438 and references

To the Editor: Amphibian species are declining at an alarming rate on a global scale (1). One of the major reasons for these declines is chytridiomycosis, caused by the chytridiomycete fungus, Batrachochytrium dendrobatidis (1,2). This pathogen of amphibians has recently emerged globally (2,3) and has caused mass die-offs and extensive species declines on 4 continents (1,3); knowledge of its distribution and effects on amphibian populations remains poor. In Europe, little is known about B. dendrobatidis distribution, which is disturbing when one considers that at least 3 European amphibian species are undergoing chytrid-associated die-offs that will likely lead to local extinction (4,5) (J. Bosch et al., unpub. data). 
 
We screened 1,664 current and archived samples of wild amphibians collected in Europe from 1994 to 2004 by researchers using amphibians as study organisms. B. dendrobatidis infects the skin of adult amphibians and the mouthparts of anuran larvae; samples included toe clippings and skin samples from adults and mouthparts of tadpoles. Our sampling was opportunistic, including both caudates and anurans. We screened all samples for chytrid fungus with quantitative real-time polymerase chain reaction (PCR) of the ITS-1/5.8S ribosomal DNA region of B. dendrobatidis (6), including appropriate positive and negative controls. We confirmed real-time PCR positives by amplifying a subset of these positives with a second B. dendrobatidis–specific PCR with a nested reaction developed from the ctsyn1 locus (3). To confirm that detection with real-time PCR indicated a viable chytrid infection, when actual tissue samples were available, we examined a generous subset using histologic features for typical signals of pathogenic B. dendrobatidis infection. Specifically, we found intracellular zoospore-carrying sporangia within the stratum corneum and stratum granulosum of toe and skin samples. We also compared real-time PCR amplification profiles of suspected positives to those generated from samples from animals involved in chytrid-driven die-offs and found these results to be comparable. Furthermore, attempts to isolate the fungus from dead animals were successful when animals were obtained in a suitable condition for this purpose (see below). 
 
Our survey found B. dendrobatidis in amphibians in 5 European countries, Spain, Portugal, Italy, Switzerland, and Great Britain. Previously, chytrid infection has been reported in wild amphibians only in Spain, Germany, and Italy (4,5,7,8). We detected chytrid fungus in 20 of 28 amphibian species examined, representing 9 different genera, 5 anuran, and 4 caudate, in 6 families. We found signs of chytrid in archived samples from as early as 1998. The number of infections per country we found were Austria 0/24, Croatia 0/8, Czech Republic 0/18, Italy 2/101, France 0/60, Germany 0/51, Greece 0/88, Portugal 1/25, Slovenia 0/29, Spain 108/345, Sweden 0/197, Switzerland 63/252, and United Kingdom 2/466. Infection prevalence was exceptionally high in Spain and Switzerland. In Spain, ongoing chytridiomycosis-driven declines of midwife toads (Alytes obstetricans) and salamanders (Salamandra salamandra) have been documented since 1997 (4) and 1999 (5), respectively, and confirmed with scanning electron microscopy, histologic examination, and molecular detection methods (4,5). Common toads have been suffering apparently minor chytrid-related die-offs in Spain for several years, but mass die-offs were observed in 2004 (5) (J. Bosch et al., unpub. data). No chytrid-related die-offs have been reported in Switzerland. Furthermore, the infected animals from Switzerland were all adults in good breeding condition, many of which reproduced successfully in behavioral and ecologic experiments. Real-time PCR amplification profiles for the Swiss samples were quantitatively equivalent to those generated from samples of A. obstetricans collected during mass die-off events in Spain; from these latter samples, we successfully isolated viable B. dendrobatidis cultures from 2 geographically distinct areas. In Great Britain, we found chytrid in 2 of 14 introduced North American bullfrogs (Rana catesbeiana) caught in 2004 but did not find it in wild-captured native species. Examination by microscope and electron microscope of 180 native British amphibians from 1992 to 1996 did not find chytrid infection (A.A. Cunningham, unpub.data). The ability of the North American bullfrog to act as a vector for chytrid range expansion has been hypothesized (9,10). Our data may indicate that bullfrogs can fulfill this role in Great Britain and other areas; we have found the molecular signal of chytrid infection from introduced North American bullfrogs collected on 3 separate continents (T.W.J. Garner et al., unpub. data). 
 
