ThesisPDF Available

Developing a Tool for Carbon Accounting through Hydrological Modeling in the Sebangau Peatlands



Content may be subject to copyright.
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Emi ssion s (t/ ha)
3-1 -1 9 9 7
3-1 -1 9 9 8
3-1 -1 9 9 9
3-1 -2 0 0 0
3-1 -2 0 0 1
3-1 -2 0 0 2
3-1 -2 0 0 3
3-1 -2 0 0 4
3-1 -2 0 0 5
3-1 -2 0 0 6
3-1 -2 0 0 7
3-1 -2 0 0 8
3-1 -2 0 0 9
3-1 -2 0 1 0
Developing a Tool for Carbon Accounting through
Hydrological Modeling in the Sebangau Peatlands
Central Kalimantan, Indonesia
MSc Thesis by Ben DeVries
Developing a Tool for Carbon Accounting through
Hydrological Modeling in the Sebangau Peatlands
Master thesis Land Degradation and Development Group submitted in partial
fulfillment of the degree of Master of Science in International Land and Water
Management at Wageningen University, the Netherlands
Study program:
MSc International Land and Water Management (MIL)
Student registration number:
LDD 80324
Supervisor: H. Wösten
Examinator: L. Stroosnijder
Date: 27 September, 2010
Wageningen University, Land Degradation and Development Group
Alterra, Soil Science Centre
World Wildlife Fund (WWF Indonesia; WWF Germany)
This study is a follow-up to my major thesis, entitled “Monitoring the Effects of Hydrological
Restoration Efforts in Degraded Tropical Peatlands”. As in my previous thesis, the work in this
study was made possible with the help of several key people. First I would like to thank my
primary supervisor, Henk Wösten, for introducing me to the project and for helpful discussions
about theoretical and practical matters. I would also like to thank my second supervisor, Leo
Stroosnijder, for his helpful comments on my report drafts. The work done in this study would
not have been possible without the technical assistance from Christian Siderius, and
especially Ab Veldhuizen, of Alterra-WUR. Finally, I would like to recognize the efforts of the
WWF-Indonesia and Taman Nasional Sebangau staff for their continued work in peatland
restoration in the Sebangau National Park. Thank you also to those who made the training
workshop in Palangkaraya possible, and who participated actively in the animated
Hydrological restoration of drained tropical peatlands is a vital step in curbing CO
from these ecosystems. In this study, the SIMGRO-MODFLOW regional hydrological model
was coupled with a linear model relating average groundwater levels in tropical peatlands with
annual CO
emissions. Modifications to the model were introduced to tailor it to a hydrological
restoration project in Sebangau National Park, Central Kalimantan, Indonesia. Using this tool,
average CO
emissions due to peat subsidence over the study area between 1997 and 2009
were calculated to be 43t/ha/y. In a scenario where 240 hypothetical dams are constructed,
potential prevented emissions were calculated to be an average of 56,000tCO
/y aggregated
over the 150,000 ha study area. Initial testing of the model, as well as an interactive training
course in Palangkaraya, Central Kalimantan, indicate that several constraints exist to
implementing the tool in its current form towards linking project activities to carbon savings in
the Sebangau National Park. First, surface drainage simulation by SIMGRO yielded
unrealistic results due the absence of accurate watercourse data. Second, modifications
which simulate water flow through watercourses and the function of dams in the canal
drainage network did not produce results that could be corroborated by field observations.
Notably, the hydrological function of dams constructed in the Sebangau peatlands to date
was under-represented by the model according to training participants. The lack of surface
water data, including stage-discharge data for watercourses and ongoing monitoring of head
differences across dams, was identified as a key constraint to modeling dam performance.
Finally, the SIMGRO-MODFLOW model was found to consistently underestimate
groundwater levels throughout the study area. Comparison of predicted groundwater levels
between 2005 and 2010 to measured levels within the National Park yielded a normalized
root mean square error (nRMSE) of 28%, indicating that further calibration is needed to make
the tool applicable. Parameters governing hydrological processes in the unsaturated domain,
such as the storage capacity, are recommended as targets of further calibration. This report
describes the development of a carbon calculation tool based on the SIMGRO-MODFLOW
hydrological model, and offers concrete recommendations for further improvements, given the
constraints to the tool’s practical use.
Abbreviations Used
ASL – Above Sea Level
bda – Binary Direct Access (file format)
DEM – Digital Elevation Model
GIS – Geographic Information System
nRMSE – Normalized Root Mean Squared Error
NSE – Nash-Sutcliffe Efficiency
RMSE – Root Mean Squared Error
RSS-GmbH – Remote Sensing Solutions
SSI – Sanintra Sebangau Indrah (former owner of SSI canal)
SVAT – Soil-Vegetation-Atmosphere Transfer (SIMGRO unit cell)
swnr – Surface Water Number (SIMGRO-SurfW trajectory)
TNS – Taman Nasional Sebangau (Sebangau National Park)
UN-REDD – The United Nations Collaborative Programme on Reducing Emissions from
Deforestation and Forest Degradation in Developing Countries
VCS – Voluntary Carbon Standard
WUR – Wageningen University and Research Centre
WWF – World-Wide Fund for the Conservation of Nature
Table of Contents
Preface .......................................................................................................................................4
Summary ....................................................................................................................................5
Abbreviations Used.....................................................................................................................6
Table of Contents .......................................................................................................................7
Materials and Methods ...................................................................................................... 11
Study Area .............................................................................................................. 11
Climate Data ........................................................................................................... 12
Channel Flow Measurements and Modeling .......................................................... 13
Regional Hydrological Modeling ............................................................................. 14
Model Evaluation .................................................................................................... 15
Carbon Calculations ............................................................................................... 15
Model Training and Feedback from Users ............................................................. 16
Developing a Carbon Accounting Tool ............................................................................. 17
Project Goals and Objectives ................................................................................. 17
Pre-processing Tools ............................................................................................. 17
Revision of Drainage Procedures .......................................................................... 18
Modeling of Dams .................................................................................................. 19
Results and Discussion ..................................................................................................... 20
Channel Flow Modeling .......................................................................................... 20
Sensitivity to Surface Water Parameters ............................................................... 22
Surface Drainage Modeling .................................................................................... 23
Effect of Dams on Model Results ........................................................................... 26
Model Performance ................................................................................................ 30
Calculating Carbon Emissions and Savings .......................................................... 32
User Feedback and Conclusions ...................................................................................... 37
References ........................................................................................................................ 39
Appendix 1 – SIMGRO-MODFLOW Model Theory ................................................................. 41
Overview of the SIMGRO-MODFLOW Model ................................................... 41
Above-Surface Storage ...................................................................................... 42
Soil Moisture Storage ......................................................................................... 43
Groundwater Flow .............................................................................................. 44
Appendix 2 – Precipitation Data .............................................................................................. 45
Appendix 3 – Surface Water Data ........................................................................................... 48
Appendix 4 – Stage-Discharge Estimations ............................................................................ 51
Appendix 5 – Carbon Calculations .......................................................................................... 53
1. Introduction
Peatlands are recognized as a major sink for greenhouse gases. Due to chronically saturated
conditions, undisturbed peatlands accumulate organic material more quickly than it can be
decomposed, making them a sink for carbon over time. This sink function breaks down,
however, as peatlands are drained and degraded. Rapid land use change on peatlands,
involving drainage, deforestation, and burning, has actually triggered some peatlands to
become sources of carbon, rather than sinks. Interest in the function of peatlands as a carbon
sink or source has followed growing concern over the adverse effects of global climate
Despite its classification as a developing economy, Indonesia is currently the third largest
emitter of CO
after China and the United States, with the vast majority of CO
being emitted
from burnt or degraded peatlands (Schwarz 2010). Indonesia’s position among the top
emitters is largely due to the fact that it holds upwards of 50% of all tropical peatland area
(Page, Rieley et al. 2010), and the massive effect of ongoing land use change on these
delicate ecosystems. Current estimates of CO
emissions due to land use change on
peatlands in Indonesia run between 200 to 700 megatonnes per year (Hooijer, Page et al.
2010). In Central Kalimantan province on the island of Borneo, huge tracts of peatlands were
slated for large-scale agricultural developments some decades ago. This ‘Mega Rice Project
(MRP)’, now largely recognized to be a failure, has left a legacy of degradation on Central
Kalimantan’s peatlands.
Figure 1 – Peatland distribution throughout Indonesia and Malaysia (figure taken from Wösten, Clymans
et al. 2008). The study area in Central Kalimantan is denoted by a square.
A major factor in the degradation of Indonesian peatlands is the drainage, subsidence, and
eventual collapse of ombotrophic peat domes (Kool, Buurman et al. 2006). A substantial
component of the process of peat subsidence in Indonesian peatlands is the microbial
oxidation of peat soils where unsaturated peat substrate is converted to CO
. This process
has been shown to occur at a much higher rate in tropical peatlands than in temperate or
boreal peatlands (Wösten, Ismail et al. 1997). Tropical peatlands also differ from their
temperate and boreal counterparts in the fact that they do not experience such drastic
seasonal fluctuations, and the effect of temperature on emissions remains relatively constant
throughout the year (Chimner 2004). CO
emissions from tropical peatlands are tightly linked
to their hydrological integrity, a reality that must be taken into account by those interested in
peatland restoration.
With global attention turning towards the role of tropical peatlands as a sink (and potential
source) of CO
, commitments are being made at the national and international level to reduce
carbon emissions from Indonesian peatlands through wise use and restoration, through such
mechanisms as the The United Nations Collaborative Programme on Reducing Emissions
from Deforestation and Forest Degradation in Developing Countries (UN-REDD) and the
Voluntary Carbon Standard (VCS) (UNFCCC 2007; Couwenberg, Dommain et al. 2010;
Schwarz 2010). In this environment of increasing environmental awareness, various peatland
research and restoration projects in Indonesia have been initiated in recent years, adding to a
growing database of tropical peatland knowledge. One such project under which the current
study was conducted, is the Sebangau Restoration Project, undertaken jointly by Taman
Nasional Sebangau (TNS; the Sebangau National Park Authority) and the Indonesian World
Wide Fund for Nature (WWF-Indonesia) in association with WWF-Germany, Remote Sensing
Solutions in Germany (RSS-GmbH), and Alterra-WUR in the Netherlands. This project seeks
to restore the protected peatlands of the Sebangau National Park, whose history of illegal and
concession logging have left an extensive network of drainage canals. A central feature to the
project is the rewetting of peat soils in the Sebangau National Park through the construction
of dams across the small canals. Although some work has already contributed to knowledge
on the effectiveness of these measures (Jaenicke, Wösten et al. 2009; DeVries 2010), the
regional effects of canal blocking are poorly understood. In the interest of continued
hydrological monitoring and eventually linking these activities to a carbon trading scheme like
the Voluntary Carbon Standard (VCS), a set of tools for hydrological monitoring and carbon
accounting are needed.
This study was undertaken with the primary aim of developing the tool described above for
the Sebangau Restoration Project. Central to the tool’s function is the combined SIMGRO-
MODFLOW hydrological model. SIMGRO (SIMulation of GROundwater) was developed by
Alterra-WUR in the Netherlands as a hydrological model with a regional focus (Walsum,
Veldhuizen et al. 2010). Where other regional models model hydrological processes in
different domains separately, SIMGRO includes vital feedbacks between hydrological
domains, an important feature in understanding a hydrological system as an integrated whole.
In this project, SIMGRO was coupled to the MODFLOW groundwater model, developed by
the US Geological Service (USGS) to model later groundwater flow after Darcy’s Law
(Harbaugh, Banta et al. 2000). Together, the SIMGRO-MODFLOW model was used to
calculate average groundwater levels given daily precipitation data, topography of the study
area, and watercourse data. Average groundwater levels were used to predicted the effect of
canal blocking activities on predicted CO
emissions from the Sebangau peatlands.
This report describes the development of a carbon calculation tool for the peatlands of the
Sebangau National Park. Modifications to the SIMGRO-MODFLOW hydrological model
necessary for catering to the needs of the project partners are described, including an added
feature for modeling dams which allows users to adjust the strength of dams on canal
drainage. Comparison to groundwater data currently available from TNS and WWF show that
further calibration is needed for the model to perform effectively. A method for coupling
hydrological predictions to carbon emissions based on linear relationship derived from
literature values is also described in this report. An initial training course with WWF and TNS
staff was held in Palangkaraya, Central Kalimantan, Indonesia, during which reactions and
feedback to the carbon calculation tool were given. Finally, recommendations for further
refinement of the monitoring tool and carbon calculation tool based on experiences in the
study and training workshop are given in this report.
2. Materials and Methods
2.1 Study Area
This study took place within the Sebangau National Park (Taman Nasional Sebangau),
located within a large expanse of peatlands in Central Kalimantan, Indonesia. The National
Park encompasses approximately 570,000 ha of peat swamp forest, and is bound in the east
and west by Sungai (S., “river”) Sebangau and S. Katingan, respectively, and generally
follows the natural dome’ shape of the peatlands (Jaenicke, Rieley et al. 2008).
Palangkaraya, the provincial capital of Central Kalimantan is located directly to the northeast
of the National Park. The Sebangau peatlands are shown in Figure 2, where the actual study
are is outlined in black, and the three weather stations are indicated as blue triangles. A
detailed description of the Central Kalimantan peatlands, specifically those within the National
Park boundaries, can be found in DeVries (2010) (DeVries 2010).
