Satellite LiDAR reveals patterns in water table depth and flood risk in SE Asia coastal lowland and peatland plantations

Agriculture in SE Asian coastal lowlands is vulnerable to inundation and associated crop production loss. Quantifying and managing flood risk, surface subsidence, and carbon emissions (in peatlands) requires an understanding of water table depth patterns and controls, for which no comprehensive data currently exists. Here, we present a dataset of 15,575 measurements of canal water table depth below the soil surface (CWD) in large-scale industrial tree plantations across the islands of Sumatra and Borneo, derived from ICESat-2 satellite LiDAR data. A pronounced relation is evident between CWD and land surface elevation, which applies equally in different regions, both in peatlands and mineral soil areas, and in oil palm as well as Acacia plantations. Mean CWD varies consistently across elevation zones, from −0.6 to −0.85 m for oil palm and Acacia plantations, respectively (i.e., ∼-0.5 – ∼-0.75 m in peat soil) where the surface is 5–9 m above global mean sea level, to ∼ -0.2 m at 1–3 m elevation near mean sea level. The area of industrial plantations in the 1–3 m elevation range alone, which likely has the highest flood risk, is 984,700 ha. Accounting for larger areas of smallholder plantations, and for areas at higher elevations with impeded drainage, the total area of lowland plantations likely experiencing regular flooding could exceed 2.8 million hectares. The close relation with elevation indicates that plantation water levels, especially in lowest-lying areas, are often controlled by landscape drainability conditions rather than management targets, and that optimum water table depths for production and peat loss prevention are not achieved over large areas. We propose that such difficult-to-manage landscape elevation controls on water table depth are considered in yield projections and land use planning, as well as fire and peat carbon emission reduction policies. Unproductive flood-prone plantations as well as peatlands with very low water tables may be logical priority targets for restoration planning.

SE Asian coastal lowlands, often consisting of peatlands separated by mineral soil areas along rivers, are vulnerable to inundation and production loss as a consequence of land surface subsidence and sea level rise. In drained peatlands especially, subsidence occurs at high rates, accompanied by high carbon emission rates and fire risk (Hooijer et al., 2012; Page and Hooijer, 2016; Taufik et al., 2022b). When the land surface is near tidal levels, drainage becomes impeded, and flooding becomes more frequent. While there is substantial policy focus on reducing drainage intensity in peatlands (Joosten and FAO, 2012; Wetlands International and Tropenbos International, 2016), and there have been some studies quantifying increasing flood risk due to subsidence (Hooijer et al., 2015; Sumarga et al., 2016), no comprehensive and verified information is available on actual current water level conditions in the region.

Some progress has been made on water table depth determination in peatlands using satellite data, although methods have so far relied on the indirect use of optical and radar data to estimate soil moisture conditions, and are mostly focused on temperate peatlands (Räsänen et al., 2022; Toca et al., 2023). In tropical peatlands, some promising results were reported for areas with limited vegetation cover (Koupaei-Abyazani et al., 2024), relying on calibration for different vegetation types. However, in common tropical conditions where vegetation cover tends to be dense, such indirect approaches may remain of limited practical use.

LiDAR (Light Detection and Ranging, i.e., laser altimetry) data is widely used to directly and accurately measure land elevation (Baltsavias, 1999; Vernimmen et al., 2019; Wehr and Lohr, 1999). While initially only airborne LiDAR data (collected from planes) were available, these techniques were recently upscaled using satellite LiDAR data (Vernimmen and Hooijer, 2023; Pronk et al., 2024). Similarly, satellite LiDAR data are increasingly being used to measure water surface levels (Ryan et al., 2020; Madson and Sheng, 2021; Coppo Frias et al., 2023). We explored whether such data can be applied to measure water table depth below the soil surface (CWD) in large-scale industrial plantation canals, by measuring both land and water surface elevation at the same location. CWD is a commonly used measure of water table depth in drained agricultural areas, sometimes referred to as ‘freeboard’ in plantations, as it is easy to measure and verify.

This study combines the two capacities of land and water surface elevation detection into one measurement using a single LiDAR sensor. This results in a relative measurement of elevation difference where differences in vertical accuracy and referencing between multiple sensors are not an issue, reducing uncertainties.