This survey shows that B. dendrobatidis is widely and irregularly distributed in Europe and infects a broad range of amphibian species. Furthermore, because of the opportunistic nature of our sampling strategy, our results certainly underestimate the overall prevalence of B. dendrobatidis in Europe. These findings are surprising considering that chytrid-related die-offs have been infrequently described in Europe. This may be because B. dendrobatidis has only recently and rapidly expanded its range into Europe (3), and the consequences are only now being detected in wild amphibian populations; because the expression of chytridiomycosis is environmentally limited (11); or because European amphibians exhibit highly variable levels of resistance to chytrid infection. Notwithstanding, our knowledge of the epidemiology of B. dendrobatidis is insufficient to effectively manage wildlife and conduct disease abatement. As data regarding the distribution of chytrid fungus accumulate and the ecologic requirements for disease persistence and transmission are identified (11), management of the pathogen can become more predictive. Basic management practices, such as restricting transportation of potential carriers and restricting pet trading and reintroduction projects, coupled with field monitoring, must be improved to prevent further global emergence of this pathogen. Our results also show that asymptomatic amphibians must be included in any broad-scale epidemiologic screening for this emergent pathogen.


Introduction
Cropland expansion, the land conversion from natural vegetation (e.g., forests and grasslands) to cropland, is historically the most dramatic land-use change across the globe (Tilman, 1999;Foley et al., 2005;Chen et al., 2006).Between 1700 and 1992, about 1,135 M ha (22.9 %) of forest and woodland were converted to agricultural use Figures

Back Close
Full The terrestrial ecosystems in South and Southeast Asia play a crucial role in regional and global carbon cycling (Brown et al., 1991(Brown et al., , 1995;;Flint and Richard, 1991;Richard and Flint, 1994;Houghton, 2002;Tian et al., 2003).As regions with large populations and growing economies, South and Southeast Asia have witnessed unprecedented rapid land-use change in the 20th century.This is characterized by cropland expansion and natural forest shrinkage, with cropland expansion undertaken to meet growing demands for food, bioenergy and urban development (Richard and Flint, 1994;Houghton, 2002;Tian et al., 2003Tian et al., , 2008Tian et al., , 2010a)).Cropland covers about 73 % of the total land area in South Asia today, while it covers about 50 % in Southeast Asia (Wood et al., 2000).For South and Southeast Asia as a whole, cropland has increased by approximately 106 M ha and approximately 81 % of forests and wetlands have been converted during the period 1880-1980 (Flint and Richards, 1991).In India, cultivated area increased by more than 40 % while 40 % of forest area was lost during 1880-1980 (Flint and Richards, 1991).It has been estimated that forest cover in Indonesia decreased by up to 20 million ha in the 1990s as a result of logging, transmigration, spontaneous settlement, and estate crops (Pagiola, 2000).Compared to other regions of the world, major cropland expansion has occurred in South and Southeast Asia over the past 20 years (Millennium Ecosystem Assessment, 2005).Around the globe, the tropical area in Southeast Asia had the highest deforestation rate, followed by Latin America and Africa (Achard et al., 2002).In most of the regions where a high rate of deforestation occurred, forest land was converted to cropland.Tropical forest has the highest biomass and relatively high soil carbon density, so land conversion from tropical forest to other vegetation types will significantly reduce carbon storage in the terrestrial ecosystem (Achard et al., 2002(Achard et al., , 2004;;Canadell, 2002).
In recent decades, a number of studies have investigated the potential impacts of land-use change on the carbon cycle in South and Southeast Asia using the bookkeeping model, remote sensing data, inventory data, and process-based models at various scales, ranging from site, local, and regional to continental scales (Brown et al., 1991;Houghton and Hackler, 1999;Houghton, 2002;Canadell, 2002;Tian et al., Introduction Conclusions References Tables Figures

Back Close
Full 2003; Suh et al., 2004;Zhao et al., 2006;Adachi et al., 2011).The bookkeeping and inventory methods (e.g.Houghton and Hackler, 1999) are able to calculate changes in carbon storage induced by land area change; however, these approaches lack the capability of providing a quantitative understanding of ecosystem processes that control carbon dynamics; due to the shortage of spatially-explicit, time-series data of biomass and land use/land cover, they rarely provide information on interannual and spatial variations in carbon sources and sinks.In an earlier study, Tian et al. (2003)  generalized crop to represent all types of crops and did not take into account cropspecific information on phenology and cropping systems.To reduce the uncertainty in estimating spatial and temporal patterns of carbon fluxes and pools in association with cropland expansion in South and Southeast Asia, therefore, we have to better understand cropland expansion-driven carbon dynamics by using updated data sets and an improved process-based ecosystem model.In this study, we applied a process-based Dynamic Land Ecosystem Model (DLEM), which fully couples biogeochemical processes with an agricultural submodel (Tian et al., 2011a;Ren et al., 2011a) 2 Materials and methods