Figure 2 – Area covered by the Sebangau National Park (Taman Nasional Sebangau). Locations of
climate stations from which data was used in this study are indicated by a star (Figure adapted from
Central Kalimantan Peatlands Project website:
The area on which this study focuses comprises of the eastern region of the Sebangau
National Park. The study area covers approximately 150,000 ha and is bordered on the
eastern side S. Sebangau, a black-water river originating in and flowing from north to south
through Central Kalimantan’s vast peatlands, finally draining into the Java Sea. The study
area is comprised mainly of three sub-watersheds, S. Bakung, S. Rasau, and S. Bangah
(from north to south, respectively), all of which drain into S. Sebangau. A fourth tributary of S.
Sebangau, S. Paduran Alam, is located at the southwest boundary of the study area, and the
point at which it drains into S. Sebangau represents the southern-most extreme of the study
area. The drainage area of this river is included in the study area, but did not factor heavily in
the analysis included in this study. A prominent feature of the study area is the large SSI
canal, situated between S. Rasau and S. Bangah, as well as a large network of other small
Figure 3 – Digital Elevation Model (DEM) of the study area. Watercourses are outlined in dark blue
(Jaenicke, Wösten et al. 2009).
The study area, shown in Figure 3, was chosen to encompass the eastern half of the
Sebangau peat dome. As such, the surface elevation rises from about 5m (above average
sea level) at banks of S. Sebangau to a maximum of approximately 20m at the northwest of
the study area. The area is comprised almost entirely of peat swamp forest, although there
has been considerable alterations to these forests due to illegal logging and forest fires during
the past 20 years. For the purpose of hydrological modeling, the entire area was assumed to
be intact peat swamp forest, since detailed analyses of the hydrological characteristics of the
degraded portions of these peatlands have not been carried out.
2.2 Climate Data
Climate data for the Sebangau area were compiled from a number of different sources
covering the period from 1 January, 1997 until present. Daily or hourly precipitation data were
collected from Palangkaraya Airport, Setia Alam research station in the north of Sebangau
National Park, and the TNS/WWF SSI field station according to Table 1. The locations of
each of the data collection locations are shown in Figure 2. Precipitation data aggregated per
year from these three sources are shown in Figure 4. Based on near uniform conditions
throughout the year, reference evapotranspiration rates were taken to be 3.5mm/d, based on
hourly evaporation data from the SSI field station.
Palangkaraya Airport 1/1/97 – 13/5/02
18/9/02 – 15/3/03
Setia Alam (CIMTROP) 14/5/02 – 17/9/02
16/3/03 – 14/10/03
19/3/04 – 8/12/08
SSI Field Station (TNS/WWF) 9/12/08 – present Hourly
Table 1 – Summary of precipitation data collected in or near Sebangau National Park.
Total P(mm)
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
P (mm / y)
Figure 4 Total annual precipitation in or near Sebangau National Park. Precipitation for 2010 is
aggregated from 1 January 2010 until 3 May, 2010. Annual trends in the precipitation data collected
from the sources indicated in Table 1 correspond well with published data from the same region
(Wösten, Clymans et al. 2008).
2.3 Channel Flow Measurements and Modeling
Surface water flows in open water channels were simulated using the SurfW functionality of
the SIMGRO model. SurfW conceptualizes watercourses as reservoirs of water connected in
tandem. At each time step, water is transferred from a given reservoir to its downstream
neighbour based on a explicitly defined stage-discharge relationship. Instances of backflow in
a watercourse were simplified by employing a ‘stop-flow’ strategy, in which downstream flow
is stopped when surface water head is higher in a downstream trajectory.
Discharge data was collected on two separate occasions from several small canals in the
Sebangau National Park using the float method as described in detail in DeVries (2010).
From these data, stage-discharge relationships were derived for other watercourses in the
Sebangau peatlands using the Chézy formula for surface water velocity (v; m/s):
[Equation 1]
where C is the Chézy coefficient (m
), R
is the hydraulic radius (m), and S is the average
canal bed slope (m/m). Assuming a rectangular channel cross section, the hydraulic radius
was determined by dividing the cross-sectional area by the wetted perimeter according to
Equation 2.
wh wh
[Equation 2]
where h
, and w are average water depth (m), and width of the wetted canal (m) respectively.
The Chézy coefficient for the peat canals was estimated by fitting collected stage-discharge
data to Equation 1. Stage-discharge tables were constructed for each canal ‘type’ (small
canal, large canal, natural tributary, and natural regional watercourse) using the calculated
Chézy coefficient and known or estimated watercourse dimensions according to the continuity
[Equation 3]
where Q is the discharge (m
) and dA is the cross-sectional area (m
) of a unit volume.
Assuming a constant velocity throughout the cross-section, Equation 3 is simplified to:
[Equation 4]
where A is the cross-sectional area (m
) of the entire wetted portion of the channel. The
combined Chézy-Continuity Equation method for estimating surface water discharge was
made into a preprocessing model in Visual Basic 6.5 using Microsoft Excel as a user
interface. The Chézy coefficient and average slope of some of the Sebangau watercourses
were adjusted using this interface, and the sensitivity of the SIMGRO-MODFLOW model to
these two parameters was investigated.
2.4 Regional Hydrological Modeling
The SIMGRO-MODFLOW model was used to gain insight into the regional hydrology of the
Sebangau National Park area using existing climatic, topographical, hydrological input data.
The theoretical basis of SIMGRO and MODFLOW model components are described in detail
in Appendix 1. Climate data were collected and compiled as described in Section 2.1.
Topographical data were based on a digital elevation model (DEM) derived from Shuttle
Radar Topography Mission (SRTM) images and verified using LIDAR data (Jaenicke, Rieley
et al. 2008). Hydrological data included a watercourse map prepared by Remote Sensing
Solutions (RSS GmbH) using qualitative and quantitative data collected by WWF-Indonesia
field staff (Jaenicke, Wösten et al. 2009). Land use was assumed to be “peat swamp forest”
over the entire study area. Other constant parameters were set based on Wösten et al. (2008)
(Wösten, Clymans et al. 2008).
The above data were preprocessed in a GIS environment using ArcGIS 9.2. Topographical
and hydrological data were extracted from the DEM and watercourse map respectively.
Drainage catchment areas of all watercourses were calculated under ‘natural’ conditions
(without canals) and under current conditions (with canals) using the DTM2CAT programme
The results of the SIMGRO-MODFLOW model were analyzed on both local and regional
scales. For local-scale analysis, groundwater or surface water hydrographs were produced for
one SIMGRO-MODFLOW unit cell (SVAT) or surface water number (swnr) in Microsoft Excel.
For regional-scale analysis, bimonthly groundwater data were extracted from the output
binary direct access (bda) file using a post-processing programme and ArcGIS 9.2 to produce
a raster image showing average groundwater levels above mean sea level (m+ASL) over the
entire study area. These rasters were converted to average groundwater levels relative to the
soil surface by subtracting DEM values in ArcGIS 9.2.
2.5 Model Evaluation
Results of the SIMGRO-MODFLOW model were evaluated by comparing output groundwater
data from the same location as data collected from the WWF/TNS SSI Field Station between
September 2005 and May 2010. Model and field groundwater hydrographs were compared
using the root mean square error (RMSE) and normalized RMSE (nRMSE) according to the
following equations:
( )
[Equation 5]
minmax xx RMSE
[Equation 6]
is the model-predicted value, x is the measured value, and n is the sample size of
the measured dataset.
2.6 Carbon Calculations
emissions due to oxidative decomposition of peat were estimated over the entire study
area by converting average peat drainage depth to CO
release according to the formula:
[Equation 7]
where E is the average CO
emissions (in tonnes CO
per annum) and D is the average
drainage depth (in metres relative to the soil surface) over an area dA (in metres). The
conversion constant, k, was taken to be 0.009t/m
/a based on Couwenberg et al. (2009)
(Couwenberg, Dommain et al. 2010). The drainage depth, D, was taken to be negative when
the water table was situated below the soil surface, and positive in flooding conditions. The
study area was stratified based on existing SIMGRO unit cells (SVAT’s). As such, for a study
area of n SVAT’s, Equation 7 is approximated as:
)0(, =
[Equation 8]
where A = 8100m
for all SVAT’s. In SVAT’s where the average annual groundwater level is
above the soil surface (D > 0), annual emissions were assumed to be zero.
To estimate carbon ‘savings’ due to canal blocking activities, the SIMGRO-MODFLOW model
was run separately for the cases without dams and with dams, respectively, and output maps
were generated using average groundwater levels for each year of the model run. Output
groundwater levels were converted to relative values by subtracting DEM-derived surface
elevation values, and the resultant drainage depths were converted to carbon emissions
using Equation 8. The difference in carbon emissions between the two situations were
determined in the last stage of the procedure. This method for calculating carbon savings
from project activities is summarized in Figure 5.
Figure 5 – Flow chart for calculations of carbon savings. Polygons represent raster files which were
manipulated in ArcGIS according to the scheme shown. Hgw refers to groundwater levels obtained from
SIMGRO-MODFLOW model results. Relative carbon savings (red) are calculated in the last step by
subtracting carbon emissions with dams from carbon emissions without dams.
2.7 Model Training and Feedback from Users
The SIMGRO-MODFLOW model was developed into a carbon calculation tool based on the
needs of the Sebangau Conservation Project. Included in this model package were several
preprocessing programmes designed specifically for the Sebangau project using Visual Basic
6.5 in a Microsoft Excel environment. A training was then held in Palangkaraya, Central
Kalimantan, Indonesia, involving field staff and project leaders from WWF-Indonesia and
Taman Nasional Sebangau (TNS). During the workshop, users’ encounters with and
reactions to the model were translated into concrete recommendations for further
modifications of the model and a revised data collection plan in the field.
3. Developing a Carbon Accounting Tool
3.1 Project Goals and Objectives
This study was undertaken in the context of a partnership between Alterra-WUR and the
Indonesian and German chapters of the World Wide Fund for Nature (WWF), as well as
Remote Sensing Solutions (RSS GmbH), Germany, initiated in 2009 towards establishing a
strategy for the monitoring of hydrological restoration efforts in the peatlands of the Sebangau
National Park (Jaenicke et al., 2009). This work of this partnership follows the ongoing field
work of the Central Kalimantan team of WWF-Indonesia, who have began canal blocking
activities in the Sebangau National Park as early as 2005. Monitoring of groundwater levels
surrounding one of the larger dams has been taking since the constructions of the first dams
in the SSI canal in 2005. As such, WWF-Indonesia has a modest dataset from which
preliminary indications of the local effects of canal blocking can be inferred (DeVries, 2010).
The next step in this project involves the spatial scaling-up of hydrological monitoring
methods and ultimately translating changes in water levels into prevented CO
Given the difficulties associated with large-scale monitoring of groundwater and surface water
dynamics throughout the Sebangau peatlands, an integrated hydrological model is necessary
to predict changes in groundwater and surface water levels in response to water management
interventions. The goal of this study was thus to develop a tool for calculating hydrological
implications of canal blocking activities, and to translate these output data into predicted
carbon savings. A pre-processing and post-processing interface was developed in Visual
Basic 6.5 to allow for easy access to the tool by WWF-Indonesia and TNS (Sebangau
National Park) staff. Conceptual modifications to the surface water, surface drainage, and
water management functions of SIMGRO-MODFLOW were included in the tool, and are
described in detail in the following sections. Though not specifically addressed in this chapter,
carbon calculations were coupled to the output of this tool as described in Section 2.6.
3.2 Pre-processing Tools
The SIMGRO-MODFLOW model is the central feature of the carbon accounting tool
developed in this study. An overview of the theoretical basis of the model is provided in
Appendix 1 of this report. The SIMGRO-MODFLOW model contains a myriad of options of
processes and parameters, which can be daunting for the inexperienced user. For this
reason, several pre-processing programmes were designed in Visual Basic (VBA 6.5) for use
by WWF/TNS field staff for easy processing of input files. First, a programme was designed to
translate hourly precipitation data from the WWF/TNS SSI meteorological station into daily
aggregated precipitation, and print the appropriate SIMGRO-MODFLOW meteorological input
file. After establishing the climatic record, a list of SIMGRO unit cells (SVAT’s), their
elevations, and watercourse locations (swnr’s) specific to the study area were exported from
ArcMap 9.2 as a comma-sparated-value file (csv). A programme was designed to translate
this list into input files relating to surface water numbers (swnr’s), including a swnr routing file
and a swnr management file (weir levels). A second programme was designed to take this
surface water data and incorporate it into SVAT-specfic input files, including an elevation
input file, an infiltration input file, a run-off input file, and a surface drainage input file. Finally,
a separate programme was designed to link stage-discharge information for each
watercourse with the previously generated swnr-routing file. This last programme was also
designed to allow users to optimize stage-discharge curves for each watercourse.
Specifically, the Chézy coefficient, channel width, maximum depth, and channel bed slope
can be adjusted by users to yield revised stage-discharge curves for each watercourse
category. Using this tool, the application of SIMGRO-MODFLOW to surface hydrology in the
Sebangau National Park can be improved with an increase in stage-discharge data, as well
as more information regarding stream and canal dimensions.