The study builds on an earlier demonstration of the accuracy of CWD measurements by airborne LiDAR, in a controlled experiment where water levels below the surface were measured in the field on the same day as the overpass of the plane conducting the LiDAR measurements (Vernimmen et al., 2020b). It was found that LiDAR-based CWD measurements correlated well with 145 field site measurements across a wide range of validation areas in Sumatra and West Kalimantan, yielding an R 2 of 0.74, with differences being within 0.25 m for 86% of measurements. Moreover, it was found that part of the differences could be attributed to inaccuracies in field measurements, especially difficulties in measuring a representative ground level, rather than inaccuracies in the LiDAR measurements that were considered to be small.

In this study, we applied ICESat-2 satellite LiDAR data to derive CWD in large-scale industrial oil palm and Acacia pulp wood plantations, using published datasets of canal location and plantation extent available for the islands of Sumatra and Borneo. To provide context to these LiDAR-derived measurements, we compared results with water table depths reported for plantations on peatland in published sources.

The primary aim of the study was to demonstrate this new method of determining plantation water table depths using satellite LiDAR for use by researchers and practitioners, possibly after refinement and downscaling. We also determined how much land is in the elevation categories that we defined, in relation to CWD, with an emphasis on areas where water levels appear too high for productive agriculture, and those where low water levels in peatlands are a concern.

Study area

The study area covers the coastal lowlands of Sumatra and Borneo, defined here as land less than 9 m above global mean sea level as referenced to EGM2008 and determined with the DeltaDTM elevation model (Pronk et al., 2024). The majority of such lowlands consist of peatlands, but there are large mineral soil areas as well (Figure 1).

Satellite LiDAR data selection

ICESat-2 satellite LiDAR data were available from October 2018 to September 2025 (7 years). We used version 7 as published by the US National Snow and Ice Data Center (Neuenschwander et al., 2025; Neumann et al., 2025). Our analysis applied both the ICESat-2 ATL03 data product, which is available at 0.7 m intervals and which we used to determine water surface elevation, and the ATL08 product, which is aggregated and filtered at 100 m intervals and which we used for land surface elevation detection.

ICESat-2 data points were selected that are indicated in the product metadata to have a high confidence (signal_conf_ph = 4) and to be of good quality (quality_ph = 0). Moreover, data points classified as noise by the ATL08 DRAGANN algorithm (classed_pc_flag = 1) were removed. For data points passing the selection, we extracted the surface height (h_ph for ATL03 and h_te_median for ATL08) for further processing.

The ICESat-2 data are referenced to the WGS84 ellipsoid. These data were transformed to the EGM2008 vertical reference, i.e., global mean sea level (MSL), using the 2.5-min version of the EGM08 geoid available from https://github.com/OSGeo/PROJ-data/tree/master/us_nga.

Canal data

The canal dataset used was published by Dadap et al. (2021), who determined canal locations from 2017 Planet optical satellite data at a ∼5 m horizontal spatial resolution using a convolutional neural network in automated routines. While this dataset aims to identify canals in peatlands, canals in surrounding lowland mineral soil areas were also mapped, where they were located within the processed Planet tiles. An advantage of this dataset is that since canal detection depends on space-based visibility, only canals that present open water surfaces of considerable width and without vegetation cover were included, reducing error sources in our analysis. While the dataset presents a canal extent situation 1–8 years before the LiDAR data collection, we found that the position of canals in industrial plantations is fixed, and these plantations have hardly expanded or declined (at least on peat) since Industry and Government moratoriums were announced over the 2012–2016 period. We vectorized the gridded dataset before further analysis.

Supporting data layers

To determine lowland extent and elevation distribution, the DeltaDTM v1.1.1 model (Pronk et al., 2024) was applied, a global coastal digital terrain model with a horizontal spatial resolution of ∼30 m, covering land up to 30 m + EGM2008. This model was created from a combination of ICESat-2 and GEDI satellite LiDAR and Copernicus DEM satellite radar data and is currently considered the most accurate global lowland DTM with an absolute error within 0.5 m outside dense natural forest.

Published peat extent maps were applied. For Sumatra and Kalimantan (Indonesian part of Borneo Island), we used the peat extent published by the Indonesian Ministry of Agriculture (Ritung et al., 2011). For Malaysia, we used maps provided by the European Digital Archive of Soil Maps (Selvaradjou et al., 2005), following the approach described by Miettinen and Liew (2010).

Boundaries of large-scale industrial tree plantations, consisting of oil palm (Elaeis guineensis) and Acacia pulp wood for paper production (mostly Acacia crassicarpa), were extracted from Version 2 of the ‘planted tree database’ (Richter et al., 2024).