Study area
The region of South and Southeast Asia extends from 38 • north to 16 • south latitude and from 60 • to 140 • east longitude, encompassing a land area of approximately 5.0 × 10 6 km 2 and 4.7 × 10 6 km 2 , respectively (Fig. 1).South Asia, which includes the countries of Afghanistan, Bangladesh, Bhutan, India, Nepal and Pakistan in this study, is one of the most populated regions in the world and has a long history of cultivation.Southeast Asia covers the countries of Sri Lanka, Brunei, Burma, Indonesia, Cambodia, Laos, Malaysia, the Philippines, Thailand, and Vietnam.Southeast Asia contains the world's third largest tropical rainforest but this has been heavily deforested over the past decades (UNEP, 2009).Both South Asia and Southeast Asia are influenced by monsoons that bring strong seasonal changes in precipitation, and are subject to climate extremes, particularly floods and droughts.

Model description
The Dynamic Land Ecosystem Model (DLEM) is a highly integrated process-based ecosystem model which couples biophysical characteristics, plant physiological processes, biogeochemical cycles, and vegetation dynamics and land use to make daily, spatially-explicit estimates of carbon, nitrogen and water fluxes and pool sizes in the terrestrial ecosystems from site and regional to global scales.The DLEM has been documented in previous studies through extensive applications to investigate the impacts of multiple environmental factors, including changes in climate, atmospheric composition (CO 2 , O 3 , reactive nitrogen), land use and land cover change, and land management (harvest, rotation, fertilization, irrigation etc.), on the structure and functioning of terrestrial ecosystems over China, Monsoon Asia, the conterminous US, and North America (e.g., Tian et al., 2005Tian et al., , 2008Tian et al., , 2010aTian et al., , b, 2011a, b, c;, b, c;Chen et al., 2006;Ren et al., 2007Ren et al., , 2011a, b;, b;Zhang et al., 2007Zhang et al., , 2010;;Liu et al., 2008;Lu et al., 2011;Xu et al., 2010).Introduction

Conclusions References
Tables Figures

Back Close
Full The DLEM model has improved process-based simulation to track the effects of land-use change on ecosystem processes that control the terrestrial carbon cycle.An agricultural module is specifically developed to simulate impacts of agricultural activities (such as seeding, planting, irrigation, fertilization, tillage, genetic improvement, and harvest) and environmental factors on carbon, water and nitrogen cycles in agricultural ecosystems (Ren et al., 2011a).Different crop types and rotations are specifically parameterized.Currently, about 20 major crop types (e.g., corn, rice, wheat, barley, soybean, sorghum, cotton, maize, sugarcane) and three cropping systems (i.e., single, double, and triple harvests) are included in DLEM simulations.The main crop categories in each grid were identified according to the global crop geographic distribution map with a spatial resolution of 5 min (Leff et al., 2004), and were then modified with regional agricultural census data derived from FAOSTAT (http://faostat.fao.org/).DLEM simulates crop growth according to prescribed phenology derived from remote sensing methods (i.e., leaf area index, LAI) and large numbers of field observations.Phenological metrics include seeding, germination, development, flowering, fruiting and harvest.
Four general land-use change categories are simulated by DLEM, namely land conversion from natural vegetation types to cropland, land conversion among different natural vegetation types, cropland abandonment, and urbanization (Fig. 2).During land conversions, partial vegetation carbon will be removed as product pools, partial vegetation and soil carbon will be released to the atmosphere through land conversion flux, and the rest will enter the litter carbon pool.Three kinds of product pools are defined in DLEM: 1-(PROD1), 10-(PROD10), and 100-(PROD100) year product pools, which represent 1-, 10-and 100-year turnover time, respectively.Partial vegetation biomass will be burnt (site preparation) immediately after land-use change and then directly enters the atmosphere as land conversion fluxes.The rest of the vegetation biomass will enter different aboveground or belowground litter carbon pools.DLEM separates three litter carbon pools: very labile, labile, and resistant litter carbon pools.Accompanying carbon redistribution after land-use change, the ecosystem nitrogen and hydrological cycles will be correspondingly changed.The changed Introduction