3.3 Revision of Drainage Procedures
Since floods in lowland tropical peatlands are a common occurrence, considerations for
surface water drainage as overland flow and canal flow must be made. Modeling flood
hydrology in this study involved the revision of two main processes: channel flow and surface
drainage to watercourses. Channel flow was modeled as described in section 2.3, but was
extended to follow a simple two-stage conceptualization, where flow through the rectangular
channel cross-section was accompanied by flow through a wider floodplain as shown in
Figure 6.
Figure 6 – Two-stage model for channel flow in flooding situations (when h > h
). Discharge is assumed
to occur in the regular rectangular channel as well as on a floodplain defined to have a width (2w) twice
that of the original channel.
According to the schematic in Figure 6, channel width (w) was doubled when channel stage
(h) exceeded the maximum channel depth (h
). To account for the increase in cross-sectional
area and hydraulic radius (R
) , equation 2 (section 2.3) was modified to yield:
wh whhwh
[Equation 9]
In this modified equation, only the cross-sectional area (numerator) is changed because
accounting for an increase in wetted perimeter (denominator) caused a temporary decrease
in discharge at stages exceeding maximum channel depth (data not shown), which is contrary
to common sense. Using this two-stage discharge model and the preprocessing tool
described in section 3.2, stage-discharge curves were generated for each watercourse
category, and are shown in Appendix 3.
The second aspect of drainage modeling that was revised in the tool was the drainage of
surface water into small canals. This revision followed the problems encountered unrealistic
flooding near small canals described in section 4.3. The hydrological prediction tool was
modified to alleviate excess water retention upland of small canals through an alternate
drainage method. Here, two different flow directions were defined for each small canal
segment (surface water number; ‘swnr’): one natural direction from an upstream swnr to its
downstream neighbour, and a second ‘bypass’ from one swnr to the swnr of the natural river
into which the canal drains. Under the assumption that surface water tends to follow natural
drainage basins during flooding periods, the model was set to follow this second route when
the swnr in question floods its banks. To accomplish this procedure, the ‘regular’ routing was
defined with a channel bed elevation of the soil surface minus the canal depth. A second
routing was then defined with a different stage-discharge relationship and a channel bed’
elevation equal to the soil surface. In this way, excess floodwaters around small canals was
quickly evacuated to natural watercourses, and artificial flooding (in the absence of dams)
was avoided.
3.4 Modeling of Dams
The construction of small dams in artificial canals is central to the restoration activities of
WWF and TNS in the Sebangau National Park. To model the retention of surface water based
on the location of the dams, an additional feature was added to the preprocessing
programmes described in Section 3.2, affecting surface water and drainage input data. This
feature closely follows SIMGRO’s current methodology for weirs, where weir levels can be set
per surface water number (swnr) (Walsum, Veldhuizen et al. 2010). In the absence of any
structures in the swnr, the weir level is set to equal the actual elevation of the channel bed of
the swnr. Raising the weir (or canal bed) elevation by a given value effectively halts canal flow
until the surface water head reaches the new weir level. There is still uncertainty, however,
about the extent to which each of the dams constructed in the Sebangau National Park
actually retain water. For this reason, a ‘dam factor’ was introduced into the preprocessing
code. The dam factor can take on a value between 0 and 1 and sets the height of the new
canal bed level, relative to the maximum canal depth. A dam factor of 0.75 thus raises the
canal bed by 75% of the canal depth, a dam factor of 1 raises the canal bed up to the bank
level, and a dam factor of 0 cancels out the effects of dams completely. These changes occur
only on swnr’s explicitly indicated by the user as “blocked canals” (swnr’s on which at least
one dam has been constructed). The dam factor was introduced into the pre-processing code
affected all input files related to weir (canal bed) elevation, stage-discharge input data, and
SVAT-specific drainage parameters.
4. Results and Discussion
4.1 Channel Flow Modeling
The SIMGRO model normally offers users the choice between several surface water sub-
models, including the SOBEK-CF and SWQN models based on the Saint Venant equations
(Walsum, Veldhuizen et al. 2010). The SurfW model, when not used in conjunction with other
physically based surface water models, requires the explicit input of stage-discharge
relationships for each watercourse trajectory rather than relying on physical equations. The
lack of surface water data from the Sebangau peatlands presents a considerable challenge to
hydrological modeling in any of the above options, where physically based models may be
sensitive to uncertain physical parameters, and the input data required by the SurfW must be
inferred in many cases. To accommodate the lack of physical data for most of the surface
water channels in the model area, surface water characteristics were separately modeled
using Visual Basic (VBA 6.5) according to the Chézy formula for flow in an open-water
channel (Equation 1, Section 2.3).
The empirically derived Chézy and Manning formulas account for the effect of canal
roughness elements on the flow and can explain unpredictable flow characteristics of natural
streams and rivers. To arrive at an appropriate quantitative flow equation for each of the
different watercourses within the study area, discharge data from several small canals were
collected in the Sebangau National Park on two separate days in 2008 and 2009 using the
simple float method. These discharge data were measured in or near the Rasau River sub-
catchement and are shown plotted against surface water head (relative to the channel bed) in
Figure 7.
0 0.2 0.4 0.6 0.8 1 1.2
h (m)
Q (m3/s)
Figure 7 – Stage (h in metres) - Discharge (Q in m
/s) relationship for several small canals in the
Sebangau National Park. Data clusters for canal 21 and the Rasau canals are indicated.
canal 21
Rasau canals
Despite the scarcity of discharge data in the Sebangau National Park, these data were fit to
the Chézy and Manning flow equations as shown in Appendix 4 to determine the most
applicable model for channel flow. From this comparison, the Chézy formula for flow velocity
appeared to have the best fit to the data and was thus chosen to generate stage-discharge
relationships for all other watercourses in the study area. The Chézy coefficient for a ‘typical’
watercourse in the Sebangau peatlands can be estimated by expressing the stage-discharge
relationship in terms of the logarithmic forms of corrected surface velocity (kv
) and hydraulic
radius (R
) as shown in Figure 8. R
is calculated by dividing the area of the canal cross-
section by the wetted perimeter. Assuming a rectangular cross-section, R
can be expressed
in terms of average surface water head (h
) and canal width (w) as described in Equation 1
(Section 2.3). Expressing the Chézy formula in logarithmic form can thus be used to solve for
unknown parameters using linear regression as shown in Figure 8 and the following equation:
log 2
CSRv h+=
[Equation 10]
where v is the corrected surface water velocity (the measured surface water velocity, v
multiplied by a constant correction factor k), R
is the hydraulic radius, C is the Chézy
coefficient, and S is the bed slope. Assuming a common bed slope of approximately
0.0006m/m throughout the area in which data were collected (the Rasau sub-catchment), the
Chézy coefficient was calculated to be approximately 10m
/s for the Sebangau canals.
y = 0.4742x - 0.5993
-1 -0.8 -0.6 -0.4 -0.2 0
Figure 8 – Logarithmic form of the Stage-Discharge relationship in Figure 7. The Chézy coefficient for
small canals was estimated based on a linear regression of this relationship.
Given marked differences in channel morphology, natural watercourses within the Sebangau
peatlands are expected to yield different Chézy coefficients. However, given the lack of
surface water data for these watercourses, it is not possible to work with accurate Chézy
coefficients for all watercourses. For this reason, Chézy coefficients used for calculating
stage-discharge tables for natural watercourses were increased to account for different canal
dimensions and roughness elements. Given the substantial uncertainty surrounding the
Chézy coefficient, a basic sensitivity analysis was undertaken to investigate the response of
SIMGRO-MODFLOW model to adjusted Chézy coefficients and channel bed slope values
and is described further in Section 3.2.3.
In order to strengthen the current dataset, WWF and TNS field staff are recommended to
select two or three small canals at least one blocked canal and one open canal and one
large canal (SSI) for regular monitoring of stage and discharge. On a bimonthly basis, stage
readings should be taken from the same location throughout the year. At the same time,
discharge measurements should be taken whenever possible within a stretch encompassing
the point at which stage was read. From these data, a channel rating curve (stage discharge
curve) can be constructed, and parameters adjusted in the preprocessing tool to fit the refined
4.2 Sensitivity to Surface Water Parameters
To test the sensitivity of the SIMGRO-MODFLOW model to the surface water parameters
described above, surface water levels in S.Bangah, one of the natural tributaries of
S.Sebangah, were simulated for the period between 2007 and 2009 inclusive, using varying
values for the channel bed slope and the Chézy coefficient. The results of the sensitivity test
are shown in Figure 9 and Figure 10. In both of the tests, it is clear that both the bed slope
and Chézy parameters have a larger effect during high flows, with virtually no effect during
drier periods. As summarized in Table 2, surface water head is inversely proportional to both
the Chézy coefficient and bed slope. This relationship is intuitive, as an lower Chézy
coefficient implies higher flow resistance in the channel, thus reducing flow velocity and
discharge and increasing water retention (and therefore water levels) within the channel.
Likewise, an increase in bed slope assumes an increase in energy slope and surface water
velocity in the channel, ultimately lowering the average surface water level over time.
Surface Water Head (m )
Figure 9 – Surface water hydrographs showing surface water head (m above channel bed) for the lower
reaches of S.Bangah with varying channel bed slope
Surface water head (m)
Chézy = 15
Chézy = 20
Chézy = 25
Chézy = 10
Figure 10 – Surface water hydrographs showing surface water head (m above channel bed) for the
lower reaches of S.Bangah with varying Chézy coefficients.
The hydrographs shown in Figure 9 and Figure 10 reveal the tendency of the SIMGRO-
MODFLOW to completely drain the natural watercourses during the dry periods. In reality,
water flows through the tributaries throughout the year, indicating that the model
underestimates surface water levels during these periods. This inaccuracy is likely linked to
drainage calculations performed by SIMGRO in transferring water from the drainage basin to
appropriate watercourses (discussed in detail in Section 4.3). Further calibration of the model
could therefore use surface water head (river or canal stage) for a given surface water
number (swnr) as focal output. However, there is currently a lack of river or canal stage data
from which such a comparison can be made. Future field monitoring plans should therefore
include surface water parameters (stage and discharge) in predefined rivers and canals in
order to establish a dataset for these parameters.
Variable Slope (m/m)
(C = 15m
/s) Mean SW
Head (m) Maximum
SW Head (m)
Variable Chézy (m
(S = 0.0003m/m)
Table 2 Results of surface water sensitivity tests. Mean and maximum surface water head (m) are
shown for each model run.
4.3 Surface Drainage Modeling
When infiltration of water into the soil column is limited by infiltrability or saturation of the soil,
transfer of this excess water to an appropriate watercourse as overland flow is an important
component of hydrological modeling. While some hydrological models route overland flow
from one cell to its neighbouring cells (especially those models concerned with soil erosion
and sediment delivery), SIMGRO directly transfers water determined to drain as overland flow
directly to watercourses as shown in Figure 11, based on predefined drainage basins.
Drainage basins for the study area were calculated using the DTM2CAT model (Alterra-
WUR), which predicts catchment areas based on watercourse and topography (DEM) grids.
Figure 12 shows the results of this calculation for the current situation (left) and the assumed
‘natural’ situation (right), where all small canals were removed. From this figure, the supposed
effects of small canals on overall drainage patterns is visible in green. According to these
results, water drained over a surface area of over 55,000 ha has been diverted to small
canals, most of which was originally draining into one of the tributaries in the study area. In
the SIMGRO drainage calculations, any surface water appearing in this area is transferred to
the appropriate small canal based on the schematic in Figure 11.
Figure 11 – Conceptualization of surface drainage to watercourses in SIMGRO. Depressions on the soil
surface (“field drains”) of each soil-vegetation-atmosphere transfer (SVAT) cell. Infiltration-limited or
saturation-limited surface water entering these drains is automatically transferred to the watercourse
according to predefined drainage basin areas and resistance parameters.
Figure 12 Calculated drainage basins for the study area including small canals (left) and excluding
small canals (right). Areas draining into S.Sebangau are shaded in light blue, areas draining into
tributaries or the large SSI canal are shaded in dark blue, and areas draining into small canals are
shaded in green.
Implementing SIMGRO’s surface drainage method, where drainage basins are explicitly input
into the model rather than routing surface water to neighbouring cells, requires thorough
knowledge of the watercourse network, including natural streams, rivers, and artificial canals
and ditches. In the case of the Sebangau project, the current knowledge of the extent of the
network of small drainage canals is lacking, which has implications on the quality of drainage
modeling by SIMGRO-MODFLOW. Calculated average groundwater levels between 1997
and 2007 using the standard SIMGRO drainage method show that small canals with large
drainage areas do not have the capacity to store the large amounts of water transferred from
their respective drainage basins (Figure 13, left). Instead, excess water seems to be retained
in the vicinity of the canal (labels A and B in Figure 13), which is contrary to the understanding
that drainage canals promote accelerated drainage and local drawdown of the water table
(DeVries 2010). The problem of insufficient drainage is partly a due to difficulties in predicting
watershed boundaries in low altitude wetlands as well as uncertainty surrounding the
behaviour of surface water during flooding periods. Implementing the alternate drainage
method described in Section 3.3 largely alleviated the problem of insufficient drainage as
shown in Figure 13 (right), where large areas of retained water (label A, left) were drastically
reduced (label C, right). However, since some of the canals in upstream reaches of the
tributaries were calculated to have very large drainage basins (Figure 12, left), this revision to
the drainage method was not sufficient to completely eliminate falsely retained water. A
complete watercourse map of the study area would be expected to alleviate this problem, as
additional watercourses would take on the excess water left behind the small canals.
