prefer δ14N over δ15N which leads to an enrichment of δ15N (Craine et al. 2009). Another factor leading to an enrichment of δ15N is a high concentration of NO3- in the soil leading to a higher denitrification (Craine et al. 2015).
2.4.3.3 Carbon isotopes
Similar like δ15N, two stable carbon isotopes exist 13C and 12C. The ratio of these isotopes can be written as (Tieszen, 1989):
δ13C = (13C12C)sample
(13C12C)standard− 1 ‰ (9)
with (13C12C)sample the isotopic ratio of the sample and (13C12C)standard the isotopic composition of a standard. The ratio of these two isotopes in a sample tissue can help determining carbon sources. Organic material ranges from -50 ‰ to -5 ‰ (Oikawa, 2018). The major factor influencing the ratio in organic matter is the biological fixation of 13CO
2, the transfer of inorganic carbon in the atmosphere to organic carbon. Land plants have a lower δ13C value than atmospheric CO2 with a δ13C value of -7 ‰ because of the biological fixation process of photosynthesis (Deines, 1980). An important enzyme active in the photosynthesis pathway is rubisco, it catalyses the reaction where inorganic carbon is transformed to organic. 13C and 12C are competing with each other for an active site on rubisco, but 12C is favoured by rubisco because of the lower activation energy (Arens, 2000). The δ13C value of plant leaves can also provide information regarding the WUE (water use efficiency) of the plant. When plants have enough water available and the WUE is low than the stomatal conductance is maximised. When the stomata conductance is high, atmospheric CO2 enters without any problem, which leads to a maximised biological fixation of inorganic carbon to organic carbon, hence the low value of δ13C (Keller, 2017).
2.4.3.4 Wood density
Wood density is less studied than leaf traits, but it is an important variable in allometric equations to estimate AGB. Wood density is characterised by tree species and local factors. Some tree species invest more in nutrients for lifting its canopy over other trees to reach a maximum amount of sun. Trees that grow on hills will invest more in thicker wood tissue at the downslope side to prevent it from falling down (Chave et al. 2009).
3 Material and Methods
The practical work can be divided into two parts. In the first part fieldwork was performed in Ecuador, inventories were performed and samples were collected on different altitudes along the elevational transect on the west side of the Andes in Northern Ecuador (Figure 4). The second part of the practical work is the processing of all the samples in the lab of ISOFYS, at the faculty of Bioscience engineering at the Coupure in Ghent. In this chapter, every part from inventory, data sampling to processing will be described in detail.
3.1
Study area
Figure 4: The left map represents Ecuador and the right map is a close up on the study area with the five strata, a line is drawn following the elevational transect. The altitude is increasing from the left to the right. The strata are located in the provinces Imbabura and Pichincha. Map data from google maps.
In previous field campaigns performed by CAVE lab (Ghent University), 17 permanent sampling plots (PSP) were set up on four altitudinal strata in undisturbed natural forest, each stratum had at least four plots on similar altitude (Table 3). The location of these strata were chosen to span an elevational transect in the TMCF. Next to the existing four strata, a new location El Cedral was added with existing plots. The plots had a 40 by 40 m dimension, with exception for El Cedral containing 60 by 60 m plots. In the RAINFOR protocol an area of 100 by 100 m is recommended as a plot, but smaller plot sizes have proven to be more practical and easier to sample in the scope of this project. Square sized plots are preferred, since they have a low edge/area ratio and are easy to make, which helps reducing the problem of trees located at the edge of the plot. (note: a circle has a lower edge/area ratio but is more difficult to construct in practice)
Table 3: The different strata on the west side of the Andes where fieldwork was conducted. The number of plots and altitude are also represented in meters above sea level (masl)
Location Altitude asl Number of plots
Strata 1 Rio Silanche 400 m 5
Strata 2 Milpe 1100 m 4
Strata 3 Maquipucuna 1900 m 4
Strata 4 El Cedral 2300 m 3
Strata 5 Puranqui (Intag) 3200 m 4
The location of these permanent plots was delicately chosen. Several constrains limited the selection of plots:
• Natural disturbances such as rivers were not tolerated within the plot • Human disturbances (e.g. pathways) were not allowed
• The slope had to allow sampling work (<40%) • Canopy closure
• Accessibility of the plots • Homogeneous soil
a
b 3.1.1 Climate
Mean annual temperature shows a linear decline along the elevational transect with a decrease of 4.3 °C for 1000m increase in altitude (P<0.001, R2 = 0.97, Figure 5a, Table 4). The mean annual precipitation is large over the entire transect > 1000mm and follows a linear decline over the elevational transect with a decrease of 948 mm for 1000m increase in height (P<0.001, R2 = 0.87, Figure 5b, Table 4). Climate data retrieved from NOAA (NOAA National Centers for Environmental information , 2020).