Conclusions References
Tables Figures

Back Close
Full biogeochemical and hydrological cycles will in turn feedback to ecosystem restoration and development after land-use change.This coupling approach differs from bookkeeping or statistical estimation methods that are used to predict land-use change effects because the latter do not address these mentioned feedbacks.If land conversion occurs from natural vegetation to cropland, crop will establish in the ecosystem after site preparation.The cropland ecosystem will then develop according to the prescribed phenology of individual crop types.If land conversion occurs from cropland to natural vegetation (i.e., cropland abandonment) or conversions between different natural vegetation types, natural vegetation will be established in the ecosystem as a secondary succession according to the phenology and vegetation dynamics module in DLEM.The ecosystems that experience land-use change might not be able to restore the initial carbon, water and nitrogen conditions even after a long time period, for example, 100 years.A process-based modeling could fully track the successional stages of natural vegetation and could make a more accurate estimation of carbon dynamics after land-use change than statistical methods can.
In the DLEM, the annual net carbon exchange (NCE) between terrestrial ecosystems and the atmosphere represents the overall carbon budget in the terrestrial ecosystem, which is calculated as: where R a , R h are carbon loss from autotrophic respiration and soil heterotrophic respiration; E c is the total carbon loss due to land use conversion; E p is the total carbon loss from products decay.

Land use and land cover data
In this study, the historical , gridded (0.5 • × 0.5 • ) land-cover data set driving the DLEM was reconstructed by incorporating historical cropland and urban distribution (HYDE v3.0, Klein Goldewijk, 2001) and a potential vegetation map.The potential vegetation map was generated from multiple data sources.We first combined the MODIS global land cover map (http://modis-and.gsfc.nasa.gov/landcover.htm)with the global potential vegetation map developed by Ramankutty and Foley (1998) to derive the first draft of the potential vegetation map.Both data sets (with 1 km × 1 km resolution) were aggregated using the majority rule and their vegetation classes were regrouped to match the respective plant functional types defined in DLEM.Then, the global C4 grassland percentage map developed by Still et al. (2003) was used to determine the distribution of C4 grassland in the study region.Finally, we identified the wetland area based on the half degree resolution Global Lakes and Wetlands Database (GLWD) developed by Lehner and D öll (2004).
Over the past century, cropland area increased by 191.7 M ha in South and Southeast Asia, while the areas of forest, shrubland and grassland decreased by 110.3, 49.1 and 18.7 million ha, respectively.In the early 20th century, forests covered about 90 % of Southeast Asia and cropland covered approximately 5 % of the total land area.From 1901 to 2000, the cropland area increased rapidly while forest area decreased.
By 2000, forest area had decreased by 15.6 % and cropland increased nearly 3 times when compared with the rate of increase in the early 20th century.Most of these changes in land use occurred in Thailand, Indonesia, and the Philippines (Fig. 3).South Asia has a long history of cultivation and 27.1 % of the total land area has been converted to cropland in the early 20th century (Ramankutty et al., 2006).From 1901 to 2000, cropland continued to increase rapidly until the 1960s, when the increasing trend slowed and then leveled off after the 1980s (Fig. 4b).Natural forests, grassland and shrubland were converted to cropland by 54 %, 5.8 %, and 30 %, respectively, Introduction

Conclusions References
Tables Figures

Back Close
Full during the study period.Most of these conversions occurred in southwestern India.
The decrease in shrubland occurred primarily in Pakistan and in northwestern India (Fig. 3).In South and Southeast Asia as a whole, cropland area followed a continuous increasing trend during the 20th century, with a relatively slower rate ensuing after the 1960s (Fig. 4c).The major land conversions were forest-to-crop and shrub-to-crop, which accounted for 71.0 % and 17.4 % of cropland expansions, respectively.Forest coverage on average declined from 56.3 % in 1901 to 45.4 % in 2000.

Other datasets
Cropping system: The cropping system database in South and Southeast Asia includes thirteen crop categories (e.g., wheat, corn, soybean, cotton, groundnuts, millet, barley, sorghum and rice), three rotation types (one harvesting, double harvesting and triple harvesting) and their corresponding crop phenology information, which was developed based on MODIS data, agricultural census data and a global crop distribution map.The main crop categories in each grid were identified according to the global crop geographic distribution map with a spatial resolution of 5 min based on the work of Leff et al. (2004).The rotation type in each grid was developed using phenological characteristics and census data at the national level and state level.The phenology information obtained from MODIS LAI (at a spatial resolution of 1 km) was used to identify the rotation type and was calibrated using census data and site data before application.