Figure 13 Comparison of average groundwater levels between 1997 and 2007 predicted by
SIMGRO-MODFLOW using existing drainage modeling (left) and the modified drainage method (right),
where flood-waters are diverted to natural watercourses.
Figure 14 Comparison of average groundwater levels between 1997 and 2007 predicted by
SIMGRO-MODFLOW with the current watercourse map (left) and with the S.Bangah extension (right).
The extent of the added section of the river is indicated in yellow.
To demonstrate the potential impact of an improved watercourse map on model results, an
extension to the upstream reaches of S.Bangah was made based on free data available on
GoogleEarth (not shown). From these images, it was apparent that the current watercourse
map did not represent the river in its entirety, as S.Bangah extends a further 10km to the
northwest. Average groundwater levels between 1997 and 2007 were calculated using the
current watercourse map (Figure 14, left) and with the S.Bangah extension (Figure 14, right).
In the absence of the river extension, the small canal labeled ‘A’ in Figure 14 (left) still retains
a reservoir of water from its catchment area. Including the extension dramatically reduces the
volume of retained water, as the river extension cuts the canal’s drainage basin into a
presumably realistic area. By extending S.Bangah based on satellite imagery, the potential to
improve SIMGRO-MODFLOW for hydrological monitoring, and ultimately for calculating
carbon emissions and savings, has been demonstrated. Generating an improved watercourse
map through remote sensing techniques combined with ground truth data should therefore be
a priority for hydrological restoration work in the Sebangau National Park. While ground truth
data is essential in confirming canal locations and canal dimensions (and therefore flow
characteristics), appropriate remote sensing data provides the advantage of large spatial
coverage, which is especially vital given the accessibility of some of the areas of the
Sebangau peatlands.
4.4 Effect of Dams on Model Results
The central activity of the hydrological restoration of the Sebangau peatlands is the
construction of simple dams in artificial drainage canals, with the aim of retaining water and
promoting the rewetting of the surrounding peat. In small canals, simple dams were
constructed using two rows of gelam poles crossing the canal, in between which bags of
compressed peat soil were placed. Using DEM-derived slopes, ‘ideal’ dam spacing was
calculated to maintain a desired head difference across each dam (Jaenicke, Wösten et al.
2009; DeVries 2010). Since practical considerations often prevent the WWF and TNS field
teams to construct dams in these exact locations, the tool developed in this study allowed for
the dams locations to be explicitly provided by the user, as described in section 3.4.
The SIMGRO-MODFLOW model was run for the years 2007 to 2009 inclusive using a variety
of dam factor values to test the sensitivity of the model to this added parameter. The surface
water hydrograph for one swnr in the S.Bangah catchment (indicated by a black arrow in
Figure 16) are shown in Figure 15. From these hydrographs, it seems that the surface water
head relative to the canal bed is only affected by the presence of the dam in that swnr during
dry periods, whereas water levels in wet periods remain unchanged with increasing dam
strength. While this outcome may be an accurate qualitative description of the function of
dams, where water retention is expected to occur mostly during the dry season, the current
dam simulation method does not allow for ‘leakage’ of the dam beyond the adjusted canal
bed level. For example, in the hydrograph corresponding to the simulation where the dam
factor (df) is 0.3, the surface water head is not allowed to drop below 0.3m, as this is the
depth of the ‘virtual’ canal bed. Given the materials used to construct the dams, and the low
bearing capacity of the peat soil surrounding the dams (Islam and Hasim 2009), prolonged
dry periods may still lead to infiltration of water through the dam material and the surrounding
peat soil, allowing water levels to drop below the virtual canal bed level.
surface water head (m)
df = 0
df = 0.3
df = 0.6
df = 0.9
Figure 15 Predicted surface water hydrograph between 2007 and 2010 in a blocked surface water
number (swnr) in the S.Bangah catchement using various values for the dam factor (df). Surface water
head (m) is relative to the original canal bed.
Figure 16 – Average groundwater levels between 2007 and 2010 for the S.Bangah and SSI catchments
with varying dam factors (df). The location from which surface water hydrographs in Figure 15 is
indicated as a black arrow.
From the SIMGRO-MODFLOW model results using varying dam factor values, average
groundwater levels for each dam factor were calculated for the study area grid, and are
shown for the S.Bangah and SSI catchments in Figure 16. Here, the spatial extent of the
dams’ influence can be seen for each scenario. There is a marked difference between the
action of dams in the S.Bangah catchment and along the SSI canal, which can be explained
by a number of factors. First, the size of the catchment associated with the swnr being
blocked determines the volume of water entering that swnr over time, and therefore the
amount of water potentially blocked by the dam. The SSI canal was calculated to create a
large drainage basin to the north of the canal as shown in Figure 12, while most of the small
canals are associated with very small drainage basins. The second and probably most
influential factor explaining the difference in rewetting patterns between the two catchments is
the drainage method used. In the case of the S.Bangah catchment, floodwater were diverted
to the river itself to avoid artificial water retention in the canal itself as described in Section
3.3. Since the SSI canal is large and behaves like a natural river, this alternate drainage
method was not applied to the SSI catchment area. As a result, water retained in the canal is
not automatically diverted to another river system, and instead remains behind the dam and
contributes to the rewetting of the surrounding peat.
Another challenge with the dam simulation methodology arises from the placement of dams
on the watercourse map. The inconsistencies between the watercourse map and the actual
dam coordinates are evident in the top left panel of Figure 16. To circumvent this problem,
preprocessing codes were made such that users input a list of swnr’s that they wish to be
represented in the model as ‘blocked’. Although this strategy allows the users to easily
calculate the consequences of adding dams to the current watercourse map, it is impossible
to account for already existing dams when the actual canal is missing from the watercourse
map. A strategy for improving the current watercourse map, as discussed in section 4.3,
should therefore be a priority in future project plans.
Figure 17 Predicted rise in groundwater levels after including approximate locations of actual dams
(left; constructed as of May 2010), and suggested dams (right).
Deficiencies in the currently available watercourse map notwithstanding, it can still be used in
conjunction with the SIMGRO-MODFLOW model as a useful planning tool. Changes in
groundwater levels due to dam placement were calculated for the period between 2007 and
2010 and are shown in Figure 17. In this figure, it is clear that with the suggested dam
locations, the rewetting effects are spread much further along the canals. This effect is likely
due to the fact that in an area of very low slopes, as in the S.Bangah catchment, optimal dam
spacing for maintaining a constant head difference across each dam is higher than for
catchments with higher slope. Closely spaced dams in the Bangah catchment are therefore
expected to give rise to redundancy, with some dams maintaining little to no head difference
(DeVries 2010). As a result, the potential of the dams for rewetting of surrounding peat is not
fully realized, as is evident in the Figure 17 (left). An improved dam simulation procedure may
include considerations of the number of dams per swnr, as well as the average slope of the
swnr. The difference between extent of rewetting using actual dam locations and suggested
locations based on a constant head difference demonstrates the potential use of the model as
a tool for maximizing the spatial impact of the dams.
4.5 Model Performance
The SIMGRO-MODFLOW model has been calibrated and used in previous hydrological
studies in Central Kalimantan peatlands, covering the Sebangau National Park as well as
neighbouring developed peatlands to the east (Wösten, Clymans et al. 2008). Although this
study uses many of the same parameters as that of Wösten et al. (2008), it also uses a more
recent release of the SIMGRO-MODFLOW model. Since the scale, objectives, and partners
involved in the current project are different from that of Wösten et al. (2008), additional
validation of the model results against measured data is necessary in the development of a
hydrological monitoring tool. To this end, groundwater hydrographs derived from measured
data at the WWF/TNS SSI field station were compared with hydrographs extracted from the
model at corresponding SVAT cells. The SIMGRO-MODFLOW model was run without dams
and with dams (with a dam factor of 0.9) for a location 250m north of the SSI canal (Figure
18) and for a location adjacent to the SSI canal (Figure 19). Root mean square error (RMSE)
and normalized RMSE (nRMSE) were calculated for each of these model runs, and are
shown in Table 3.
2-1-2005 17-5-2006 29-9-2007 10-2-2009
Groundwater level (cm)
no dams
dams (df = 0.9)
Figure 18 Groundwater hydrographs derived from SIMGRO-MODFLOW outputs without dams (blue
line) and with a dam factor of 0.9 (green line) at a location 250m north of the SSI canal. Measured
groundwater levels from the same location (SEL5) are shown as pink points.
2-1-2005 17-5-2006 29-9-2007 10-2-2009
Groundwater level (cm)
no dams
dams (df = 0.9)
Figure 19 Groundwater hydrographs derived from SIMGRO-MODFLOW outputs without dams (blue
line) and with a dam factor of 0.9 (green line) at a location adjacent to the SSI canal. Measured
groundwater levels from the same location (SCL1) are shown as pink points.
SEL5 model
(w/o dam) model
(w/ dam) SCL1 model
(w/o dam) model
(w/ dam)
RMSE 60cm 56cm 80cm 61cm
nRMSE 28% 26% 36% 28%
Table 3 – Root mean square error (RMSE in cm) and normalized RMSE (nRMSE as %) for the model
runs shown in Figure 18 and Figure 19.
Though general patterns of groundwater drainage and recharge are observed in both
measured and predicted hydrographs, sub-surface drainage in dry periods exaggerated by
the model in all cases. This consistent overestimation of sub-surface drainage rates may also
explain why model output grids shown earlier in this report consistently show average
groundwater levels that are much lower than expected. Although the objective of this study
was not to re-calibrate the SIMGRO-MODFLOW model for the Central Kalimantan peatlands,
further calibration is indeed needed for the model to be effective as a monitoring tool, focusing
especially on reducing the rate of groundwater drawdown during the dry season. Previous
exercises using the SIMGRO-MODFLOW model for Central Kalimantan peatlands have
shown that model results are sensitive to changes in drainable pore space and specific yield
in the unsaturated zone of the soil matrix (Clymans 2005; Wosten, personal communication
2010). If such changes are successful in increasing groundwater levels during dry periods,
these parameters may indeed be ideal candidates for further calibration of the model.
Between the two locations, a better fit was accomplished at the SEL5 location, 250m away
from the canal. Here a normalized root mean square error (nRMSE) for simulation with and
without dams was found to be 28% and 26% respectively, as opposed to 36% and 28%
respectively for the SCL1 location (adjacent to the canal). In both cases, the nRMSE can be
considerably reduced (indicating higher model precision) by correcting the overestimation in
drawdown in dry periods. Adjacent to the canal at SCL1, there is a slight difference between
model results with and without dams, especially in dry periods. However, the effects of canal
blocking are still noticeably underestimated by the model with and without dams. This
inaccuracy suggests that more work is needed to optimize the simulation of water retention by
the SSI dam.
4.6 Calculating Carbon Emissions and Savings
A major objective of the Sebangau Conservation Project is to link hydrological restoration
efforts with reductions in carbon emissions. This interest follows global concern over the
levels of carbon emissions resulting from land use change in tropical peatlands. Nearly half of
all peatlands were estimated to have been deforested and drained by 2006, resulting in
estimated emissions between 355Mt/y to 855Mt/y (Hooijer, Page et al. 2010). As Indonesia
contains the majority of these tropical peatlands, considering emissions from such land use
change places Indonesia among the top emitters of CO
in the world. The government of
Indonesia has thus placed considerable interest in the implementation of schemes to reduce
the rate of deforestation of its peatlands, including UN-REDD (The United Nations
Collaborative Programme on Reducing Emissions from Deforestation and Forest Degradation
in Developing Countries) and the Voluntary Carbon Standard (Couwenberg, Dommain et al.
2010). The restoration activities of the Sebangau National Park Authority (TNS) and WWF-
Indonesia in the Sebangau peatlands fall in line with these schemes. Estimates of the impact
of hydrological restoration activities on carbon emissions from the Sebangau peatlands must
therefore be based on robust and verifiable methodologies (UNFCCC 2007).
In this study, a direct connection between average groundwater depth and CO
from tropical peatlands was drawn based on Couwenberg et al (2010). Here, the authors
conducted a meta-analysis of existing peat subsidence and emissions data to arrive at an
estimate of 900gCO
/y for every 10cm of drainage depth (Couwenberg, Dommain et al.
2010). Expressed per unit volume of peat, this quantity becomes 0.009t/m
/y, which can be
used to estimate carbon emissions based on average groundwater depths over an entire
region as described in Section 2.6. In this study, the soil-vegetation-atmosphere transfer
(SVAT) units of the SIMGRO model were used as strata for the carbon calculations, allowing
for easy translation of SIMGRO-MODFLOW output grids into estimated emissions grids as
shown in Figure 5, Section 2.6.
Total Emissions (Mt)
No dams
Figure 20 Estimated CO
emissions per year over the entire study area (150,000ha) without dams
(red bars) and with dams (blue bars).