Table 4: Climatic data for the different altitudes, mean annual temperature (MAT) in °C and the mean annual precipitation (MAP) in mm.
Altitude (masl) 400 1100 1900 2300 3200
MAT (°C) 23.8 20 17.4 16.1 11.2
MAP (mm) 3447 3066 1492 1374 1181
Figure 5: Variation of climate along the 3200 masl transect, including mean annual temperature (a) (MAT) in °C and mean annual precipitation (b) (MAP) in mm. A linear regression with 95% confidence interval, P - value and R2 are
3.1.2 Forest description
Figure 6 contains pictures from the different locations.
3.1.2.1 Rio Silanche
The plots in Rio Silanche were located in the hilly evergreen forest on an altitude of 400 masl. Rio Silanche is a bird sanctuary and is owned by the Mindo Cloudforest foundation. The region is known for its bird diversity, a lot of endemic bird species are located here. The plots in Rio Silanche are flat compared to other locations. Around the area of Rio Silanche palm trees are grown on an agricultural scale. The area is under deforestation to provide more grazing area for cattle (Mindo Cloudforest Foundation, 1).
3.1.2.2 Milpe
The Milpe site is located on an altitude of 1100 masl and owned by the Mindo Cloudforest foundation. It is known as one of the most unique sites in Ecuador to see bird species. Milpe exists out of natural forest, but the highest plots showed parts of the forest that were more recently grown. Plots were located on steep hills, which made the data sampling difficult. (Mindo Cloudforest Foundation, 2).
3.1.2.3 Maquipucuna
The Maquipucuna Cloud Reserve is a 6000-hectare primary forest reserve on an altitude of 1900 masl located in the Pichincha province of Ecuador. It is a lower montane cloud forest where 10% of all the plant species of Ecuador are located (Maquipucuna Cloud Forest reserve, 1). The field campaign of this stratum was performed by botanist German Toasa.
3.1.2.4 El Cedral
The plots in El Cedral were located on an altitude of 2200 masl near a scientific station owned by botanist German Toasa. The reserve spans an area of 71-hectare. It is the only place where 60 by 60 m plots were used, which were established by the owner German.
3.1.2.5 Puranqui (Intag)
Puranqui is situated in the Imbabura province in Ecuador. The plots were located on an altitude of 3200 masl, trees at this altitude reach a height of 10-15 m and grow slow. The most abundant tree species in all the plots was Freziera canescens. The area is under deforestation to provide graze land for cattle.
Figure 6: Pictures from the different locations during the field campaign in Milpe (a), El Cedral (b), Puranqui (c) and Silanche (d).
a b
3.1.3 Field protocol
This field work was further build on research by Bauters (2013), Demol (2016) and Bruneel (2016). The field protocol was based on the RAINFOR protocol (Philips et al. 2009), which is a manual for fieldwork in the tropics. The protocol helps understanding the study for people who were not involved in the field work, it is also a tool for comparing results from different projects.
