Return to ecosystems summary page

1. Tropical rainforest (evergreen or semi-evergreen forest of the humid tropics, usually tall)

(corresponds to Olson et al. seasonal tropical forest and broad-leaved humid forest.).

References directly cited in these pages (does not at present include secondary citations)


The various types of forest existing in the moist tropics are widely regarded as an important source or sink of carbon in the present day world, and also during the late Quaternary. However, there is still considerable uncertainty over the carbon storage values of tropical forests under 'natural' (i.e. mid-Holocene) conditions. These problems are discussed below, separately for each of the main ecosystem components.

Suggested summary values for the preanthropogenic state

Storage (tC/ha) Ecosystem component
10 tC/ha Dead standing trees, coarse woody litter, leaf litter and other debris
100 tC/ha Soil organic carbon
210 tC/ha Above and below-ground vegetation
320 tC/ha Total carbon storage



Storage (tC/ha) Location Author(s)
10-15 tC/ha (1.) Global average (ad hoc estimate) M. Harmon, pers. comm.
2.5-3 tC/ha (2.) Tropical lowland moist forest Brown & Lugo (1984)
20 tC/ha (3.) Amazon, Rondonia (primary forest) Salati et al. (1990)

(1.) Coarse woody debris only.

(2.) Litter generally. Based on a global literature survey, of 13 samples. Standard error +/- 0.3.

(3.) Litter and fallen trunks.

Conclusion; The dataset on this reservoir is very poor, but a figure of around 10 tC/ha is conservatively suggested as representative of the total coarse and fine litter and debris reservoir.



There is a great deal of uncertainty over the 'natural' preagricultural state of tropical forest carbon storage. The subject is presently in a state of overhaul with the much more thoroughly researched findings of Brown & Lugo (1984), Brown et al. (1992), and Iverson et al. (in press) that are replacing the views based on the purely present-day study of Olson et al. (1983). These more recent authors have argued that many forests in all three tropical regions, that were often sampled under the belief that they were not anthropogenically disturbed, in fact have a long history of selective logging or shifting cultivation. They suggest that, even allowing for the effect of natural site disturbances such as landslides, the 'average' carbon storage of non-anthropogenic forests would be much higher, due to the presence of very large trees which are either removed by selective felling ('log poaching') or do not have time to grow full size under a shifting cultivation regime. Shifting cultivation has existed for thousands of years in many areas (especially the more seasonal rainforest zones), but its intensity is likely to have grown since its inception in the early-to-mid Holocene, and especially within the past century or so as the local populations have increased.

Such considerations are of relevance to reconstructing carbon storage in the forests that existed in the earlier Holocene and also in historical times, under lower intensities of human-induced disturbance. It may also be relevant to note that several hundred years after abandonment, forests that have recolonised old fields of the Mayan (Central America) and Khmer (South-east Asia) civilisations retain detectable structural and compositional differences from the surrounding areas of relatively undisturbed forest (Burrows 1990, Huston 1994). This perhaps emphasises the long-term sensitivity of forest structure to anthropogenic disturbance, even though a superficially identical closed high forest can form in only a few decades following clearance (Burrows 1990, Huston 1994).

The intensity of natural disturbance events in tropical forests varies greatly with local site conditions. Hurricanes only affect areas of forest that are relatively close to the coasts, such as in Central America. In floodplain forests of the Manu River in the western Amazon, disturbance due to river channel shifting may occur once every 200 years on average. However, only a small proportion of forest is actually floodplain forest (Huston 1994 p.550). Earthquake and volcanic damage to forests is significant on the timescale of centuries in some areas of the tropics that are seismically active, (e.g. Puerto Rico; Crow 1980) but this does not apply to most areas.

Drought and fire may be a more significant disturbance factor in limiting tree size and biomass in many parts of the tropics, but it generally seems to be a very rare event in the rainforest zones. Mueller-Dombois (1978), in his extensive review of the incidence of fire in the tropics, concludes that naturals fire almost only ever occurs on peat swamp soils in the tropical rain forests, burning through the peat layer during drought periods (this seems to have been the case for example in the severe drought in parts of Borneo in 1982-3). Otherwise, he concludes, the only fires that occur on latosoil forests seem to be in areas extensively disturbed by humans, where humans actually set the fires (again, this seems to have been true in Borneo in 1982-83). Even these fires are rarely spread through the forest on anything other than a very localised scale.

