Drainage of organic soils and GHG emissions: Validation with country data

Drainage of large areas with organic soils was conducted over the past century to free land for agriculture. A significant acceleration of such trends was observed in recent decades in South-East Asia, largely driven by drainage of tropical peatlands, an important category of organic soils, for cultivation of oil palm. This work presents methods and main results of a new methodology developed for FAOSTAT, whereby the overlay of dynamic maps of land cover and the use of information 10 on histosols allows the production of a global annual dataset of drained area and emissions over a time series, covering the period 1990–2019. This is an improvement over the existing FAO approach, which had produced only a static map of drained organic soils for the year 2000. Results indicate that drained area and emissions increased by 13 percent globally since 1990, reaching in 2019 24 million ha of drained organic soils, with world total emissions of 830 million tonnes of carbon dioxide (CO2) equivalent. Of these totals, the largest contribution was from the drainage of tropical peatlands in South-East Asia, 15 generating nearly half of global emissions. Results were validated against national data reported by countries to the UN Climate Convention and to well-established literature. Overall, the validation yielded a good agreement with these sources. FAOSTAT estimates explained about 60 percent of the variability in official country reported data. The predicted emissions were virtually identical – with over 90 percent of explained variability – to official data from Indonesia, currently the top emitting country by drained organic soils. Also, calculated emissions factors for oil palm plantations in Indonesia and Malaysia were in the 20 same range and very close to emissions factors derived from detailed field measurements. This validation suggests that the FAO estimates may be a useful and sound reference in support of countries reporting needs. Data are made available as open access via the Zenodo portal (Tubiello and Conchedda, 2020) with DOI 10.5281/zenodo.3942370.

typically last for several decades after the drainage event, due to the large quantities of organic substrate available. Agriculture is a major cause of drainage of organic soils around the world, and especially since 1990 due to the cultivation of permanent crops such as oil palm. Restoration of degraded organic soils is currently a priority in several countries as part of their greenhouse gas mitigation and ecosystem restoration commitments under the UN climate convention (Leifeld and Menichetti, 2018;Tiemeyer et al., 2020). Measuring current trends, globally and with country detail, is therefore important to identify and 35 quantify existing and fast-developing new hotspots of degradation and to help reduce emissions from drained organic soils in future decades. Estimates of drainage area and greenhouse gas (GHG) emissions from organic soils for the year 2000 were developed by FAO and used by the Intergovernmental Panel on Climate Change (IPCC) for global analysis (Tubiello et al., 2016;Smith et al., 2014). That preliminary work was based on the geospatial overlay of two static maps, one for land cover, indicating presence of agriculture, and one for soil characteristics, indicating presence of organic soils, through the use of 40 histosols as proxy. This paper describes additional methodological developments made possible by the availability of time dependent land cover maps, resulting in the production, for the first time, of estimates over a complete time series .

Material and Methods
Organic soils are characterized by high concentrations of organic matter. They mostly develop under poorly drained, wetland 45 conditions and are found at all altitudes, with the vast majority occurring in lowlands (Rieley and Page, 2016). Peatlands are an important type of organic soils (Page et al., 2011;IPCC, 2014a). According to IPCC (2006), organic soils can be largely identified with the histosols group of the FAO-UNESCO classification. FAO and Wetlands International (2012) indeed described histosols as soils that develop in (predominantly) "moss peat in boreal, arctic and subarctic regions, via moss peat, reeds/sedge peat and forest peat in temperate regions to mangrove peat and swamp forest peat in the humid tropics". Common 50 names for histosols are `peat soils', `muck soils', `bog soils' and `organic soils' (FAO et al., 1998). In this work, we follow IPCC guidelines and identify organic soils with histosols. Cropland and grassland organic soils are drained permanently or semi-permanently, as well as regularly limed and fertilized, to permit annual or permanent crop cultivation, including tree plantations, or to support livestock grazing. For these reasons, the presence of cultivated crops or grazing animals on organic soils (Noble et al., 2018) -which have otherwise low productivity-may be associated with drainage, causing N and C losses 55 the drainage for agriculture. Drainage stimulates the oxidation of organic matter previously built up under a largely anoxic environment. The rates of emissions are influenced by climate, with warmer climates accelerating the processes of oxidation 95 of soil organic matter hence causing higher emissions than in temperate and cooler climates. The emission factors by gas thus are climate-dependent. In this methodology, we spatialized the relevant IPCC emission factors following a global map of climatic zones (JRC, 2010) to produce global maps EF for the two gases. The pixel-computations then multiply the area of drained and managed organic soils from Eq. (2) and Eq. (3) above by global maps of emission factors to derive estimates of annual N2O and CO2 emissions by pixel as summarized in Eq. 1. 100 As described in Tubiello et al., (2016), the approach is based on reclassification tables to extract the proportions of cultivated and grassland area from the yearly land cover maps. When all input layers overlap, the underlying assumption is that of an equal likelihood within each pixel to find cultivated (or grassland) area and organic soils. Operationally, the methodology multiplies the area of organic soils in the pixel by the area of the pixel that is cropped or has grassland cover. In this way, we derived by pixel the area of organic soils that is drained for agricultural activities. Organic soils must be indeed be drained to 105 allow for crop cultivation activities. In the case of grassland, livestock grazing beyond the carrying capacities of organic soils leads instead to peat degradation and drainage. The following sections provide more details about the information necessary to implement the computations above.