3.1.4 Permanent Monitoring Plot establishment 3.1.4.1 Orientation
The plots were established with a north-south and east-west orientation for the main axes. This was overall the most consistent way to work, nevertheless, sometimes deviations were necessary due to irregularities or overlap of plots. The centre point and the corners of the plots were registered using a GPS. The borders of each plot were made using a rope or measuring tape while doing the fieldwork.
3.1.4.2 Dimensions
For the plots, a dimension of 40 by 40 m was used with an exception for the plots established by German in El Cedral, which had a dimension of 60 by 60 m. This is smaller than recommended, but in previous field campaigns, it has proven that larger plots were not efficient due to disturbances and the topography which made it difficult to work in (Bauters, 2013).
3.1.4.3 Topography
At each site, all the plots were located in a cluster at more or less the same altitude. For every plot the slope was determined for every border using the Nikon Forestry Pro.
3.1.4.4 Fixation of the boundaries
The borders of each plot were made visible using plastic tubes, which were painted red using spray paint. Once the first tube was placed, a measuring line was used to determine the centre of the plot after 20 m. Once the centre was determined, the other border could be measured using a 20m measuring line. This method was repeated to determine the other sides. The subplots were named according to their wind direction.
3.1.5 Marking, measuring and counting of trees
All the trees were marked using numeration tags. In the past tags were nailed in the tree, but since some trees suffer from termites afterwards and eventually die, tags were attached to the tree using a simple rope. The POM of the tree was marked using red/orange spray paint. Spray paint was added on the same side of every tree within a subplot, the orientation of the spray paint was changed 90° for every subplot, which makes it easier in the future to detect different subplots. A drawing of every plot with details like coordinates, number of plots, slope was made to help future measurements. For trees standing at the borders, a simple rule was used: If more than 50% of the root system of the tree was included in the plot then the tree was included in that plot. Nevertheless, some cases were given more attention:
• If a tree was leafless, it was checked if it is was still alive, if dead this was noted down. • Sometimes trees were multiple-stemmed. These trees were only tagged on stems with a
diameter bigger than 10 cm at a height of 130 cm. Trees standing close to each other were checked if they were not belonging together.
• Fallen trees were checked for death.
The diameter at breast height (DBH) was measured at the POM (130 cm). During the fieldwork, a plastic tube with POM painted in red was used to mark the POM by putting it next to the tree. In case of deformities around the tree base at 130 cm, the POM was taken higher. The tree was cleaned of epiphytes or debris at the POM to have an accurate measurement of the DBH. Lianas covering trees were lifted in order to measure the diameter of the tree under the lianas. In some cases when the tree was buttressed at 130 cm, the POM was taken 50 cm higher at 180 cm. The POM was always measured on the downhill side, if the tree was bended the POM was measured along the trunk (Figure 7).
Figure 7: Measuring the POM on a slope and a bended tree. Figure by Philips et al. 2009
The tagged trees were identified on species and genus level by botanist German Toasa. The identified species were then compared to species names of 2015 and if necessary corrected. The identification was done by looking at leaves of trees, even by cutting in the trunk and smelling. Fallen fruits also helped in identifying species. A lot of species in the plots were endemic, which
made it sometimes difficult to identify. Tree by tree was classified. Once the total inventory of each plot was made, the species that were not present in the DRYAD database and were not measured in previous field campaigns were sampled for stem density. When the species for wood sampling were determined. Wood samples were taken from these species outside the plot. In this way, tree species inside the plot were not disturbed. The wood sampling was done by using a Pressler borer, the samples were put into paper straws and dried in the sun or oven if available.
Tree height was measured in 2015 using the Nikon Forestry Pro during a field campaign by Miro Demol and Stijn Bruneel (2016) these were vital to make height diameter relationships which were used to determine the height of trees based on their diameter measurements in 2019. From experience, it can be said that height measurements in these tropical forests are a challenge because of the densely overgrown vegetation.