Using the carbon calculation methods described in section 2.6, total emissions were
estimated over the entire study area between 1997 and 2009. The total emissions per year (in
megatons per year; Mt/y) are shown in Figure 20. During this period, average emissions per
hectare were calculated to be 43t/ha/y and were found to fluctuate inversely with annual total
precipitation (Figure 4, Section 2.1). In comparison, Hooijer et al. (2010) estimated CO2
emissions from large croplands, mixed cropland, and shrubland to be 86t/ha/y, 48t/ha/y, and
15t/ha/y, respectively. While a drained peat swamp forest would be expected to have
emissions lower than these three land use types, the fact that the estimate from this study
falls within the same range as other studies indicates that the SIMGRO-MODLOFW model is
an applicable tool for estimating average annual CO
emissions. This estimate can be
improved with further calibration of the model, as the current form of the model was found to
consistently over-predict the groundwater depths throughout the study area (described in
section 4.5).
Estimations of CO
emissions over the study area were calculated for the same time period
using the hypothetical dam locations shown in Figure 17 (right panel) and are shown in Figure
20 as blue bars. Compared to total emissions over the entire study area, the change in
emissions due to canal blocking seems very small. On average, installing dams at these
locations before 1997 would have prevented the emission of nearly 56,000tCO
/y over the
entire study area (approximately 150,000 ha). Since a quantitative analysis of the effect of
canal blocking on CO
emissions in tropical peatlands has not been carried out before this
study, it is difficult to verify the validity of this estimation. However, the revised drainage
procedure described in Section 3.2.4 may indeed overestimate the drainage of floodwater in
the canal, which would in turn underestimate the extent of rewetting by the dams. With
improved watercourse data and a better understanding of the function of dams over time, this
drainage procedure can be modified, and a revised estimate of the prevented emissions by
canal blocking can be calculated.
Figure 21 Predicted change in emissions (E; tonnes per year) over the S.Bangah and SSI
catchments using suggested dam locations (left) and actual dam locations as of May 2010 (right).
To gain insight into the spatial patterns of CO
emissions with and without dams over the
study area, average groundwater levels over a 2-year simulation were converted to carbon
emissions according to the method described in Section 2.6. The difference in calculated
emissions E between the ‘unblocked canal’ and ‘blocked canal’ scenario was calculated
given the locations of suggested dams and those of actual dams, as shown in Figure 21. Not
surprisingly, the spatial patterns of carbon savings due to dam construction are identical to
the spatial rewetting patterns evident in Figure 17, given the linearity of the relationship
between drainage depth and carbon emissions. Since this relationship is bounded by the soil
surface (positive groundwater depths are assumed to have zero emissions), the linearity
breaks down where SVAT’s are calculated to have an average positive groundwater level. As
the SIMGRO-MODFLOW currently seems to consistently underestimate groundwater levels,
this trend is not evident in Figure 21. With further calibration and improved model results,
however, regions where this boundary conditions is met may become visible when carbon
savings are calculated. The physical significance of such a phenomenon can be explained by
the fact that a saturated soil column does not allow for oxidation of soil carbon, a reality which
is assumed to be unaffected by increasing floodwater levels above the saturated soil column.
The method for calculating carbon emissions and savings based on canal blocking activities
presented in this section is a simple tool that can be used for planning and monitoring
purposes in the Sebangau conservation project. There are several factors, however, that cast
a shadow of uncertainty over the formula, including the extent of the linearity of carbon
emissions and drainage depth, variance of local topography, variance of peat depth, and the
effects of forest fires on carbon emission. These aspects warrant a discussion on the extent
of the tool’s validity in calculating carbon emissions from the Sebangau peatlands.
Though data on carbon emissions from tropical peatlands are not as abundant for their
temperate and boreal counterparts, it has been suggested that the linear relationship
presented in this report holds true for drainage depths up to 50cm below the soil surface
(Couwenberg, Dommain et al. 2010). Factors introducing non-linearity to this relationship
include an increase in bulk density in subsided peat (Couwenberg, Dommain et al. 2010) and
different reactions to temperature variation among different peat material (Davidson and
Janssens 2006). As has been noted for temperate peatlands, emissions at drainage depths
below this maximum significantly slow down, likely owing to a factor which limits the microbial
oxidation of organic carbon to CO
. An additional boundary condition should therefore be
added to the relationship used in this study, whereby SVAT’s with lower drainage depths than
that of the equation’s boundary are assigned a different emissions constant, accounting for
lower emission rates. For such a boundary condition to be introduced, more data is needed
on the relationship between CO
emissions and peat drainage in tropical peatlands with very
low groundwater levels than is currently available.
Another feature of tropical peatlands which may affect rates of carbon emissions is the local
topography. The floor of the peat swamp forests in Central Kalimantan is composed of
hummocks (raised surfaces associated with tree roots) surrounded by a network of hollows,
which can give rise to variations within a small area of over 50cm (Shimamura and Momose
2005). Microtopography has implications on both the quality of the DEM used in this study
(DeVries 2010) as well as actual carbon emissions because of differences in inundation
periods (Jauhiainen, Takahashi et al. 2005). However, these local scale variations in carbon
emissions are likely cancelled out due to the fact that this study was carried out on a
landscape scale.
Variance of peat depth over the study area is a factor that likely warrants further modifications
to the carbon calculation tool presented in this report. The Sebangau peatlands are believed
to from a ‘dome’ shape, where surface elevation increases moving inwards from S.Sebangau
and S.Katingan at the eastern and western edges, respectively, to a maximum depth of over
10m in the centre (Page, Rieley et al. 1999; Jaenicke, Rieley et al. 2008). At the edges of the
peat dome adjacent to a natural river, peat depths are typically shallow, under which a
mineral soil layer is found. At such depths, emissions of soil carbon are not expected to follow
the same relationship presented in this report. As such, it may be necessary to introduce
another boundary condition for such SVAT’s, where the depth of the peat layer is the
maximum drainage depth at which emissions occur. While limited data regarding peat depths
continues to challenge researchers interested in estimating carbon storage in tropical
peatlands (Hooijer, Page et al. 2010), there are likely very few SVAT’s in the study area
where average groundwater levels descend below the peat-mineral soil interface. A revised
stratification method for aggregating carbon emissions should nevertheless consider peat
depth classes, applying this boundary where applicable.
A final factor not accounted for by the carbon calculation methodology presented in this report
is the effect of peat firest on carbon emissions. During year in which the El Niño Southern
Oscillation (ENSO) occurs, Central Kalimantan experiences relatively low rainfall and an
extended dry season, as is evident in Figure 4 (Section 2.2), substantially increasing the risk
of forest fire. Peat soils are very porous, and are therefore quickly aerated after drainage,
making them very sensitive to fires during dry periods. During 1997, an especially severe
ENSO-year, carbon release from peat fires alone in Central Kalimantan was estimated to be
up to 0.23 gigatonnes, not include that of burnt vegetation (Page, Siegert et al. 2002). Large-
scale peat fires, occurring especially on extensively drained peatland (Wösten, Clymans et al.
2008), are thus responsible for accelerated carbon release. While there is not enough data on
fire incidence associated carbon release in the Sebangau National Park, continued monitoring
of fire events is an important feature of a monitoring plan, as predicted carbon emissions will
inevitably be affected by incidence of peat fires.
5. User Feedback and Conclusions
The hydrological monitoring and carbon calculation tools developed in this study and
presented in this report were delivered to WWF-Indonesia and Taman Nasional Sebangau
(TNS) during training in Palangkaraya, Central Kalimantan, Indonesia, in July 2010. In the
days following the training, participants were able to successfully prepare input data with the
aid of the Visual Basic 6.5 macros developed in this study and run the SIMGRO-MODFLOW
model. The participants also carried out simple carbon emission and carbon savings
calculations in ArcGIS using the SIMGRO output grids. File packages containing all input
files, executable files and preprocessing macros were handed over to WWF and TNS staff for
use in their monitoring programmes. These folders contain all the files needed to successfully
run the model and to visualize the results.
Following the training, participants were given the opportunity to give their feedback on the
model theory and implementation in the context of the restoration project. From these
discussions, two main themes relating to constraints to hydrological modeling in Sebangau
National Park arose. First, the quality of the watercourse map was identified as a major
constraint to accurately modeling surface water drainage and the placement and effect of
dams. Second, some of the pre-processing methods pertaining to dam simulations do not
satisfactorily represent the reality of the restoration work on the ground.
The quality of watercourse data has a large effect on the model results. First, inaccurate or
incomplete canal and river courses lead to incorrect calculation of hydrological catchment
areas using the DTM2CAT model, which in turn leads to the misdirection of drainage water to
these canals. As a result, water levels tend to appear elevated near some canals, where they
would expect to be decreased due to increased drainage. As shown in Figure 14 (Section
3.2.4), extension of S.Bangah according to GoogleEarth imagery already improved some of
the anomalies seen in the S.Bangah catchment. A more accurate and extensive watercourse
map would therefore be expected to improve drainage simulations across the study area.
Second, where canal courses do not correspond to actual dam locations (as measured by the
WWF-Indonesia Sebangau field team), the modeled dam locations were estimated by eye.
While this method is satisfactory for obtaining results for blocked canals, a more accurate
watercourse map would allow for easier and more consistent placing of dams in the model
preprocessing stage.
From the development of the Sebangau hydrological monitoring and carbon calculation tools
described in this report, it is clear that these tools require further refining before they can be
used towards verifying data for voluntary carbon trading schemes (UNFCCC 2007). A number
of concrete recommendations are included herein to integrate hydrological modeling, remote
sensing data, and field monitoring towards improving the current methodologies. First,
improving the model’s simulation of drainage through watercourses requires a more accurate
watercourse map. An effective strategy to improve the watercourse map of the Sebangau
National Park will be comprised of a combination of methods, including the sourcing of high-
quality remote sensing data and ground truthing or surveying methods. Remote sensing data
may include high-resolution satellite images as well as aircraft-mounted LIDAR or
photography. Taman Nasional Sebangau (TNS; Sebangau National Park) has access to an
ultralight aircraft, which may be used to capture images of certain parts of the National Park.
Surveying data by the field team should include GPS-based mapping of canals and rivers.
This data can not only be used for generating updated maps of the region, but also for
verifying remote sensing data. A detailed and comprehensive plan for improved mapping of
the Sebangau watercourse network will likely be comprised of a combination of the above
data sources and should be formulated involving all of the relevant parties involved in the
Sebangau restoration project (WWF-Indonesia field staff, TNS staff, and technical staff from
WWF-Indonesia in Jakarta).
Second, in its current state, the simulation of dams using the ‘dam factor’ does not fully
represent the action of a dam cascade. The simulation of dams in SIMGRO-MODFLOW
should therefore be altered to take into account the work of several dams over the span of
one swnr. To this end, one of two strategies could be employed. First, the length of swnr’s
could be reduced from their current 1km, to allow for more precision in the blocking of swnr’s.
Second, an algorithm could be written which takes into account the number of dams per swnr,
and the average slope of each swnr. The second method accounts for the fact that closely
spaced tandem dams on an area of low slope tend to maintain little to no head difference, and
therefore become redundant (DeVries 2010).
Simulation of dam function can also be improved with increased availability of field data,
specifically relating to surface water discharge from blocked and unblocked canals, and head
difference across specific dams. Surface water monitoring is very important for understanding
the effects of canal blocking on surface water hydrology. Field staff are recommended to
select several canals for regular monitoring (once every 2 weeks), including at least one
blocked canal and one open canal. Measurements should include stage (water level above a
stationary reference point) and discharge, and should be taken from the same location each
time. In case discharge cannot be measured (during flooding periods, when water in some
canals becomes stagnant), the stage should be measured as before, accompanied by a
qualitative description of the state of the surface water. Prolonged records of stage and
discharge can be used to come to arrive at a better expression of surface water discharge
using the Chézy (or other) formula for each watercourse.
To come to a better understanding of dam function, head difference over several
representative dams should be measured at regular time intervals (once every 2 weeks). The
same dams should be measured every time, and a record over time of head difference should
be kept. These data will help to understand the change in dam function over time, and will
also aid in improving the simulation of dam function in SIMGRO-MODFLOW.
6. References
Baird, A.J., 1997. Field Estimation of Macropore Functioning and Surface Hydraulic
Conductivity in a Fen Peat. Hydrological Processes, 11(3): 287-295.
Camporese, M., Ferraris, S., Putti, M., Salandin, P. and Teatini, P., 2006. Hydrological
modeling in swelling/shrinking peat soils. Water Resour. Res., 42(6): W06420.
Chimner, R., 2004. Soil respiration rates of tropical peatlands in Micronesia and Hawaii.
Wetlands, 24(1): 51-56.
Clymans, E., 2005. Hydrological modelling for protection and restoration of degraded tropical
peatlands. Wageningen University and Research Center, Wageningen, The
Couwenberg, J., Dommain, R. and Joosten, H., 2010. Greenhouse gas fluxes from tropical
peatlands in south-east Asia. Global Change Biology, 16(6): 1715-1732.
Davidson, E.A. and Janssens, I.A., 2006. Temperature sensitivity of soil carbon
decomposition and feedbacks to climate change. Nature, 440(7081): 165-173.
DeVries, B., 2010. Monitoring the Effects of Hydrological Restoration Efforts in Degraded
Tropical Peatlands. Central Kalimantan, Indonesia., Wageningen University and
Research Center, Wageningen, The Netherlands.
Harbaugh, A.W., Banta, E.R., Hill, M.C. and McDonald, M.G., 2000. MODFLOW-2000, The
U.S. Geological Survey Modular Ground-Water Model - User Guide to Modularization
Concepts and the Ground-Water Flow Process, United States Department of the
Interior and United States Geological Survey.