Determining WSE, LSE and CWD in canals

Where ICESat-2 flight lines cross canals at angles >30° (to minimize error), WSE was determined as the median of ATL03 values within the canal, assuming a width of 5 m around the centerline, where at least 5 data points were available and if the range between the 25 and 75 percentile values did not exceed 0.25 m. Land surface elevation (LSE) was determined from ATL08 data between 100 and 250 m away from the canal side, to avoid inclusion of data points over the canal water surface or too close to the next canal, and the mean LSE was calculated as the mean of LSE on either side of the canal. Locations where the difference between mean LSE on either side of the canal exceeds 0.5 m were excluded from further analysis, as this suggests uneven or relatively high slopes exceeding 1 m km -1 (slopes over 1 m km -1 are less common in lowlands and peatlands; Vernimmen et al., 2019).

The canal water table depth below the soil surface (CWD) was then determined as the difference between WSE and LSE (Figure 2). Where WSE was found to be above LSE, this was considered potentially due to measurement inaccuracy, and the result was not included in further analyses.

In the final analysis, LSE values below 1 m + EGM2008 were excluded as they were rare and rarely yield valid CWD values. Locations above 9 m + EGM2008 or more than 100 km from the coastline were also excluded, to limit the analysis to the extent of coastal lowland.

CWD results were summarized into key statistics by Province, presented in maps and tables and linked to LSE, for four elevation ranges: 1–3, 3–5, 5–7 and 7–9 m.

Assessment of possible sources of bias in canal sampling

In Supplementary Information, we present detailed analyses, summarized in Results, of how the distribution of [a] canal length and density and [b] canal orientation, as included in the analysis, relates to elevation zones, industrial plantations, soil types and regions.

Sensitivity assessment of filter parameter settings

We also assessed what the effect on overall CWD results would be of applying ‘looser’ criteria to filter valid measurement values; the results are presented in Supplementary Information.

While we in principle exclude CWD values above the surface, as these could in some cases incorrectly suggest the land is flooded, we determined what the effect on overall statistics would be when including values up to 0.25 m above the surface. Likewise, we determined the effect of inclusion of CWD measurements where the maximum difference between LSE points is within 1 m (instead of 0.5 m), i.e., tolerating elevation differences in the reference landscape that well exceed most CWD values.

Interannual variability

In Supplementary Information, we analyzed whether there was an impact of the El Niño Southern Oscillation on mean CWD values between the full years in the data acquisition period and if there was a bias in the temporal distribution of the measurements.

Literature data for comparison

As the CWD measurement locations are randomly distributed, being dependent on chance intersections between satellite flight lines and canals, they cannot be compared with simultaneous ground measurements in the same locations directly. Instead, publications were identified that present time series graphs of water table depth (WTD) in plantations on peatlands across the study area. Monthly data were imported from online sources or scanned from graphs; in some cases, multiple similar records from nearby locations were combined into a single record. Statistics were derived from grouped records per region (Table 1).

Statistical analysis

CWD distributions were compared across surface elevation ranges and years and between soil types using non-parametric statistical tests.

Within soil type comparisons across elevation ranges and years

For each region/soil type combination, differences in CWD among the four surface elevation ranges (1–3, 3–5, 5–7, and 7–9 m above EGM2008) and years were assessed using the Kruskal–Wallis H-test (Kruskal and Wallis, 1952). Where the Kruskal–Wallis test indicated a statistically significant overall effect (p < 0.05), pairwise post hoc comparisons between all elevation range combinations and years were performed using Dunn’s test (Dunn, 1964), with p -values adjusted for multiple comparisons using the Holm–Bonferroni sequential correction (Holm, 1979).

Between soil type comparisons within elevation ranges and years

Differences in CWD between peat and mineral soils within each elevation range and year were assessed using the two-sided Mann-Whitney U test (Mann and Whitney, 1947). To account for the four and six simultaneous comparisons (one per elevation range and year) within each region, raw p -values were adjusted using the Holm–Bonferroni correction.

All analyses were conducted in Python using the SciPy 1.15.2 (Virtanen et al., 2020) and scikit-posthocs 0.14.0 (Terpilowski, 2019) libraries.