3.1.6 Leaf sampling
Species determining 80% of the basal area were calculated. From these species, leaf samples were taken (Table 5). At least two individuals were sampled but preferably more. Leaves were only taken from trees with a diameter bigger than 10 cm. Preferably leaves exposed to the sun (sun leaves) were sampled. Leaves from the lower canopy were sampled when sun leaves were impossible to reach. A telescopic pruner which could reach up to 12 m was used to collect leaves (Figure 8b). The fresh weight was determined when leaf samples were collected.
The specific leaf area (SLA) was measured using an application on the smartphone called ‘Easy Leaf
Area free’ (Ealson, 2014) available in the google play store. The leaf area was determined on the
day of sampling, since the app determines the leaf area based on the green surface of the leaf, which fades away after some hours. The app was used in the following way:
• A white background was used for the leaves e.g. paper • A red square, 2 by 2 cm, was drawn on the white background
• The app on the phone uses the camera to detect the square as a reference area to determine the leaf area
The SLA was calculated by dividing the leaf area over the fresh weight. After calculating the SLA, leaves were dried in the sun and if possible an oven to prevent fungi from growing on the leaves. After the leaves were dry, they were put into envelopes for transport to Ghent for further analysis.
Table 5: No. of leaf and wood samples taken for the different strata. No samples were taken in stratum 3 at 1900 masl, since the fieldwork was performed by botanist German Toasa.
Altitude (masl) 400 1100 1900 2300 3200
No. of leaf samples 32 96 0 115 14
No. of wood samples 11 19 0 23 6
Figure 8: Two pictures during the field campaign. Sampling of soil (a) and on the collection of leaves using a telescopic pruner (b).
3.2
Lab analysis
3.2.1 Leaf samples
All leaf sample material was grounded using a bullet grinder at 200 tpm (Figure 9a), approximately 3 microgram was weighted into tin cups using a precision scale (Figure 9b). These sample cups were then analysed for C, N and its isotopes δ13C and δ15N by using an element analyser at ISOFYS, Department of Green Chemistry and Technology, Ghent University.
3.2.2 Wood samples
The dry weight of every wood sample was measured after samples were taken out of the oven at 103 °C for 48 hours, since atmospheric air could be water saturated. A method was chosen that approaches the green volume by submerging the air dried samples. The samples were submerged for 24 hours, afterwards the diameter (D) and length (L) were measured to calculate the volume π
4D2L. The wood density was then measured out of the dry weight and the approached green volume.
3.3
Data Processing
All statistical analyses were done in the software R (version 3.6.3), an open source software program (RStudio Team, 2019). Data from two time periods 2015 and 2019 were used in the analysis. El Cedral (cluster 4) was measured in 2015 and 2018 and calculated to the same period as the other locations, by calculating the yearly diameter increment for this period (three years). The yearly diameter increment was then used to determine a diameter increment for four years. This allowed us to compare results for all the strata.
Figure 9: Two figures with material used for analysing the leaf tissue, the grinder (a) and the precision scale (b) used to weigh leaf samples for further analysis.
b a
Surface
Planimetric surfaces were calculated and converted to hectare to describe variables per unit surface. For every plot the slope is determined per border using the Nikon Forestry Pro. The ground segments were corrected to planimetric lengths using the following formula:
lplanimetric =lsegment
cos (θ) (10) with lsegment the length of a segment of a subplot, which is 20 in the case of 40 by 40 m plot, lplanimetric is the planimetric length and θ is the slope. Based on the planimetric length of all the subplots the planimetric surface of each plot was calculated.
Tree Height
In 2015 height and diameter measurements at breast height were taken for all trees in the different plots. These are used to make diameter-height models for each plot. The function
modelHD from the R package Biomass (Réjou‐Méchain et al. 2017) was implemented with diameter and height measurements for each plot. Different methods (log10, log2,..) were analysed and the best suited one was used for the height diameter relationship. This model was then used to estimate tree height in 2019 based on the diameter measurements of 2019. To work consistent all tree heights from 2015 were also estimated. In this way, all calculations for aboveground biomass for trees in 2015 and 2019 used the estimated height.