Soil properties:
We obtained the spatial maps for soil bulk density, soil pH, and texture, (i.e., percentage of clay, sand, and silt content), of the study region from the International Satellite Land Surface Climatology Project (ISLSCP) Initiative II Data Collection distributed by the Oak Ridge National Laboratory Distributed Active Archive Center (http://daac.ornl.gov/).This Collection provided spatially-explicit global soil information derived from data and methods developed by the Global Soil Data Task, which was coordinated by the Data and Information System (DIS) of the International Geosphere-Biosphere Programme (IGBP).Introduction

Conclusions References
Tables Figures

Back Close
Full Topography map: Topography maps required by DLEM include elevation, slope and aspect.We first aggregated the Global 30 Arc Second Elevation Data (GTOPO30) developed by the United States Geological Survey (Bliss and Olsen 1996) to half degree resolution to develop the digital elevation model of the study region.Then we derived the half degree resolution aspect and slope maps from the digital elevation model.

Simulation experiments and implementation
To address the effects of land-cover change on carbon dynamics in South and Southeast Asia during the 20th century, we ran DLEM using the historical gridded land-cover data set from 1900 to 2000 and kept other input data unchanged.We generated the long-term average climate spanning 1961 to 1990 to represent the initial state of the year 1900, and the daily pattern in 2000 was selected to manifest the daily climate variation.The climate data in 1900 were used for the non-climate-transient simulation experiments.There were no ozone effects in this simulation experiment because the entire daily ozone AOT40 index was zero before 1940 (Felzer et al., 2005).In addition, CO 2 and nitrogen deposition in 1900 were used in this experiment.
The model simulation began with an equilibrium run to develop the baseline carbon, nitrogen, and water pools for each grid.Finally, the model ran in transient mode driven by transient data of land-conversion while other datasets remained constant.

Changes in net carbon exchange (NCE)
The carbon storage in terrestrial ecosystems in South and Southeast Asia decreased due to land-use change since 1901 (Table 1).The simulation result indicated that a total of 18.26 Pg C (0.18 Pg C yr −1 ) was released to the atmosphere from changes in land use induced by crop expansion during the 20th century in South and Southeast Asia (Table 1).The mean annual emission of 0.18 Pg C in South and Southeast Asia Introduction

Conclusions References
Tables Figures

Back Close
Full in the 20th century is slightly lower than the previous estimation (0.23 Pg C yr −1 ) for 1860-1990(Tian et al., 2003)).The carbon emissions from South Asia (4.19 Pg C) and Southeast Asia (14.07 Pg C) accounted for 23 % and 77 %, respectively, of the total emissions from both regions during the 20th century.Although a much larger land area in South Asia (Fig. 4) was converted to cropland, carbon losses in this region were significantly lower than those in Southeast Asia.Nearly half of the cropland areas in South Asia were converted from shrubland, grassland and other natural vegetations, while in Southeast Asia almost all the cropland areas were converted from forest.In addition, most areas of Southeast Asia are located in the tropical climate zone where densities of forest biomass and soil carbon are significantly higher than those in South Asia.
When partitioning the NCE into different components, we found that most of the carbon emissions were from land conversion fluxes (Ec, 10.83 Pg C), which accounted for 59 % of total carbon emission (Table 1).This suggested that carbon flux from land conversion was the major contribution to carbon emission in both South and Southeast Asia.Net ecosystem production (NEP = NPP-Rh) and product decay (Ep) accounted for 23 %, and 18 % of the total emission, respectively.
The accumulated net carbon exchange rate per decade varied significantly over time across South and Southeast Asia (Fig. 5).Decadal accumulated carbon emission showed an increasing trend and peaked in the 1980s (2.56 Pg C) and the 1990s (2.38 Pg C) for Southeast Asia, while carbon emission peaked in the 1950s (0.90 Pg C) and has no obvious increasing tendency during the entire study period in South Asia.Different land-use change patterns for Southeast and South Asia can explain the difference (Fig. 3, Fig. 4).Land area converted to cropland nearly leveled off around the 1950s in South Asia (Fig. 4b).Carbon emissions after the 1950s were primarily from land use legacy effects.In contrast, land area converted to cropland kept a relatively constant rate of increase from 1901 to 2000 in Southeast Asia (Fig. 4a).Partially driven by land-use change legacy effects, decadal carbon emissions continuously increased from the 1900s to the 1990s in this region.Introduction