However, Huston (1994 p.531) cites evidence for a 300-600 year return cycle of fires, even in the high rainfall parts of the Amazon Basin. To what extent these fires actually destroy the canopy trees, rather than just burning through the litter or understory layer, is not clear. Even the periodic destruction of forest on these time scales seems unlikely to act as a very major suppresser on forest biomass, since most tropical trees seem likely to reach their maximum size in less time than this (although the life span of tropical trees is generally quite poorly known) (Walter 1971).

Mueller-Dombois (1978) suggests that semi-deciduous forests might have a higher fire frequency because of the regular seasonal incidence of drought. However, he notes, it may be difficult for any fires to get started on the ground because of the lack of a good humus or litter layer (due to high rates of decomposer activity year-round). Hence, he is also sceptical of the importance of fire in the semi-deciduous forests.

The proportion of roots typical of tropical forests is another major uncertainty. Because of the difficulty of measuring root mass, many biomass estimates deal only with above-ground biomass, and leave it up to the reader to guess what the below-ground mass was. Other authors calculate the below-ground mass on the assumption that the proportion of root biomass is as low as 0.1; it is often not clear from published figures what ratio they have used and on what basis this figure was accepted. J. Grace (University of Edinburgh, pers. comm. 1991) suggests that a value of at least 15% of total plant biomass below ground seems appropriate. This would seem a conservative figure; according to J. Grace, Douglas Deans (ITE Aberdeen, UK) has recently obtained figures for above and below ground forest biomass in Cameroon which show the root/shoot biomass ratio between 0.2 and 0.4, with a mean of around 0.25 (J. Grace, Edinburgh University, pers. comm. Jan. 1994; submitted manuscript). If these figures are representative of tropical forest vegetation in general, the approximate 'additional 20%' figure I use here will tend to underestimate the actual forest biomass (this would assume a root:shoot ratio of around 0.19). In this sense, my estimate should be seen as a conservative one.

The following are for the most part estimates for 'present-day' carbon storage above and below ground in the less anthropogenically disturbed tropical forests. These estimates (e.g. in Brown's papers) do not generally include understory/shrub biomass within the forest, which would account for an extra 3-4 % (Brown & Lugo 1984).

Storage (tC/ha) Location Author(s)
200tC/ha (1.) Nonseasonal equatorial forest Olson et al. (1983)
290tC/ha (2.) Tropical rainforest Rodin et al. (1975)
164tC/ha (3.) Tropical rainforest Duvignead (s.)
201 tC/ha (4.) Africa and Americas (7 sites) Tanner (1985)
335 tC/ha (5.) Malay Penin. Everwet rainforest Iverson et al. (1994)
212 tC/ha (6.) Thailand. Seasonal rainforest Iverson et al. (1994)
170 tC/ha (7.) Rio Negro basin. Semi-evergreen forest Medina & Ceuvas
185 tC/ha (8.) Tropical lowland moist forest Brown & Lugo (1984)
140 tC/ha (9.) Moist tropical forest Olson et al. (1983)
171 tC/ha (10.) Primary forest. Central Amazonia McWilliam et al. (s.)
209 tC/ha (11.) Mean preanthropogenic, Africa Brown & Gaston (1996)

(1.) Based on a global literature review. 180 tC/ha as low figure, 200 as medium figure, 250 as high figure for 'wet nonseasonal evergreen equatorial forest' For present-day (c.1980) forests. Incorporating sources 1. and 2. below. This probably assumes too low a proportion of root biomass, and incorporates many forests that are extensively disturbed by anthropogenic factors. However, also many of the IBP-derived values it refers to have also been criticised for concentrating on atypically high biomass sites.

(2.) Based on a review of IBP studies (cited by Olson et al., above). This estimate has been criticised as unrepresentative of overall present-day biomass for the site studies on which it was based were selectively placed in high-biomass forest patches.

(4.) E. Tanner, University of Cambridge pers. comm. (data table published by him in 1985, collected from various other literature sources) gives living above-ground standing crop in lowland tropical forests. These are here converted into carbon storage equivalents (x 0.475) = 161 tC/ha above-ground biomass (this needs a conversion factor for root mass to be added). If one conservatively assumes the root mass to be an extra 20%, this would make 201 tC/ha. Following this conversion with Tanner's collected data, the following figures are available; 190 tC/ha for a site in Brazil, 100 Ghana, 148 Panama, 241 Ivory Coast, 203 Ivory Coast, 153 Colombia, 86 Colombia.