A different method was necessary for tree height estimates for the plots located in El Cedral (cluster 4). No tree height measurements were available here. Diameter measurements of El Cedral were available. For cluster 4 a diameter-height model was constructed, using the same method as above, but based on cluster 3 (Macuipucuna), under El Cedral.
Wood density
Wood samples were taken from trees which had no wood density data in the DRYAD database. These samples were analysed for wood density in the lab. The DRYAD database was used to assign wood density at trees on genus and species level. The wood density measurements from the lab were assigned at plot level to the different trees based on genus and species name. If wood density measurements contained two or more values for the same genus and species an average value was assigned. Trees without wood density determination were given a plot average.
Aboveground Biomass
Based on the diameter measurements, the estimated height and the wood density the AGB for each tree was calculated. The function computeAGB from the Biomass package (Réjou‐Méchain et
al. 2017)was used for the calculations. The function calculates the AGB for each tree based on the estimated height (H), wood density (WD) and measured diameter (DBH) by pan-tropical allometric relationship:
AGB = (0.0673×(WD×H×DBH1000 2)0.976 (11) This equation was made by Chave et al. (2014) with height in m, diameter in cm and wood density in g cm-3, the model performs well for all different forest types and climatic conditions in the tropics (Chave at al. 2014).
The total AGB was measured for every plot and converted to total AGB ha-1. It is important to distinguish different sorts of total AGB ha-1:
• The total AGB ha-1 for all trees in 2015 and 2019.
• The total AGB ha-1 for all trees that died in the time interval (AGB dead trees). Based on this number the mortality was calculated as the ratio of AGB dead trees in 2019 and the total AGB of 2015.
• The total AGB ha-1 for the new recruits in the time interval (AGB new recruits). Based on this number the vitality was calculated as the ratio of AGB new recruits 2019 and the total AGB of 2015.
The productivity was calculated as is described in section 2.3.2 Quantification of forest biomass (Figure 2, approach 2).
Isotopes
To calculate community levels of δ13C and δ15N, the basal area of the community was calculated and used to determine basal area weights for the isotopes.
Statistical analysis
Differences in variables between strata were tested using the Mann-Whitney two-sample test from the R package mtcp. Tables are made with averages for each stratum and their standard error. Different letters are used to indicate significant differences between strata (P=0.05). The different traits WD, SLA and LNC were used in mixed effect models with diameter increment at breast height as response variable. The clusters and different plots were implemented as a random nested effect, since plots were spatially clustered. The altitude, traits and diameters were implemented as fixed effects without interaction. Altitude was rescaled to km. The response variable DBH increment and the fixed effect diameter were log transformed for normality. The models were performed using the maximum likelihood method from the R package lme4 (Bates et al. 2007). A Linear model was used to estimate the yearly wood productivity (ton ha-1 year-1) for the different clusters. The P-values are reported in the results.
a
b
4 Results
4.1
Stand structure
Tree height decreased linearly over altitude for the forest both inventories 2015 and 2019 (Figure 10a, 10b). Tree height increment followed a linear decline over altitude (P<0.01, R2 = 0.29, Figure 10c). The total basal area increased over altitude for both time periods (Figure 11a, 11b). There was not a significant trend for the basal area productivity (Figure 11c). The stand variables for the different strata were calculated (Table 6).
Figure 10: The two figures 10a and 10b show tree height (m) in function of altitude (masl) for the different time periods 2015 (a) and 2019 (b), figure 10c shows the increment in tree height (m) over the time interval. Linear regressions with 95% confidence interval were added and R2 and P-values are given.
a
b
Figure 11: The two figures 11a and 11b show the total basal area in m2 ha-1 in function of altitude in masl for the
different time periods 2015 (a) and 2019 (b). Figure 11c shows the basal area productivity in m2 ha-1 year-1. Linear
regressions with 95% confidence interval were added and R2 and P-values are given when significant.