Distribution of CWD measurements

A total of 15,575 high-quality CWD measurements were selected for analysis over the 7-year study period. We found few CWD measurements for Acacia plantations in Borneo and have therefore removed such occurrences from further CWD analysis. At 64% (9,946 measurements), most CWD measurements are located on peat, with the majority of those in oil palm (69%) compared to Acacia plantations (31%) (Table 2). On peat, measurements in oil palm plantations are in most regions reasonably well distributed over the 1–9 m elevation range, with the highest number in the 3–5 m range, but measurements on mineral soil are concentrated (67%) in the 1–3 m zone. The distribution of measurements in Acacia plantations is less even, with the highest numbers in the 1–3 m (mostly in South Sumatra) and 7–9 m (mostly in Riau) elevation ranges. CWD measurements in Acacia plantations on mineral soil are limited in number (264) compared to peat (3,056) but were nevertheless included in further analysis for consistency.

Relation of CWD with land surface elevation

CWD values for oil palm plantations, for different areas and elevation ranges, are shown in Figures 3, 4. Mean CWD is greater (i.e., the water table is further below the surface) at higher elevations in all areas (Figure 3). Values in oil palm plantations range from −0.22 m in the 1–3 m elevation range to −0.63 m for 7–9 m (Table 2; Figure 3) on peat and from −0.19 to −0.52 m on mineral soils. For all regions combined, regardless of being on peat or mineral soils, mean CWD in oil palm plantations is significantly different between the 1–3, 3–5 and 5–9 m ranges, while mean CWD is not significantly different between the 5–7 and 7–9 m elevation ranges. Comparing peat and mineral soils, mean CWD is significantly greater in peat compared to mineral soils for all regions combined (Figure 3). Across the three different regions (Sumatra, South Borneo, North Borneo), CWD patterns are consistent with regional mean CWD being within 0.19 m at most from overall mean CWD values.

CWD patterns and variation are similar for Acacia plantations in Sumatra, albeit generally ∼0.20 m lower compared to oil palm plantations in Sumatra (over the 1–9 m elevation range; Table 2; Figure 4).

Non-linear relationships fitted through the elevation range and mean CWD values shown in Figure 4 are as follows:

All OP on peat:

Sumatra Acacia on peat:

All OP on mineral soil:

Sumatra Acacia on mineral soil: where elev_mean is the mean value of the elevation range, CWD is the mean CWD in meters, and the standard error (SE) is in brackets. It should be noted that due to the limited number of elevation bins (n = 4), the quadratic fit has only 1 residual degree of freedom. The standard errors on the quadratic and intercept terms are large relative to their estimates (in some cases exceeding the estimate itself), indicating these terms are not well constrained by the available data.

Frequency distribution of CWD values

At lowest elevations, 1–3 m above global MSL, 68% of CWD values across combined plantation and soil types are in the −0.2 – 0 m range (Table 2), which likely indicates frequent flooding; this reduces with elevation, to 23% in the 7–9 m range. Conversely, only 6% of CWD values at 1–3 m elevation are below −0.8 m, indicating very dry conditions, but as much as 40% at the 7–9 m range. CWD percentages in the −0.2 – 0 m range are higher for oil palm than Acacia, and higher on mineral soils than on peat, across all elevation zones and regions.

Sensitivity of mean CWD values to filter settings

Inclusion of CWD values between 0 and +0.25 m brings the overall mean across all regions, elevations, and industrial plantations up by 0.13 m, from −0.42 m (Table 2) to −0.29 m (Supplementary Table S1). At elevations of 1–3 m, CWD values would become −0.13 m overall and −0.10 m on mineral soils. Such extremely high water tables would indicate near-permanent waterlogging and flooding, and would rule out any successful agriculture; we find these values inconsistent with observed conditions. Therefore, we conclude that exclusion from analyses of CWD values above the soil surface is justified, while acknowledging that this may result in some overestimation of CWD, especially at lowest elevations.

Inclusion of CWD values where the difference in LSE values is between 0.5 and 1 m, i.e., of areas with greater surface roughness and/or slopes, brings the overall mean CWD somewhat down by 0.04 m, from −0.42 m (Table 2) to −0.46 m (Supplementary Table S2). We conclude that this LSE difference threshold does not affect the outcome much and therefore consider the 0.5 m threshold appropriate.

Comparison of satellite LiDAR derived CWD with field measurements of WTD

All published field studies were found to be located at elevations above 4 m + EGM2008, with an average above 8 m + EGM2008 (Table 1). If we consider all satellite LiDAR CWD measurements in OP plantations over the 3–9 m elevation range (Table 2; Figures 3, 4), mean values range from −0.42 to −0.60 m. This is near the mean value of −0.52 m as calculated from 16 published OP plantation WTD records (Table 1). Equally, the mean satellite LiDAR CWD range of −0.63 to −0.87 m found for Acacia plantations over the 3–9 m elevation range is comparable to the literature mean of −0.71 m across 4 publications (Table 1).