(5.) The findings of Iverson et al. (1994) are based on an extensive compilation of forest inventory data from several tropical SE Asia countries. Noting that literature data from ecological studies on rainforests in remote uninhabited regions, and taking the relationships between biomass density and the proportion of the biomass in large trees, they infer that in fact most of the rainforest in SE Asia in relatively recently disturbed by anthropogenic factors such as logging. Applying predictive equations to information on tree girths and density, they suggest 154 t C/ha for above-ground biomass of SE Asian forests (both monsoon forest and rainforest) when undisturbed by humans (as an average over whole region; including dry forests), or if one assumes a root mass of an extra 20%, then this would give 192 t C/ha in total. This is probably a good, reliable estimate (Tanner pers. comm., 1992), but it may include too much dry forest, tending to give a lower figure than for true rainforest alone. In the paper by Iverson et al. (1994), there is a summary diagram showing abundance distribution of potential forest biomass in Peninsular Malaysia (moist aseasonal tropics) taken at 550 t/ha aboveground biomass (x 0.475= 244 t C/ha, or 305 tC/ha if one includes adds on an extra 20% of root mass. Brown et al. (1991) suggest that an above ground biomass figure of greater than 225 t/ha of carbon - and perhaps as high as 350 t/ha carbon - may be representative of truly 'primary' (unlogged and not subject to shifting cultivation) forests in the Indo-Malayan region. If an additional 20% belowground biomass is assumed, this gives a total of 270 tC/ha (for the lower figure) to 420 t/ha (for the higher figure). Similarly, for sites in the Brazilian Amazon, Brown and Lugo (1992) note that the anomalously low proportion of large trees present may be due to a history of human disturbance of various kinds during the past few centuries (many of these sites are in areas subject to logging and with a relatively high density of human habitation), with true 'primary' forest biomass being higher than the figures normally obtained for forests in that region. It is of course difficult to allow here for the great heterogeneity in tropical forests (swamp forests, limestone forest etc.), and the role of natural disturbance events (such as drought and fire) in keeping down the biomass of pre-anthropogenic forests. The available data on disturbance intensity simply do not allow this assessment to made precisely, although as I have already argued, it is unlikely to cause most current assessments of the non-anthropogenically influenced biomass to be too high. One can only hope that Iversen et al. have properly taken this into account. Iversen et al. also note that the abundance of dipterocarps in the SE Asian everwet forests may tend to mean that they have rather higher biomass values than everwet forests elsewhere, because dipterocarps do tend to form exceptionally big trees.

(6.) Similarly, for the seasonal rainforests of Thailand, Iverson et al. (1994) infer a pre-anthropogenic state (on the basis of various high biomass sites in isolated areas, and on ecological considerations) of around 350 t/ha aboveground biomass. Given an extra 20% for roots (a conservative figure), this gives 425 t/ha biomass, or (x 0.47) 212 tC/ha in tree biomass.

(7.) Tanner (University of Cambridge, pers. comm. May 1992) cites these figures from work by Medina & Cuevas (not directly cited here), which he feels are representative of the main forest in that area. Above and below-ground biomass of seasonal semi-evergreen forests in upper Rio Negro basin, Brazil. The following values (multiplied by 0.475 to convert from biomass to carbon) are given according to the soil types the forest was growing on; 146 tC/ha (Oxisol), 138 (Oxisol), 162 (Ultisol), 219 (Ultisol), 189 (Tropoquod), 168 (Tropoquod), of which the mean is 170 tC/ha. This value may be too low for the moister-climate types of forest, but semi-evergeen forests certainly cover large areas in all three main tropical regions. Note however Brown & Lugo's (6.) findings which indicate that the forests which have been sampled in the Brazilian Amazon are already disturbed by logging.

(8.) Based on a global literature review of 14 samples (standard error +/-12), of sites falling between 1700 and 3700 mm annual rainfall (mean around 2300mm). The figure was calculated by Brown & Lugo assuming a root:shoot ratio of 0.17. Note that the figure of 0.20 may be more accurate, and also that Iversen et al. (1994) have since suggested that many of these sites could be depleted in biomass due to previous anthropogenic disturbance.