Soils
Information on the geographical distribution of histosols, for use in the term WSM of Eq. (2) and Eq. (3) above, was derived 110 from the Harmonized World Soil Database (HWSD v 1.2), a raster dataset with a nominal resolution of 30 arc second on the ground (corresponding approximately to 1 x 1km at the equator) published in 2012 by FAO and the International Institute for Applied Systems Analysis (IIASA). The HWSD compiles more than 40 years of soil information from several sources worldwide, re-classified and harmonized according to the FAO-UNESCO classification. The standardized structure of the HSWD v 1.2 allows displaying and querying the composition in terms of soil units and of soil parameters such as the organic 115 carbon content, the pH, or the water storage capacity. The HSWD dataset was queried to extract values representing the percentage of the pixel area that contains histosols, as either dominant or secondary soil type (Fig. 1).

Land cover and land use
Information on the area extent of IPCC categories cropland and grassland for use in terms LUcropland and LUgrassland in Eq. (2)-120 (3) was taken from the land cover maps produced by the Catholic University of Louvain (UCL) Geomatics (UCL Geomatics, 2017), produced under the Climate Change Initiative of the European Spatial Agency (ESA CCI, 2020) and hereinafter referred to as CCI LC maps. The CCI LC maps were first released in April 2017 as 24 global annual and consistent land cover maps covering the period 1992 to 2015 (UCLouvain Geomatics, 2017). At the end of 2019 and in the framework of the European https://doi.org/10.5194/essd-2020-202 2017 and 2018 that are consistent with earlier maps (Fig. 2).
The long-term consistency of this dataset, yearly updates and high thematic detail on a global scale make it uniquely suitable to observe and assess changes in area drained and GHG emissions from organic soils. The CCI LC maps contain information for 22 global land cover classes, based on the FAO Land Cover Classification Systems (Di Gregorio, 2005), with a spatial 130 resolution of approximately 300m.
The land cover maps  were used to assign to each pixel the proportion of its area under relevant land cover categories. This information was combined to provide proxy information on the proportion of pixel area under land cover / land use classes cropland and grassland (Tables 1 and 2).

Livestock
Information on the spatial distribution of livestock for use in estimating the term LDR in Eq. (3) above, was taken from the Gridded Livestock of the World (GLW) (Robinson et al., 2014), providing geospatial data on the density of three ruminants species: cattle, sheep and goats (Fig. 3). Animal numbers by pixel were first converted in livestock units (LSU) (FAO, 2011), and pixels with values higher than 0.1 (Critchley et al., 2008;Worrall and Clay, 2012) selected for use in Eq. (3). 140