Conclusions References
Tables Figures

Back Close
Full

Changes in carbon pools
Most carbon emissions induced by land-use change were from carbon losses in the vegetation carbon pool, which accounted for about 83 % (15.09Pg C) of the total carbon emission (18.26 Pg C), while soil organic and litter carbon pools only accounted for 11.0 % (2.01 Pg C) and 8.8 % (1.60 Pg C), respectively (Table 2).In South Asia, vegetation, soil, and litter carbon pools decreased 3.2, 0.48 and 0.32 Pg C, respectively.The vegetation carbon pool accounted for about 76.4 % of the total decreases in carbon storage in this region.In Southeast Asia, vegetation, soil and litter carbon pools decreased 11.89, 1.53 and 1.28 Pg C, respectively.The vegetation carbon pool accounted for 84.5 % of the total decrease.In general, vegetation and soil carbon in Southeast Asia decreased much more than they did in South Asia.Vegetation, soil and litter carbon pools varied over time in the entire region (Fig. 6).The significant decrease in vegetation, soil and litter carbon pools occurred during the 1980s in Southeast Asia, while the significant decreases in these pools occurred in South Asia during the 1950s.The decrease in carbon pools gradually leveled off after the1980s in Southeast Asia and after 1950 in South Asia.For both South and Southeast Asia, two peaks of decreases in vegetation and soil carbon pools were found in the 1950s and the 1980s, which account for 35 % of the total decreases in vegetation and soil carbon during the period 1901-2000.

Spatial variation in changes of carbon storage
Land-use change has resulted in large spatial variation in the net carbon exchange rate in South and Southeast Asia during the 20th century (Fig. 7).Most areas in South and Southeast Asia were carbon sources during 1901-2000 because cropland expansion was the major land conversion type.Few areas were found to be carbon sinks in South Asia due to cropland abandonment, cropland expansion from shrubland, and reforestation/afforestation (Fig. 7a; Fig. 3).The largest carbon sources (>100 g C m −2 ) were found in Southeast Asia where cropland expansion from tropical forest may have Introduction

Conclusions References
Tables Figures

Back Close
Full During the first half of the 20th century , most areas undergoing landuse change were carbon sources because cropland expansion was the predominant land-use change type (Fig. 7b and Fig. 3).During the second half of the 20th century, more areas acted as carbon sources and the magnitude of change was much wider than that in the first half of the century (Fig. 7b, c).The biggest carbon sources were found in Indonesia and the Philippine Islands where tropical forests were the major land cover types.Deforestation and cropland expansion occurred in Southeast Asia during the second half of the century.In South Asia, however, more area acted as a carbon sink during the second half of the century due to cropland abandonment and afforestation/reforestation, and over 80 % of cropland abandonment occurred during this period.

Carbon fluxes resulting from different categories of land-use change
Land-use change effects on carbon fluxes varied among different categories of landuse change.The results indicated that the conversions from forest to cropland and from shrubland to cropland accounted for 79.7 % and 8.3 % of total carbon emission, respectively.Land conversion from grassland to cropland only contributed 4.2 %.Our results indicated that the 1.56 × 10 6 km 2 land area (about 10 % of total land area experiencing land-use change) was converted to cropland and the cropland was abandoned afterwards.This category of land conversion resulted in 1.5 Pg C emission during the study period, which implied that original carbon storage could not be completely restored after cropland abandonment during the study period.Introduction

Conclusions References
Tables Figures

Back Close
Full

Process-based modeling of land-use change impacts on the global carbon cycle
In this study, we link the land use data with other spatially-explicit environmental information (e.g., climate, soil prosperities, cropping system, and topography etc.) as well as ecological processes in a process modeling framework.We improved the process-based modeling methods and accordingly developed new spatial data sets to address the dynamic effects of land-use change on terrestrial ecosystem carbon cycling in South and Southeast Asia during the 20th century.To track the impacts of land-use change, DLEM coupled the ideas from statistical methods such as the book-keeping method (Houghton, 1995(Houghton, , 1999(Houghton, , 2003) ) and the process-based modeling methods of TEM (Terrestrial Ecosystem Model, McGuire et al., 2001;Tian et al., 2003).The fully coupled carbon, water and nitrogen cycles addressed in DLEM enable the model to simulate the impacts of land-use changes on carbon fluxes and other biogeochemical cycles (Ren et al., 2011a).Furthermore, like the LPJ model (Bondeau, 2007), DLEM has the ability to simulate carbon dynamics related to the land conversion between natural ecosystem and human-managed ecosystem.For a historical study at the regional scale, DLEM tried to introduce the available observational database into a mechanism analysis.For example, the least area index (LAI) derived from remote sensing was used in DLEM as input data to represent daily vegetation phenology.This enhanced the ability to capture land surface change in the real world.In addition, other new spatially-explicit environmental databases, which are crucial to describe the impacts of land-use change, were developed and applied in this study.Specifically, land management (e.g., fertilizer, irrigation, tillage) and cropping systems (crop categories and rotation types) based on substantial data sources (e.g., FAOSTAT; Leff et al., 2004) had be used.Introduction