(9.) For "evergreen or deciduous 'moist' forest, closed or regenerating well", 100 (low) - 140 (medium) - 170 (high) tC/ha are suggested. The deciduous category in fact would constitute what I have defined as 'monsoon forest' in this inventory, this inclusion tending to depress the carbon storage value for rainforests (being evergreen or semi-evergreen). Such more seasonal tropical forests are generally more suitable for agriculture, and may be relatively heavily affected by disturbance from humans. The inclusion of regenerating forest raises the suspicion that there is a substantial depression of biomass from human activity included in this estimate. Certainly, the 'medium' value given by Olson et al. is substantially lower than the indications from more recent sources on 'moist' or 'seasonal' rainforests. A value of about or above 170 tC/ha seems appropriate even for the most seasonal semi-evergreen rainforests.

(10.) J. Roberts (Inst. Hydrology, Oxfordshire, UK. pers. comm. Sept. 1995) reports a value of roughly 300 t/ha aboveground biomass for a sample of aboveground seasonal evergreen forest in the east-central Amazon basin, near Manaus. Multiplied by 0.475 to convert to carbon mass, and 1.2 to include a proportion of root material, one obtains a figure of 170 tC/ha. This is noticably lower than the estimates of Brown and colleagues. The result was published as part of a paper by McWilliam et al. in Functional Ecology, 1993 (Roberts, pers. comm.).

(11.) This figure for the 'pre-anthropogenic' state of African evergeen and semi-deciduous closed forests, is from site studies selected from forest areas that have low population densities and low anthropogenic disturbance rates. Note that Brown & Gaston suggest that the present-actual mean carbon density in the closed forests is about 209 tC/ha.

Conclusion; Considering the very persuasive arguments of Iverson et al. (5., 6. above), I suggest here a fairly conservative estimate of around 210 tC/ha in total biomass globally in tropical forests for the non-anthropogenic state. This lumps together values from everwet and seasonal rainforests, and does not include any contribution from understorey mass and smaller trees.



There is still a great deal of disagreement over what the average organic matter storage of tropical forest soils actually is, and the range of values reflects this uncertainty. The actual value certainly varies considerably with soil type and local site history. There seems to be a general intuitive feeling amongst ecologists that I have spoken to (e.g. D. Hall, University of London: J. Grace, University of Edinburgh) that the values given by Post et al. (1982) are too high to be representative of this ecosystem type, but Post (pers. comm.) makes the point that the values he and his colleagues have presented are based on a large number of samples, gathered using standardised methods.

Storage (tC/ha) Location Author(s)
97 (1.) Tropical oxisols Kimble et al. (1990)
83 (2.) Tropical ultisols Kimble et al. (1990)
200 (3.) Tropical rainforest soils Post et al. (1985)
98 Schlesinger (1984) Tropical rainforest soils
104 Zinke et al. (1984) Tropical rainforest soils
85 (4.) Lowland moist forest soils Brown & Lugo (1984)
115 (4.) Lowland wet and rainforest soils Brown & Lugo (1984)

(1.) Cited by Eswaran et al. (1993). Based on global literature review of 71 soil sections to 1m depth. 42% coefficient of variation in soil carbon values amongst the samples.

(2.) Cited by Eswaran et al. (1993). Based on 53 soil samples, 70% coefficient of variation.

(3.) the Post et al. study was based on 110 samples. Their use of the term rain forest may refer only to the Holdridge definition of very high rainfall forest (>5000mm).

(4.) Based on a global literature survey. Moist forest by their definition refers to sites with between 1700 and 3700 mm annual rainfall (mean around 2300mm) (15 samples, +/- 6 standard error). Wet and rain forest sites have a mean of around 5000mm. The values for the latter set show a fairly high variability (9 samples, +/- 32 one standard error). Since much more of the world's rainforest presently falls within the 'moist forest' category than the 'wet and rain forest' category, an overall value should be appropriately weighted according to their relative areas. As a preliminary estimate I suggest that an overall mean value of around 100 tC/ha might be reasonable on the basis of this.

Conclusion; Based on the available data summaries, a figure of around 100 tC/ha seems reasonable for tropical rainforest soils overall. This value does not include tropical peatlands, for which almost no data are available.