Climatic Zones and emission factors
As discussed above, pixel-level climatic information for use in terms EFijk in Eq. (1), was derived from a map of climatic zones. The Joint Research Centre (JRC) of the European Commission, developed this spatial layer in line with IPCC specifications based on latitude and elevation of each pixel ( Figure 4). Default IPCC emissions factor by land use and gas (Table 3) were then assigned by pixel to each climatic zone and three 145 additional geospatial layers were produced to cover possible combinations of EFs ( Figure 5). As one country may encompass more than one climatic zones, when emissions are aggregated at national level, the resulting emissions factors represent weighted averages of the various EFs assigned at pixel level. In computations, CO2-C  (FAO, 2020c) and Grassland (FAO, 2020d) are disseminated within the FAOSTAT Emissions-Land use domain. As part of ongoing efforts to provide users with reliable and transparent data, the complete spatial dataset that underlies FAOSTAT statistics will be also disseminated through FAO new maps catalog (FAO, 2020e). Under the dataset, Cultivation of Organic soils, N2O emissions are also disseminated in CO2eq by applying three 160 different sets of Global Warming Potential (GWP) coefficients (100-year time horizon) from the IPCC assessment reports: a) IPCC Second Assessment Report (IPCC, 1996); b) IPCC Fourth Assessement Report (IPCC, 2007); and c) IPCC Fifth Assessment Report (IPCC, 2014b). All data are also available at Zenodo as open access (Tubiello and Conchedda, 2020) with DOI 10.5281/zenodo.3942370. They and can be downloaded at https://zenodo.org/record/3942370#.XxWJjygzbIU.

Limitations and uncertainty 165
Previous work had estimated the uncertainty of our estimates at ±40% for the area information and an uncertainty range (−14%, +166%) for the emission estimates. These uncertainties, valid at pixel level, were assumed to also characterize the nationallyaggregated values (Tubiello et al., 2016). Furthermore, the new methodology developed herein may result in some cases in reduction in the drained area during the 30 years of the analysis (see Appendix A, Table A1). In such cases, the pixel-level proportions that are applied to identify the cropland and grassland cover, have detected corresponding changes in land cover. 170 The scale of the analysis prevents however to understand whether these changes actually happened in the area of organic soils thus resulting in a rewetting of the drained peats or are instead an artefact of the spatial methods. The IPCC Wetlands Supplement introduced already in 2014 additional methodological guidance, with a specific focus on the rewetting and restoration of peatland that was not included in the 2006 Guidelines (IPCC, 2014a). Limited country-specific activity data on rewetting prevented however implementing the Supplement refined methods in the FAOSTAT dataset. 175

Main results: Global trends
In 2019, nearly 25 Mha or about 7.5 percent of the 328 Mha of worldwide histosols had been drained for agriculture with a limited increase since 1990. Data suggest that the largest extent of organic soils in Northern America and Eastern Europe have undergone little changes during the past decades likely because these peats have been drained for agriculture already for many centuries (Joosten and Clarke, 2002). The drainage of organic soils is instead a more recent phenomenon in South East Asia. 180 In this region, the drained area grew by 5 percent points since 1990 and in 2019 more than 26 percent of the original organic soils were already drained. Asia is on average the region with the highest share of drained histosols (30 percent) while, at subregional level Western Europe had over two thirds of its organic soils that were drained already in 1990. In 2019, among countries where the area of histosols is above 1Mha (see Annex A, Table A1 In 2019, Indonesia had the largest area of drained organic soils (newly 5Mha), followed by the Russian Federation (about 1.9 Mha) and the United States of America (nearly 1.6 Mha). Among these top ten countries, Indonesia and Malaysia also registered the largest relative increases in area drained since 1990 (+5 and +10 percent for Indonesia and Malaysia, respectively). 190 Global GHG emissions from drained organic soils were 833 Mt CO2eq. In 2019, emissions were 13 percent higher when compared to 1990 and 10 percent higher when compared to 2000 ( Figure 6). This value represented almost 8 percent of total agriculture and related land use emissions.
In 2019, CO2 and N2O gas contributed 87 percent and 13 percent of global emissions. Grassland organic soils were responsible for about 10 percent of all emissions while the vast majority was due to the drainage for cropping. These relative 195 contributions have changed little since 1990 (Appendix A, Table A2).
In 2019, three-fourths of the global emissions from organic soils were from only 11 countries (Figure 7), Malaysia and Indonesia together were responsible for nearly half (47 percent) of total emissions.

Results: Data Validation
The FAOSTAT estimates of the extent of organic soils, which are used as input to Eq. (2)-(3), were compared to published 200 data at country, regional and global level. Resulting emissions and emissions factors for oil palm plantations are also included to validate FAOSTAT results.