Conclusions References
Tables Figures

Back Close
Full

Total carbon emissions resulting from land-use change
For the purpose of simulation evaluation, we compared our simulation results with previous studies using a bookkeeping model, remote sensing data, and a terrestrial carbon model (Houghton and Hackler, 1999;Houghton, 2002;DeFries et al., 2002;Tian et al., 2003) (Table 3).Our simulated results suggested that annual carbon emission induced by land-use change was about 0.28 Pg C yr −1 in the two decades before 2000, which is comparable to those estimations of 0.32 Pg C yr −1 from the book-keeping model (Houghton and Hackler, 1999), and 0.25 Pg C yr −1 from integrating remote sensing and the book-keeping model (DeFries et al., 2002).However, our results were modestly higher than the estimation of 0.13-0.17Pg C yr −1 in 1981, drawn from the statistical method and inventory data (Hall and Uhlig, 1991).The century-scale analysis showed that our estimation of 0.18 Pg C yr −1 during 1901-2000 was similar to the estimation of 0.17 Pg C yr −1 during 1850-1980 (Houghton and Hackler 1999) and slightly lower than the estimation of 0.23 Pg C yr −1 during 1860-1990 from Tian et al. (2003).The discrepancies among those studies were mainly attributed to differences in study period, data source, and methods.Even though the estimate for carbon emission in Southeast and South Asia (0.28 Pg C yr −1 ) was similar, e.g., at 0.28 Pg C yr −1 to the estimate of 0.25 Pg C yr −1 for the tropical Asia during 1981-2000 in DeFries et al. (2002), the underlying mechanisms controlling this total amount of carbon emission were found to be different.For example, in the DLEM model, we assumed that the recovery of vegetation growth would be limited by nutrient and water resources when cropland was converted to natural vegetation.In the book-keeping model (Houghton, 2002;Houghton and Hackler, 2003), abandoned croplands returned to the ecosystems from which they were initially cleared, accumulating carbon in biomass and soil at specific rates.In our study, net carbon emission to the atmosphere included the emission from other trace greenhouse gases, such as methane, which possibly resulted in higher total carbon loss (5. at least partially contribute to the discrepancies between our results and those of other studies.

Uncertainties and future needs
Although we improved the estimation accuracy, there are still many uncertainties in this study.Firstly, uncertainty may arise from the input data.Different data sources usually have big gaps for the cropland area in South and Southeast Asia.For example, according to FAO (1982), cropland accounted for about 13.2 % of the total land area in the 1980s.Wood et al. (2002) showed that almost half of Southeast Asia and 73 % of South Asia are cropland areas.Data from Ramankutty and Foley (1998)  Secondly, some processes related to land-use change were simplified in the DLEM model.For example, the parameters for allocating litters to different litter carbon pools after land-use change were kept constant.In addition, land conversions from one natural biome to another (e.g., forest to grassland or shrubland) were not included in this study due to limited data sources.
To improve our ability for simulating the interactions between the regional carbon cycle and land-use change, several key land use processes (e.g., management for plantation, urbanization, biofuel production, rotational uses of forest) should be included in the future study to better reflect the effects of land-use change on the spatial and temporal patterns of the carbon balance in South and Southeast Asia.For specific regions like South Asia that have experienced a long history of cultivation and management practices, we need to consider the legacy effects of land use before 1900 and remove associated artifacts.Also, the major source of uncertainty in estimating global carbon release is from aboveground biomass density during the land conversions from forest to other uses (Houghton et al., 2009).The importance of biomass in the carbon cycling should be given more attention in process-based modeling at regional and global scales.Moreover, further evaluation of DLEM model results is needed using accessible 11994 Introduction