Hooijer, A. et al., 2010. Current and future CO2 emissions from drained peatlands in
Southeast Asia. Biogeosciences, 7(5): 1505-1514.
Islam, M.S. and Hasim, R., 2009. Bearing Capacity of Stabilised Tropical Peat by Deep
Mixing Method. Australian Journal of Basic and Applied Sciences, 3(2): 682-8.
Jaenicke, J., Rieley, J.O., Mott, C., Kimman, P. and Siegert, F., 2008. Determination of the
amount of carbon stored in Indonesian peatlands. Geoderma, 147(3-4): 151-158.
Jaenicke, J., Wösten, H., Budiman, A. and Siegert, F., 2009. Planning hydrological restoration
of peatlands in Indonesia to mitigate carbon dioxide emissions. Mitigation and
Adaptation Strategies for Global Change, 15(3): 223-239.
Jauhiainen, J., Takahashi, H., Heikkinen, J.E.P., Martikainen, P.J. and Vasander, H., 2005.
Carbon fluxes from a tropical peat swamp forest floor. Global Change Biology, 11(10):
Kool, D.M., Buurman, P. and Hoekman, D.H., 2006. Oxidation and compaction of a collapsed
peat dome in Central Kalimantan. Geoderma, 137(1-2): 217-225.
Kruse, J., Lennartz, B. and Leinweber, P., 2008. A modified method for measuring saturated
hydraulic conductivity and anisotropy of fen peat samples. Wetlands, 28(2): 527-531.
Ong, B.Y. and Yogeswaran, M., 1994. Peatland as a resource for water supply in Sarawak.,
Tropical Peat: Proceedings of the International Symposium on Tropical Peatland, pp.
Page, S.E., Rieley, J.O. and Banks, C.J., 2010. Global and regional importance of the tropical
peatland carbon pool. Global Change Biology.
Page, S.E., Rieley, J.O., Shotyk, Ã.W. and Weiss, D., 1999. Interdependence of peat and
vegetation in a tropical peat swamp forest. Philosophical Transactions of the Royal
Society of London. Series B: Biological Sciences, 354(1391): 1885-1897.
Page, S.E. et al., 2002. The amount of carbon released from peat and forest fires in Indonesia
during 1997. Nature, 420(6911): 61-65.
Schwarz, A., 2010. Low carbon growth in Indonesia. Bulletin of Indonesian Economic Studies,
46(2): 181 - 185.
Shimamura, T. and Momose, K., 2005. Organic matter dynamics control plant species
coexistence in a tropical peat swamp forest. Proceedings of the Royal Society B:
Biological Sciences, 272(1571): 1503-1510.
UNFCCC, 2007. Report of the Conference of the Parties on its thirteenth session, held in Bali
from 3 to 15 December 2007. Decisions adopted by the Conference of the Parties.
Walsum, P.E.V.v. and Groenendijk, P., 2008. Quasi Steady-State Simulation of the
Unsaturated Zone in Groundwater Modeling of Lowland Regions. Vadose Zone
Journal, 7(2): 769-81.
Walsum, P.E.V.v., Veldhuizen, A.A. and Groenendijk, P., 2010. SIMGRO 7.1.0, Theory and
model implementation, Alterra -WUR, Wageningen.
Wösten, J.H.M., Clymans, E., Page, S.E., Rieley, J.O. and Limin, S.H., 2008. Peat-water
interrelationships in a tropical peatland ecosystem in Southeast Asia. CATENA, 73(2):
Wösten, J.H.M., Ismail, A.B. and Wijk, A.B.M.v., 1997. Peat subsidence and its practical
implications: a case study in Malaysia, 78. Elsevier, Amsterdam, The Netherlands.
Wüst, R.A.J., Bustin, R.M. and Lavkulich, L.M., 2003. New classification systems for tropical
organic-rich deposits based on studies of the Tasek Bera Basin, Malaysia. CATENA,
53(2): 133-163.
Appendix 1 – SIMGRO-MODFLOW Model Theory
A1.1 Overview of the SIMGRO-MODFLOW Model
A key feature of the SIMGRO-MODFLOW model is its ability to encompass the hydrological
features of an entire region, while taking into account feedback between sub-regional systems
(Walsum, Veldhuizen et al. 2010). To integrate hydrological processes and allow for
exchange of hydrological information between the different domains throughout a given
region, the unit cells for both SIMGRO and MODFLOW model components are explicitly
connected. In this way, excess water from the soil-vegetation-atmosphere transfer (SVAT)
domain of SIMGRO is passed onto corresponding MODFLOW cells, where water flow in the
saturated groundwater domain is modeled. The surface water component of SIMGRO, SurfW,
concurrently models surface water flow through watercourses using explicit stage-discharge
tables for each watercourse. Taken together, the SIMGRO-MODFLOW model (including
SurfW processes by default) requires spatially referenced data including a variety of
parameters. Figure A1-
shows a simplification of this schematic: moving from bottom to top,
the bottom layer represents the land-use/soil map supplied to the model, followed by the
groundwater hydrology flows, followed by a surface water channel map. The top map refers to
the overlay of these three representations.
Figure A1- 1 Visual representation of information layers in SIMGRO-MODFLOW. The bottom layer
represents land-use/soil maps (SIMGRO); the second layer from the bottom represents groundwater
flows (MODFLOW); the second layer from the top represents surface water channels (SIMGRO-SurfW);
the top layer represents the composite of all three layers (Walsum, Veldhuizen et al. 2010).
In this study, SIMGRO and MODFLOW time steps were set to one day, corresponding to the
frequency of precipitation data available. Within this time step, the model calculates water
balances at several levels, conceptualized as storage elements. The flow of hydrological
information between storage elements is shown in Figure A1-
. The coupling of
functionalities between above-surface storage reservoirs (SIMGRO), watercourse reservoirs
(SIMGRO-SurfW) unsaturated soil moisture storage reservoirs (SIMGRO-MetaSWAP), and
saturated groundwater storage reservoirs (MODFLOW) are key to the functioning of the
SIMGRO-MODFLOW model as a whole. A description of above-surface, unsaturated, and
saturated storage reservoirs are described below based on the SIMGRO and MODFLOW
manuals (Harbaugh, Banta et al. 2000; Walsum, Veldhuizen et al. 2010). A method for
modeling surface water flow through watercourses was developed in this study based on the
SurfW component of SIMGRO, and is described in more detail in Section 4.1. The simulation
of surface drainage from the SVAT domain to the surface water (SurfW) domain is also an
important feature which was modified in this study, and is thus described further in Section
3.3. Finally, the modeling of water management features (dams) was tailored to the
Sebangau situation based on demands by the users, and is discussed in Section 3.4.
Figure A1- 2 Overview of storage elements and transmission links in the SIMGRO model (Walsum,
Veldhuizen et al. 2010). Since the Sebangau National Park not used for agriculture, springling links and
subsurface irrigation were not considered in this study.
A1.2 Above-Surface Storage
Hydrological processes above the soil surface (not including surface water flow in channels)
are simulated in soil-vegetation-atmosphere transfer (SVAT) cells as described in section
A1.1. Within each SVAT, potential reservoirs include interception storage reservoirs due to
vegetation canopy and depression storage on the soil surface. Interception storage refers to
the extent at which rainfall is captured by the vegetation canopy. This storage capacity is
therefore a function of canopy cover which is inferred from vegetation and land use input
data. In this study, land use type was assumed to be uniformly “peat swamp forest”
throughout the study area.
Depression storage, on the other hand, depends primarily on soil characteristics and
topography. Storage of water in the soil depression reservoir is a result of one of two
phenomena. First, water ponds on the soil surface when rainfall intensity exceeds the
maximum infiltration rate of the soil. Second, water can appear on the soil surface as ‘visible’
groundwater, which occurs when the soil column is completely saturated. Given the high
conductivity of tropical peat forest soils and the persistently high water table, infiltration-limited
depression storage is likely to be negligible (Wösten, Clymans et al. 2008), while ‘visible’
groundwater is a common phenomenon during flooding periods. This assumption may not be
completely valid in the small regions of the Sebangau National Park, however, where forests
have been severely disturbed. Although the hydraulic properties of the soils are likely altered
in such cases, there is not enough data available to describe these changes, and such
deviations in land use were not considered in this study.
A1.3 Soil Moisture Storage
The simulation of water storage in the unsaturated layer of the soil column is accomplished
using a ‘meta-modeling’ technique, where the MetaSWAP model is first employed to construct
soil water storage tables given specific soil physical parameters before running the SIMGRO-
MODFLOW model (Walsum, Veldhuizen et al. 2010). The MetaSWAP model simulates
unsaturated soil storage through the use of a simplified version of Richard’s Equation
(Walsum and Groenendijk 2008):
( ) ( )
0,1 =Ψ
[Equation A 1]
where z is gravity head, K is the variable unsaturated hydraulic conductivity, Ψ is the pressure
head, and
is the root water extraction as a function of pressure and gravity head. This
equation is bounded at the top by the soil surface and at the bottom by the variable phreatic
level (Walsum and Groenendijk 2008), which in the case of a peat swamp is coincident with
the natural water table.
Although an evaluation of the MetaSWAP functionality is beyond the scope of this study, two
aspects of peat hydrology may affect the application of Richards Law to hydrological modeling
in tropical peatlands. First, the sole application of Richards Law to model flow in the
unsaturated domain has been questioned in the case of temperate peat fens on the basis that
it fails to sufficiently account for flow through macropores in addition to the soil matrix (Baird
1997). Since macropore flow through the woody debris commonly found in fibric tropical peat
is likely even more predominant than found in temperate peat (Ong and Yogeswaran 1994;
Kruse, Lennartz et al. 2008). Second, tropical and temperate peat soils are known to go
through swelling and shrinking cycles on both a diurnal and seasonal basis due to the effect
of changing pore water pressure on matrix structure (Camporese, Ferraris et al. 2006). It has
been proposed that Richards Law be modified in such cases to account for changes in pore
size and overall porosity with changing moisture content in peat soils (Camporese, Ferraris et
al. 2006). Taking into account these two factors, Richards Law may fall short of accurately
modeling soil moisture storage in tropical peatlands.
A1.4 Groundwater Flow
MODFLOW was developed by researchers in the United States Geological Survey (USGS) to
simulate three-dimensional groundwater flow through porous media (Harbaugh, Banta et al.
2000). Whereas SIMGRO functionalities include the direct transmission of water from one
reservoir to another without regard to spatial routing, MODFLOW is in fact a groundwater
routing model. The MODFLOW model is centred on the solution of the groundwater flow
equation, which describes the change in hydraulic head (h) over time multiplied by the
specific storage of the media (S
) as a function of
the three-dimensional hydraulic gradient,
the hydraulic conductivity tensor (K), and an additional source/sink term (W) (Harbaugh,
Banta et al. 2000).
[Equation A 2]
For the purpose of modeling lateral groundwater flow in the Sebangau peatlands, the vertical
peat profile was divided into two layers. These layers roughly represent the fibric/hemic and
sapric zones, with saturated hydraulic conductivities assumed to be on the order of 30md
and 0.5md
respectively (Ong and Yogeswaran 1994; Wösten, Clymans et al. 2008).
Although these hydraulic conductivities are supported by a previous calibration exercise using
data from the Sebangau National Park (Wösten, Clymans et al. 2008), the unique properties
of tropical peat soils may not be completely captured by modeling groundwater flow using
Darcy’s Law. Water flow through tropical peat is believed to depart from Darcy’s Law, largely
due to the prominence of woody material and the resulting macropore structure in the peat.
Specifically, Reynold’s number and specific discharges in the soil tend to exceed the limit
below which Darcy’s Law is applicable. (Ong and Yogeswaran 1994). The current
representation of the Sebangau peatlands in MODFLOW does not account for mineral layers
beneath the peat. While this omissions is of little consequence for areas with thick peat
layers, those regions bordering the natural rivers have very thin peat layers, and underlying
mineral soil layers may affect the movement of water.
Appendix 2 – Precipitation Data
Precipitation data were collected from a number of different sources, described in section 2.2.
In most cases, precipitation data were available at a daily time resolution. In the case of data
gathered from the SSI field station (WWF/TNS; Figure A2-7), data were available at an hourly
rate and were converted to total daily precipitation before entering into the SIMGRO-
MODFLOW model.
Precipitation (mm/d)
Figure A2- 1 – Daily precipitation data from Palangkaraya Airport between 1-1-1997 and 13-5-2002.
Precipitation (mm/d)
Figure A2- 2 – Daily precipitation data from Setia Alam (CIMTROP) research station between 14-5-
2002 and 17-9-2002.
Precipitation (mm/d)
Figure A2- 3 – Daily precipitation data from Palangkaraya Airport between 18-9-2002 and 15-3-2003.
Precipitation (mm/d)
Figure A2- 4 – Daily precipitation data from Setia Alam between 16-3-2003 and 14-10-2003.
Precipitation (mm/d)
Figure A2- 5 – Daily precipitation data from Palangkaraya Airport between 15-10-2003 and 18-3-2004.
Precipitation (mm/d)
Figure A2- 6 – Daily precipitation data from Setia Alam between 19-3-2004 and 8-10-2008.