Area of organic soils and peatlands
Comparison of the extent of drained organic soils is hindered by a number of factors, including the fact that the FAOSTAT data refers to area of organic soils, while a majority of published studies has focused on area of peatlands. The FAOSTAT 205 global estimates of 3.3 million square kilometers (Mkm 2 ) of organic soils (histosols) were 25 percent smaller than the published range of 4.0-4.3 Mkm 2 of peat soils. This is consistent with statements by Xu et al. (2018), who highlighted that histosols tend to underestimate areas in tropical swamp-forested peatlands. At regional level, FAOSTAT data agreed well with the most recent estimates of Xu et al. (2018) and mean estimate from Immirzi et al. (Table 5). In addition, while acknowledging the large differences existing between published estimates by regions, FAOSTAT estimates remained consistently within the 210 observed ranges More specifically, FAOSTAT estimates of area of organic soils for North America was 1.3 Mkm 2 vs. a We continued the validation analysis by comparing FAOSTAT estimates to published data for about 60 tropical countries, compiled from the widely recognized meta-analysis of Page et al. (2011), and for the same set of countries to values computed using a recent map of tropical peat distribution (Gumbricht et al., 2017) Table B1). In 2017, Gumbricht and associates published new estimates of wetland and peatland areas, depths and volumes. The expert system approach is based on three biophysical indices related to wetland and peat formation: (1) long-term water supply exceeding atmospheric water 220 demand; (2) annually or seasonally water-logged soils; and (3)  The use of observed or estimated data is hampered by the wide uncertainties that still exist in defining, mapping and measuring actual extent of peatland throughout the world. To date, no globally accepted definition of peatlands exists. To this end, ongoing international efforts such as the Global Peatlands Initiative (2020) are expected to improve and consolidate current knowledge.

Validation with country data reported to the Climate Convention of the United Nations
The FAOSTAT data uses Eq. (2)-(3) above to overlay information on organic soils extent with information on land use and 245 other geospatial characteristics, to estimate the drainage area of organic soils due to agriculture (Tubiello et al., 2016). These were in turn used as input to estimate resulting GHG emissions. We used data reported by countries to the UN Framework Convention on Climate Change (UNFCCC) for validation of these FAOSTAT estimates. We looked both at data from the National Greenhouse Gas Inventories of the Annex I Parties and to the national communication from Indonesia, a top emitter country. 250

Annex I parties
UNFCCC data were available for thirty-eight countries belonging to the Annex I parties to the climate Convention. These represent developed countries, mostly located in temperate and boreal zones of the world. First, we compared data on the area drained (activity data), which allowed to test assumptions underlying the use of Eq. (2)-(3) above. FAOSTAT country-level estimates were in good agreement with those officially reported by countries to the UNFCCC (R 2 =0.57) of area drained of 255 organic soils (Fig. 10). At regional level, FAOSTAT predicted a total of about 14 Mha of drained organic soils for Annex I parties, versus country reported figures of nearly 12 Mha for the last inventory in 2017 (Appendix B, Table B2). On the one hand, estimates in several countries with significant contributions were well in line with national reporting, including the Canada (1.3 vs 0.2 Mha). In these latter cases, differences have however opposite directions. FAOSTAT estimates were much larger than country reported data in Canada but smaller in the Russian Federation.
For the same set of UNFCCC countries as above, we also compared N2O emissions, which are reported by countries under the IPCC sector Agriculture. C fluxes from the drainage of organic soils are reported by Annex I countries under Land Use, Land 265 Use Change, and Forestry (LULUCF) categories. In the inventories, relevant reporting categories are 4.B.1 "Cropland Remaining Cropland"; 4.B.2 "Land Converted to Cropland" and 4.C.1 "Grassland Remaining Grassland"; and 4.C.2 "Land Converted to Grassland". Data for carbon are much sparser than for N2O emissions possibly due also to complexity in reporting (Berthelmes et al., 2015). Beside the differences in activity data (area drained) that were observed earlier, differences may also due to countries applying Tiers higher than the default methodology applied in FAOSTAT as well as to the definition of types 270 of land use causing drainage.
As for the area drained, FAOSTAT emissions estimates were also in good agreement with data officially reported to the UNFCCC (R 2 =0.553) (Fig.11), but with FAOSTAT consistently overestimating country data. At regional level, FAOSTAT predicted total emissions of 184 kt N2O for Annex I parties, versus country reported figures of 143 kt N2O (Appendix B, Table   B2). Estimates of annual emissions in several countries with significant contributions were well in line with national reporting, 275 including the United States of America (20 vs 27 kt N2O); Belarus (19 vs 18 kt N2O); Germany (14 vs 10 kt N2O); and Ukraine CO2eq, about three-fold FAOSTAT results (14 Mt Mt CO2eq). This suggests, that even FAOSTAT estimates may not fully grasp the potential for mitigation from the rewetting of drained organic soils. 285