Conclusions References
Tables Figures

Back Close
Full  Biogeosciences, 7, 3637-3655, doi:10.5194/bg-7-3637-2010, 2010. Zhang, C., Tian, H., Chappelka, A., Ren, W., Chen, H., Pan, S., Liu, M., Styers, D., Chen, G., and Wang, Y.: Impacts of climatic and atmospheric changes on carbon dynamics in the Great Smoky Mountains National Park, Environmental Pollution, 149, 336-347, 2007. Zhang, C., Tian, H., Wang, Y., Tao, Z., and Liu, Y.:  Full  Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | used a process-based model (the Terrestrial Ecosystem Model, TEM) to estimate changes in carbon fluxes and pools due to land-use change in Asia.Although their study simulated the monthly carbon fluxes induced by land-use change, the TEM model used a , to quantify the effects of cropland expansion on carbon fluxes and pools in South and Southeast Asia in the 20th century.The objectives of this study were to: (1) quantify the effects of cropland expansion induced land-use change on temporal and spatial patterns of carbon fluxes and pools in the terrestrial ecosystem of South and Southeast Asia; (2) examine the relative role of different landuse change types on the capacity of terrestrial carbon sequestration; and (3) identify major uncertainties and future research needs.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | A positive NCE value means a terrestrial carbon sink, whereas a negative NCE indicates that terrestrial ecosystems are carbon sources to the atmosphere.The land-use change related parameterization for DLEM follows the empirical parameters established by Houghton (2003), other process-based models (e.g.Terrestrial Ecosystem Model, TEM, Tian et al., 2003) and field observational data.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | resulted in a large amount of carbon loss.Tropical forest has very high carbon density, as high as 22 kg C m −2 in vegetation and 11 kg C m −2 in soil.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 53 Pg C) as estimated by DLEM, comparing to the estimation of 5.0 Pg C for the tropical Asia from DeFries et al. (2002) in the period of 1981-2000.These processes Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | data sources such as long-term field experiments and observations on plant regrowth after land-use change.5ConclusionsIn this study, we investigated the temporal and spatial patterns of changes in carbon fluxes and pools resulting from land-use change and specifically cropland expansion in South and Southeast Asia over the 20th century.Cropland expansion increased continuously over the 20th century, showing a relatively higher rate of increase before the 1960s.The major land conversion types were forest-to-crop and shrub-to-crop, which accounted for 71.0 % and 17.4 % of cropland increases, respectively.Results estimated with DLEM indicated that temporal and spatial patterns of change in carbon storage were significantly influenced by cropland expansion.In the 20th century, South and Southeast Asia released 18.26 Pg C to the atmosphere, of which South and Southeast Asia accounted for 23 % (4.19 Pg C) and 77 % (14.07 Pg C), respectively.This suggested that the entire region acted as a carbon source from cropland expansion.More carbon emissions were from Southeast Asia though larger land areas in South Asia were converted to cropland.Decreases in vegetation carbon (15.09Pg C) contributed the most (83 %) to the total carbon emission.The largest decrease in vegetation carbon occurred in Southeast Asia due to crop expansion that converted natural tropical forests, while an increase in soil carbon was found in some areas of South Asia as a result of afforestation/reforestation or land conversion from cropland to shrubland over the last two decades.Most areas in South and Southeast Asia were carbon sources due to vast cropland expansion, with the highest carbon loss in Southeast Asia occurring where tropical forests were cleared for cropland.During the 20th century, the carbon dynamics resulting from land-use change in these two regions were dominated by Southeast Asia where cropland expansion occurred as natural tropical forests were converted.Discussion Paper | Discussion Paper | Discussion Paper | ton D.C., 2000.Xu, X. F.,Tian, H. Q., Zhang, C., Liu, M. L., Ren, W., Chen, G. S., Lu, C. Q., and Bruhwiler, L.:   Attribution of spatial and temporal variations in terrestrial methane flux over North America, Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |
indicated cropland coverage in South Asia and Southeast Asia at 39.2 % and 19.2 %, respectively, in the 1960s, with the highest values of 42.3 % and 26.5 % in the 1990s.
Predicting response of fuel load to future changes in climate and atmospheric composition in the Southern United States, Forest Ecol.

Table 1 .
Different components of net carbon exchange (NCE, Pg C) in South and Southeast Asia during 1901-2000.NEP -Ec -Ep; negative values in NCE mean carbon release into the atmosphere; NEP: Net Ecosystem Productivity; Ec: the total carbon loss due to land use conversion; Ep: total carbon loss due to land use conversion Note: NCE =

Table 3 .
Estimations of changes in carbon storage (Pg C) resulted from land-use change in South and Southeast Asia using process-based model, book-keeping model, and statistical method.