Precipitation (mm/h)
Figure A2- 7 – Hourly precipitation data from the SSI field station (WWF/TNS) between 9-12-2008 and
Appendix 3 – Surface Water Data
The surface water modeling techniques used in this study were based on data available from
WWF/TNS field monitoring data between 2007 and 2009. Discharge from several small
canals was measured in the S.Rasau catchment in December 2007, coinciding with the wet
season. Two measurements were made in the mouth of canal 21 in July 2009, at the driest
point of the year (2009 was an exceptionally dry year). Although consistent stage-discharge
records for several canals throughout a year would be ideal for constructing the stage-
discharge input data for the SIMGRO (SurfW) model, they are not available at this time.
Existing data from Rasau canals and canal 21 were therefore used in their place.
Figure A3- 1 Locations of Stage-discharge measurement points. Two measurements were taken at
the same location in canal 21, while several canals in the S.Rasau catchment were sampled individually.
Surface water velocity was measured with the float method using pieces of waterlogged wood
approximately 10cm long and 2-3cm in diameter. The floats were released approximately 1
meter upstream of the start point (0m) to allow for acceleration to maximum velocity, and
timing began at 0m. The float travel time was recorded for 3 separate trials and averaged.
Surface velocity (v
) was calculated by dividing the length (L) by the average travel time.
Discharge (Q) was then estimated by multiplying average velocity (v) by the average cross-
section area (A) according to Equation 4 (Section 2.3). Since the surface velocity is not
necessarily representative of the vertical velocity profile of the canal, a correction factor k of
0.85 was used to convert surface velocity (v
) to average velocity according to the equation:
)85.0(, == kkvv s
[Equation A 3]
This methodology was tested in canal 21 during the month of July 2009, and a discharge of
/s was obtained. This result was compared to direct discharge measurements from
the spillway of the first dam on the same day. A calibrated forty-litre basin was placed
Canal 21
beneath the spillway and all the water was funnelled into the basin. The average time (over 3
trials) needed to fill the basin was recorded and discharge was calculated by dividing 40L by
the average time. Average discharge was thus estimated to be 0.012m
/s, which is
comparable to that obtained using the float method. However, the direct discharge suffers
from the disadvantage that the time needed to fill the basin is very short (on the order of 3
seconds). As the rains increase in the wet season, this time will decrease, and accurate time
measurements will be very difficult to obtain for this volume. The float method described
herein is therefore recommended for small canals (up to 3m in width).
Figure A3- 2 – Schematic of canal discharge measurements. Average cross sectional areas for the start
and end points (striped areas) were calculated by taking the average depth and canal width (top), and
discharge estimated by taking surface velocity and average cross sectional area for a stretch (bottom).
Raw data obtained from the Rasau canals and canal 21 are summarized in the table on the
following page.
Surface water data from Rasau canals and canal 21:
Date Region Latitude Longitude h1(m) h2(m) h3(m) h
(m) W(m) A(m
) L(m)
1 27-12-2008 Rasau -2.489 114.017 0.99 1.20 0.82 1.00 2.80 2.81 9
2 27-12-2008 Rasau -2.498 114.013 0.85 1.01 0.40 0.75 2.53 1.91 9
3 27-12-2008 Rasau -2.494 114.000 0.80 0.91 0.73 0.81 2.68 2.18 9
4 27-12-2008 Rasau -2.497 113.993 1.10 1.10 0.83 1.01 2.67 2.70 9
5 28-12-2008 Rasau -2.494 113.974 0.75 1.15 0.56 0.82 2.90 2.38 9
6 28-12-2008 Rasau -2.495 113.967 0.92 1.04 0.75 0.90 2.35 2.12 8
7 28-12-2008 Rasau -2.492 113.962 0.93 1.05 0.78 0.92 1.90 1.75 6
8 28-12-2008 Rasau -2.494 113.959 0.83 1.12 0.88 0.94 2.75 2.59 6
9 28-12-2008 Rasau -2.497 113.948 0.89 0.75 0.62 0.75 2.40 1.81 8
10 28-12-2008 Rasau -2.500 113.940 0.68 0.80 0.67 0.72 2.65 1.90 8
11 28-12-2008 Rasau -2.494 113.966 0.52 0.63 0.43 0.53 1.82 0.96 4
12 28-12-2008 Rasau -2.494 113.969 0.72 0.74 0.63 0.70 1.50 1.05 3
13 1-9-2009 Canal 21 -2.665 114.036 0.12 0.23 0.14 0.16 0.87 0.14 10
14 29-9-2009 Canal 21 -2.665 114.036 0.17 0.24 0.18 0.19 0.94 0.18 10
(s) t
(s) t
(s) t
(s) v
(m/s) k Q(m
1 12.79 12.21 14.34 13.11 0.69 0.85 1.639
2 51.87 67.17 56.43 58.49 0.15 0.85 0.249
3 123.51 158.34 132.75 138.20 0.07 0.85 0.121
4 65.76 60.18 54.24 60.06 0.15 0.85 0.343
5 54.98 68.00 50.42 57.80 0.16 0.85 0.315
6 26.26 22.21 25.38 24.62 0.32 0.85 0.586
7 23.52 24.63 23.07 23.74 0.25 0.85 0.376
8 29.92 29.10 25.63 28.22 0.21 0.85 0.469
9 26.89 31.46 44.12 34.16 0.23 0.85 0.360
10 22.54 24.88 36.18 27.87 0.29 0.85 0.463
11 24.07 31.71 24.66 26.81 0.15 0.85 0.122
12 13.33 10.69 11.68 11.90 0.25 0.85 0.224
13 53.00 78.00 53.00 61.33 0.16 0.85 0.0197
14 111.00 155.00 141.00 135.67 0.07 0.85 0.0114
Appendix 4 – Stage-Discharge Estimations
The most applicable equation for estimating channel flow was chosen based on the
logarithmic form of the stage-discharge data collected from the Sebangau National Park
(Appendix 2). The Chézy formula (Equation 1, Section 2.3) and Manning formula (shown
below) were tested against these data to find the most applicable. The Manning formula for
channel flow is as follows:
[Equation A 4]
where v is the corrected flow velocity, n is an empirical constant, R
is the hydraulic radius,
and S is the channel bed slope. Expressed in logarithmic form, Manning’s equation is:
+= n
[Equation A 5]
Plotting the stage-discharge data from the Sebangau National Park in terms of hydraulic
radius instead of surface water head (stage) allows for a linear fit to the logarithmic form of
either the Chézy or Manning equation. The linear regression of the logarithmic form of the
data, shown in Figure 8 gives a slope of 0.47, corresponding more closely to the logarithmic
form of the Chézy formula, whose slope is 0.5. The Chézy formula was therefore chosen as a
more appropriate approximation of channel flow for this study. The Chézy constant was
calculated with an assumed constant bed slope of 0.0006m/m (derived from the DEM in the
S.Rasau region) based on the logarithmic Chézy formula as follows:
10 0006.0
SCRv h
This value for the Chézy coefficient for small canals and was adjusted upwards for natural
rivers and the large SSI canal, based on different watercourse dimensions. Given the
uncertainty surrounding the Chézy coefficient, it was included among other surface water
parameters in a simple interface such that users can adjust these parameters as more
detailed data becomes available. For this study, stage-discharge curves were calculated for
all watercourses based on assumptions listed in Table A3-1 and the flooding considerations
described in Section 3.2.2 (shown in Figure 6), except where values were varied for analyzing
the response of the SIMGRO-MODFLOW model. Resulting stage-discharge curves are
shown in Figure A4- 1 and Figure A4- 2.
Chézy C
/s): Average Bed
Slope (m/m): Average
width (m) Max Channel
Depth (m)
sSebangau 20 0.0003 50 5
sSebangau West 15 0.0003 15 3
sBakung 15 0.0006 15 3
sRasau 15 0.0006 15 3
sBangah 15 0.0003 15 3
cSSI 15 0.0003 10 3
cBakung 10 0.0006 3 1
cRasau 10 0.0006 3 1
cBangah 10 0.0003 3 1
cSebangau 10 0.0003 3 1
Table A4- 1 – Parameters used for estimating stage-discharge curves for each watercourse.
0 1 2 3 4 5 6
Head (m)
Discharge (m^3/s)
S.Bakung / S.Rasau
S.Bangah / S.Paduran Alam
Figure A4- 1 Stage(head)-discharge curves for the rivers of the study area: S.Sebangau, S.Bakung,
S.Rasau, S.Bangah, and S.Paduran Alam.
0 1 2 3 4 5 6
Head (m)
Discharge (m^3/s)
SSI Canal
Small Canals (Bakung / Rasau)
Small Canals (Bangah / Sebangau)
Figure A4- 2 Stage(head)-discharge curves for the canals of the study area based on the rivers into
which they drain (except in the case of SSI): SSI canal, Bakung canals, Rasau canals, Bangah canals,
and Sebangau canals.
Appendix 5 – Carbon Calculations
The methodology for calculating CO
emissions used in this study uses average annual
groundwater depths in a linear equation to estimate annual carbon emissions, assuming that
no emissions occur when the groundwater depths are above zero (flooding conditions). The
problem with this method is that it likely neglects periods of the year where groundwater
depths drop below the surface, such that emissions still occur for part of the year. For
example, if an average groundwater level of zero is predicted by the SIMGRO-MODFLOW
model, emissions will be calculated at zero, which neglects those periods of the year where
the groundwater level was indeed below the soil surface.
For 1 SVAT cell, 2 methods of calculating carbon emissions over a 13 year period were
compared. In the first method (Method 1), emissions were calculated per day using the
emissions formula for daily groundwater levels and summing over each year. In the second
method (Method 2), emissions per year were calculated using the average groundwater level
over one year as described in section 2.6.
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Emissions (t/ha)
Annual (accumulated daily)
Annual (annual averages)
Figure A5- 1 – Estimated carbon emissions in tonnes per hectare (t/ha) for one SVAT cell over 13 years
as calculated by calculating emissions based on daily groundwater levels and aggregating them over
the year (red bars) and by calculating annual emissions from average annual groundwater levels (blue
During wet years (eg. 1999, 2005, 2007), Method 2 (annual average) under-predicts CO
emissions compared to Method 1 (daily aggregation), while during dry years the estimates are
virtually the same (Figure A5- 1). If the water table is below the soil surface for a longer period
of the year, the annual average seems to be more consistent with daily calculations. On the
other hand, years in which the water table is above the soil surface for longer periods lead to
larger deviations in carbon emission predictions between the two methods, as shown in
Figure A5- 2.
Difference between methods per year
0% 20% 40% 60% 80% 100%
% of year below soil surface
Difference: method 1 -
method 2
Figure A5- 2 – Difference in CO
emission estimates between method 1 (daily aggregation) and method
2 (annual average) as related to the percentage of each year that the groundwater level was below the
soil surface.
The two methods used in this analysis were compared using two statistical methods
commonly used for time series data: the normalized root mean squared error (nRMSE) and
the Nash-Sutcliffe efficiency (NSE). The nRMSE was calculated as described in Section 2.5.
The NSE was calculated according to the Equation A4.
( )
[Equation A 6]
An NSE of one indicates a perfect fit between the two sets of data, while an NSE of zero
indicates that the average of the original dataset (in this case, the 1
method) is as good a
predictor as the model (the approximation). Anything below zero indicates a poor fit. In this
analysis, an nRMSE of 6.1% and an NSE of 0.96 was found between the two methods, which
indicates that calculating annual CO
emissions from average groundwater depths per year is
an appropriate method for estimating carbon emissions per SVAT unit cell.