Non-Annex I parties
Over fourty percent of the global emissions from the agricultural drainage of organic soils is generated in Indonesia and Malaysia. In addition, these two countries have contributed the most to emissions increases since 1990 (FAO, 2020f) (Fig. 12).
We compared FAOSTAT estimates of GHG emissions from the drainage of organic soils in Indonesia to those reported by  (Fig. 13). As FAOSTAT only includes drainage for agriculture, part of the differences may be due to the types of land use for which the BUR reports drainage and emissions, as the BUR possibly includes drainage under forestry. 300

Emissions factors for palm oil plantations
The establishment of new oil palm plantations is recognized as main driver for the drainage of tropical peatlands in Indonesia and Malaysia (Hojer et., 2010;Hooijer et al., 2012;Miettinen et al., 2012;Dohong et al., 2018;Cooper et al., 2020;FAO, 2020f  Mha respectively for the two countries. Of these, based on our analysis of the crop drained soils and of the plantations map, about 9 percent (Indonesia) and 4 percent (Malaysia) were located in drained organic soils (Table 6). Together, the tree plantations mapped by Petersen and colleagues were responsible for almost half of the 2014 emissions from cropland organic 315 soils in Indonesia.
In Malaysia, the relative contribution of oil palm plantations to the area drained for cultivation was even as nearly half of total area of crop organic soils and emissions in this country was due to oil palm plantations. Other types of plantations contributed an additional 10 percent to area drained and emissions, which suggests the contribution of annual and temporary crops to peat conversion in this country may less important than in Indonesia. 320 The EFs for oil palm plantations derived from the analysis was around 78 CO2eq ha -1 yr -1 in the two countries, and in close agreement with published estimates (Table 10). Available literature is largely based on direct measurements and typically analyses the influence of the depth of drainage, soil subsidence rates, soil moisture and the period since the initial drainage and establishment of the oil palm plantations. Corresponding values range from minimum average losses of 13 t CO2e ha -1 yr -1 as in Hashim et al., (2018) to a maximum value of 117 t CO2e ha -1 yr -1 as in Matysek et al. (2018) and recent study by Cooper 325 et al. (2020). FAOSTAT estimated EF is therefore very close to the average value from the selected studies (73 t CO2eq ha -1 yr -1 ). This additional validation confirms that our methodology is compatible with most relevant and well-established estimates of a major source of emissions from drained organic soils in South East Asia and suggests that FAOSTAT estimates may be equally applied to other tropical countries.

Conclusions 330
Organic soils are a rich carbon pool and their drainage for agriculture has important impacts on the global carbon cycle.
FAOSTAT statistics on greenhouse gas emissions relative to the drainage of organic soils were updated for the period 1990-2019 based on geospatial computation and pixel-level application of default Tier I method of the Intergovernmental Panel on Climate Change (IPCC). In line with country reporting requirements to the Climate Convention, and following the IPCC, statistics are disseminated by gas (N2O and CO2) and land use classes, cropland and grassland in three separate FAOSTAT 335 domains. These FAOSTAT statistics represent the only avaialble global dataset in the world today showing country, regional and global time series on drained organic soils. Efforts are also in progress to disseminate publicly the underlying spatial data.
In 2019, FAOSTAT estimated that nearly 25 million ha of organic soils were drained from agriculture and were responsible for 833 million tonnes of CO2eq. This was about 8 percent of total agriculture and related land use emissions in that year.
About half of the greenhouse gas emissions was due to the drainage of organic soils in South Eastern Asia and particularly 340 Indonesia and Malaysia.
We  Table 1. Proportion of area of relevant CCI-LC pixels corresponding to land use cropland

Cropland Grassland
Range