Full-text available
Tropical forests cover a significant portion of the earth's surface and provide a range of ecosystem services, but are under increasing threat due to human activities. Deforestation and forest degradation in the tropics are responsible for a large share of global CO2 emissions. As a result, there has been increased attention and effort invested in the reduction of emission from deforestation and degradation and the protection of remaining tropical forests in recent years. Methods for tropical forest monitoring are therefore vital to track progress on these goals. Two data streams in particular have the potential to play an important role in forest monitoring systems. First, satellite remote sensing is recognized as a vital technology in supporting the monitoring of tropical forests, of which the Landsat family of satellite sensors has emerged as one of the most important. Owing to its open data policy, a large range of methods using dense Landsat time series have been developed recently which have the potential to greatly enhance forest monitoring in the tropics. Second, community-based monitoring is supported in many developing countries as a way to engage forest communities and lower costs of monitoring activities. The development of operational monitoring systems will need to consider how these data streams can be integrated for the effective monitoring of forest dynamics. This thesis is concerned with the monitoring of tropical forest dynamics using a combination of dense Landsat time series and community-based monitoring data. The added value conferred by these data streams in monitoring deforestation, degradation and regrowth in tropical forests is assessed. This goal is approached from two directions. First, the application of econometric structural change monitoring methods to Landsat time series is explored and the efficacy and accuracy of these methods over several tropical forest sites is tested. Second, the integration of community-based monitoring data with Landsat time series is explored in an operational setting. Using local expert monitoring data, the reliability and consistency of these data against very high resolution optical imagery are assessed. A bottom-up approach to characterize forest change in high thematic detail using a priori community-based observations is then developed based on these findings. Chapter 2 presents a robust data-driven approach to detect small-scale forest disturbances driven by small-holder agriculture in a montane forest in southwestern Ethiopia. The Breaks For Additive Season and Trend Monitoring (BFAST Monitor) method is applied to Landsat NDVI time series using sequentially defined one-year monitoring periods. In addition to time series breakpoints, the median magnitude of residuals (expected versus observed observations) is used to characterize change. Overall disturbances are mapped with producer's and user's accuracies of 73%. Using ordinal logistic regression (OLR) models, the extent to which degradation and deforestation can be separately mapped is explored. The OLR models fail to distinguish between deforestation and degradation, however, owing to the subtle and diffuse nature of forest degradation processes. Chapter 3 expands upon the approach presented in Chapter 2 by tracking post-disturbance forest regrowth in a lowland tropical forest in southeastern Peru using Landsat Normalized Difference Moisture Index (NDMI) time series. Disturbance between 1999 and 2013 are mapped using the same sequential monitoring method as in Chapter 2. Pixels where disturbances are detected are then monitored for follow-up regrowth using the reverse of the method employed in Chapter 2. The time of regrowth onset is recorded based on a comparison to defined stable history period. Disturbances are mapped with 91% accuracy, while post-disturbance regrowth is mapped with a total accuracy of 61% for disturbances before 2006. Chapter 4 and 5 explore the integration of community-based forest monitoring data and remote sensing data streams. Major advantages conferred by community-based forest disturbance observations include the ability to report on drivers and other thematic details of forest change and the ability to detect low-level forest degradation before these changes are visible above the forest canopy. Chapter 5 builds on these findings and presents a novel bottom-up approach to characterize forest changes using local expert disturbance reports to calibrate and validate forest change models based on Landsat time series. Using random forests and a selection of Landsat spectral and temporal metrics, models describing forest state variables (deforested, degraded or stable) at a given time are produced. As local expert data are continually acquired, the ability of these models to predict forest degradation are shown to improve. Chapter 6 summarizes the main findings of the thesis and provides a future outlook, given the prospect of increasing availability of satellite and in situ data for tropical forest monitoring. This chapter argues that forest change methods should strive to utilize satellite time series and ground data to their maximum potential. As "big data" emerges in the field of earth observation, new data streams need to be accommodated in monitoring methods. Operational forest monitoring systems that are able to integrate such diverse data streams can support broader forest monitoring goals such as quantitative monitoring of forest dynamics. Even with a wealth of time series based forest disturbance methods developed recently, forest monitoring systems require locally calibrated forest change estimates with higher spatial, temporal and thematic resolution to support a variety of forest monitoring objectives.
Full-text available
In the recent past, rapid destruction of sensitive tropical peatland ecosystems has necessitated the formulation of comprehensive strategies aiming towards restoration and sustainable environmental management. In cooperation with the World Wide Fund for Nature of Indonesia (WWF-Indonesia), research was undertaken in the Sebangau National Park in the province of Central Kalimantan, Indonesia, to assess the efficacy of hydrological restoration of the Sebangau peat swamp forest. In this study, dams were constructed across drainage canals in the Sebangau National Park, and groundwater and surface water levels in monitoring transects near the dams were monitored on a monthly basis and analyzed using ArcGIS9.3 and a radar satellite-derived Digital Elevation Model (DEM). Groundwater dynamics showed that artificial drainage by constructed canals disrupts regional subsurface drainage patterns by increasing drawdown on a local scale, while regional flows remain intact. The scale of this drawdown was found to be dependent on the direction of regional subsurface flows, suggesting that the extent of drainage by artificial canals is orientation- dependent. Dams constructed by WWF-Indonesia were found to function by reducing local drawdown, retarding subsurface drainage during the dry season, and retaining floodwater during the wet season. Water retention and associated rewetting was found not only to be dependent on local precipitation, but was also found to progressively increase over time despite varying levels of rainfall, suggesting a role for deposited organic debris in dam performance. The implications of these results on peatland ecosystem health and sustainable management, and the potential for hydrological modelling in the Sebangau peatlands are also discussed in this report.
Full-text available
Peatlands respond to natural hydrologic cycles of precipitation and evapotranspiration with reversible deformations due to variations of water content in both the unsaturated and saturated zone. This phenomenon results in short-term vertical displacements of the soil surface that superimpose to the irreversible long-term subsidence naturally occurring in drained cropped peatlands because of bio-oxidation of the organic matter. These processes cause changes in the peat structure, in particular, soil density and void ratio. The consequential changes in the hydrological parameters need to be incorporated in water flow dynamical models. In this paper, we present a new constitutive relationship for the soil shrinkage characteristic (SSC) in peats by describing the variation of porosity with moisture content. This model, based on simple physical considerations, is valid for both anisotropic and isotropic three-dimensional peat deformations. The capability of the proposed SSC to accurately describe the deformation dynamics has been assessed by comparison against a set of laboratory experimental results recently published. The constitutive relationship has been implemented into a Richards' equation–based numerical code and applied for the simulation of the peat soil dynamics as observed in a peatland south of the Venice Lagoon, Italy, in an ad hoc field experiment where the relevant parameters are continuously measured. The modeling results match well a large set of field data encompassing a period of more than 50 days and demonstrate that the proposed approach allows for a reliable reproduction of the soil vertical displacement dynamics as well as the hydrological behavior in terms of, for example, water flow, moisture content, and suction.
Full-text available
Extensive degradation of Indonesian peatlands by deforestation, drainage and recurrent fires causes release of huge amounts of peat soil carbon to the atmosphere. Construction of drainage canals is associated with conversion to other land uses, especially plantations of oil palm and pulpwood trees, and with widespread illegal logging to facilitate timber transport. A lowering of the groundwater level leads to an increase in oxidation and subsidence of peat. Therefore, the groundwater level is the main control on carbon dioxide emissions from peatlands. Restoring the peatland hydrology is the only way to prevent peat oxidation and mitigate CO2 emissions. In this study we present a strategy for improved planning of rewetting measures by dam constructions. The study area is a vast peatland with limited accessibility in Central Kalimantan, Indonesia. Field inventory and remote sensing data are used to generate a detailed 3D model of the peat dome and a hydrological model predicts the rise in groundwater levels once dams have been constructed. Successful rewetting of a 590 km² large area of drained peat swamp forest could result in mitigated emissions of 1.4–1.6 Mt CO2 yearly. This equates to 6% of the carbon dioxide emissions by civil aviation in the European Union in 2006 and can be achieved with relatively small efforts and at low costs. The proposed methodology allows a detailed planning of hydrological restoration of peatlands with interesting impacts on carbon trading for the voluntary carbon market.
Accurate inventory of tropical peatland is important in order to (a) determine the magnitude of the carbon pool; (b) estimate the scale of transfers of peat-derived greenhouse gases to the atmosphere resulting from land use change; and (c) support carbon emissions reduction policies. We review available information on tropical peatland area and thickness and calculate peat volume and carbon content in order to determine their best estimates and ranges of variation. Our best estimate of tropical peatland area is 441 025 km 2 ($11% of global peatland area) of which 247 778 km 2 (56%) is in Southeast Asia. We estimate the volume of tropical peat to be 1758 Gm 3 ($ 18–25% of global peat volume) with 1359 Gm 3 in Southeast Asia (77% of all tropical peat). This new assessment reveals a larger tropical peatland carbon pool than previous estimates, with a best estimate of 88.6 Gt (range 81.7–91.9 Gt) equal to 15–19% of the global peat carbon pool. Of this, 68.5 Gt (77%) is in Southeast Asia, equal to 11–14% of global peat carbon. A single country, Indonesia, has the largest share of tropical peat carbon (57.4 Gt, 65%), followed by Malaysia (9.1 Gt, 10%). These data are used to provide revised estimates for Indonesian and Malaysian forest soil carbon pools of 77 and 15 Gt, respectively, and total forest carbon pools (biomass plus soil) of 97 and 19 Gt. Peat carbon contributes 60% to the total forest soil carbon pool in Malaysia and 74% in Indonesia. These results emphasize the prominent global and regional roles played by the tropical peat carbon pool and the importance of including this pool in national and regional assessments of terrestrial carbon stocks and the prediction of peat-derived greenhouse gas emissions.
Most schemes in common use for field and laboratory classification of peats were developed in boreal and humid temperate regions and do not recognize the distinctive features and specific uses of tropical peats, such as those of the Tasek Bera Basin in tropical Peninsular Malaysia. The important aspects of peat texture (morphology of constituents and their arrangement) and laboratory ash content (residue after ignition) need modification to be valuable for classifying these and other tropical peat deposits. In the Tasek Bera Basin, most of the deposits would not be considered as peat according to some classification schemes, even though most have C contents >25%. We propose a new three-group (fibric, hemic, sapric) field texture classification applicable to tropical organic deposits, which is similar to the system of the US Soil Taxonomy. The classification is based on the following factors: (1) visual examination of the morphology of the peat constituents (texture); and (2) estimates of fiber content and matrix (finest fraction of peat consisting of highly humified organic matter and inorganic material). The classification is applicable to all organic deposits with 35% loss on ignition). We also present a new laboratory classification of organic soils based on ash and C content. The US Soil Taxonomy classifies organic soils as having more than 12–18% organic C, depending on clay content. Ash content and these limits for organic soils allow the discrimination of four main groups: peat, muck, organic-rich soil/sediment and mineral soil/sediment. Peat is defined as having an ash content of 0–55%, muck 55–65%, organic-rich soil/sediment 65–80% and mineral soil/sediment 80–100%. The peat class is further subdivided into very low ash (0–5%), low ash (5–15%), medium ash (15–25%), high ash (25–40%) and very high ash (40–55%) subclasses.
A limitation of existing models of water and solute movement in fen peats is that they fail to represent processes in the unsaturated zone. This limitation is largely due to a lack of data on the hydraulic properties of unsaturated peat, in particular the relationship between hydraulic conductivity (K) and pressure head (). A tension infiltrometer was used to measure K() of a fen peat in Somerset, England. It was found that macropores could be important in water and solute movement in this soil type. It was also found that (i) variability of K in this peat was less than that reported for other peats and mineral soils, and (ii) the K data were better described by a log-normal distribution than a normal distribution in accord with findings from other peat and mineral soils. Recommendations on improving the understanding of water and solute movement in the unsaturated zone of this soil type are made.
The lowland peatlands of south-east Asia represent an immense reservoir of fossil carbon and are reportedly responsible for 30% of the global carbon dioxide (CO2) emissions from Land Use, Land Use Change and Forestry. This paper provides a review and meta-analysis of available literature on greenhouse gas fluxes from tropical peat soils in south-east Asia. As in other parts of the world, water level is the main control on greenhouse gas fluxes from south-east Asian peat soils. Based on subsidence data we calculate emissions of at least 900 g CO2 m−2 a−1 (∼250 g C m−2 a−1) for each 10 cm of additional drainage depth. This is a conservative estimate as the role of oxidation in subsidence and the increased bulk density of the uppermost drained peat layers are yet insufficiently quantified. The majority of published CO2 flux measurements from south-east Asian peat soils concerns undifferentiated respiration at floor level, providing inadequate insight on the peat carbon balance. In contrast to previous assumptions, regular peat oxidation after drainage might contribute more to the regional long-term annual CO2 emissions than peat fires. Methane fluxes are negligible at low water levels and amount to up to 3 mg CH4 m−2 h−1 at high water levels, which is low compared with emissions from boreal and temperate peatlands. The latter emissions may be exceeded by fluxes from rice paddies on tropical peat soil, however. N2O fluxes are erratic with extremely high values upon application of fertilizer to wet peat soils. Current data on CO2 and CH4 fluxes indicate that peatland rewetting in south-east Asia will lead to substantial reductions of net greenhouse gas emissions. There is, however, an urgent need for further quantitative research on carbon exchange to support the development of consistent policies for climate change mitigation.
A tropical ombrotrophic peatland ecosystem is one of the largest terrestrial carbon stores. Flux rates of carbon dioxide (CO2) and methane (CH4) were studied at various peat water table depths in a mixed-type peat swamp forest floor in Central Kalimantan, Indonesia. Temporary gas fluxes on microtopographically differing hummock and hollow peat surfaces were combined with peat water table data to produce annual cumulative flux estimates. Hummocks formed mainly from living and dead tree roots and decaying debris maintained a relatively steady CO2 emission rate regardless of the water table position in peat. In nearly vegetation-free hollows, CO2 emission rates were progressively smaller as the water table rose towards the peat surface. Methane emissions from the peat surface remained small and were detected only in water-saturated peat. By applying long-term peat water table data, annual gas emissions from the peat swamp forest floor were estimated to be 3493±316 g CO2 m−2 and less than 1.36±0.57 g CH4 m−2. On the basis of the carbon emitted, CO2 is clearly a more important greenhouse gas than CH4. CO2 emissions from peat are the highest during the dry season, when the oxic peat layer is at its thickest because of water table lowering.
Indonesia has responded to worldwide concern about climate change by committing to significant carbon emissions reductions over the next decade. Achievement of this goal will face significant practical obstacles, and the opportunity cost of avoided deforestation is considerable. In our second Policy Dialogue, two climate change experts present contributions to the debate. This exchange could not be more timely, given Norway's recent offer of substantial development assistance to Indonesia in return for reductions in deforestation and forest degradation. In this first contribution, Adam Schwarz canvasses abatement options and outlines barriers to abatement and its measurement. These include capacity and data constraints, obstacles posed by decentralisation, and problems in identifying and measuring the costs of abatement measures. The author urges Indonesia to embrace the opportunity a lower carbon growth trajectory presents, and argues that this will require substantial additional funding, including from developed countries, and sustained leadership. (Ed.)