Distribution of industrial plantations over elevation ranges

The total area of industrial oil palm and Acacia plantations in the 1–9 m lowland elevation range is 39,693 km 2, or 17% of the total land area between 1 and 9 m. Adding an area of 47,412 km 2 for ‘other plantations’, total tree plantation extent makes up over a third (38%; 87,105 km 2) of the Sumatra and Borneo coastal lowlands (Table 3).

Of the industrial plantations, 30,812 km 2 (77.6%) is oil palm and 8,880 km 2 (22.4%) is Acacia, with an area of 9,847 km 2 or 25% in the most flood-prone 1–3 m range, with 7,681/2,167 km 2 in oil palm and Acacia, respectively. The peatland lowland area under industrial plantation is 22,926 km 2 (58% of all lowland area under industrial plantation), of which 4,217 km 2 is in the 1–3 m range (Table 3).

Possible sources of bias in canal sampling

Assessments presented in Supplementary Information demonstrate that canal length, canal density and canal orientation, as included in the analysis, are mostly evenly distributed over elevation zones, industrial plantations, soil types and regions. Therefore, it can be ruled out that uneven canal distribution has caused a substantial bias in CWD measurements.

Interannual variability in CWD values and counts

For all regions and large-scale industrial oil palm and Acacia plantations combined, a year-to-year CWD variation is evident in plantations on peat (Supplementary Figure S1), with the lowest values occurring in the ‘El Niño’ affected years of 2019 and 2023 that experienced less rainfall in the last quarters. Although the variation in annual CWD values is no greater than 0.1 m at this aggregation level, the difference is significant compared to other years, except 2024.

On a year-by-year basis, the number of CWD measurements (‘counts’) is rather evenly distributed in time (Supplementary Figure S2). Quarterly, large variations in counts are sometimes seen, but these do not seem to be closely related to rainfall regime, as low counts are found both in some ‘El Niño’ affected quarters and in some quarters with normal rainfall. We therefore conclude that there is no strong bias in the temporal distribution of usable ICESat-2 acquisitions that could substantially affect CWD outcomes.

Key findings of the study

A strong relation is evident between satellite LiDAR-derived CWD and land surface elevation, with water tables in the lowest areas (1–3 m) generally being less than −0.4 m below the surface, as also demonstrated in Figure 5. This relation applies across different regions, both in peatlands and mineral soil areas, and in oil palm as well as Acacia plantations (Figure 4; Table 2; Equations 1–4).

Spatial variations are evident with CWD in Acacia plantations being greater than in oil palm, and peatlands having somewhat greater CWD than mineral soil areas, but every data subset without exception shows a pattern of greater CWD (lower water tables) where land surface elevation is higher. Mean CWD in the 1–3 m elevation range just above mean sea level is within a narrow range of −0.19 to −0.24 m for all subsets (Table 2). Oil palm CWD at higher elevations between 5 and 9 m varies from −0.44 to −0.71 m (except a small group of measurements on mineral soil in North Borneo), while Acacia CWD in Sumatra reaches ∼ -0.85 m at such elevations. North Borneo has somewhat higher water tables compared to the other two areas, regardless of soil type or elevation (Table 2; Figure 4), probably due to a wetter climate with higher rainfall throughout the year (Hooijer and Vernimmen, 2026).

Accuracy of satellite LiDAR-derived CWD values

While no validation with field measurements for the same dates and locations is possible for this type of data, beyond the earlier study by Vernimmen et al. (2020b), there are several types of indirect evidence that the satellite LiDAR CWD measurements are accurate. First, our CWD findings match water table depths reported in the literature for higher elevations, including the consistent difference between oil palm and Acacia plantations. Second, the relations between CWD and land surface elevation would not have been so consistent and similar for different regions had the individual measurements not been accurate. Finally, visual inspection of water surface elevation measurements along drainage canals from higher land surface elevation to the coast shows water table gradients as expected, and as required for drainage, as is demonstrated in Figure 5.

We note that this indirect evidence of accuracy does not remove some inevitable uncertainties from the analysis. Vegetation is known to interfere with the LiDAR signal, potentially resulting in some uncertainty in land and water surface elevation (LSE and WSE) measurements. We have mitigated this by determining LSE from multiple ICESat-2 ATL08 data segments that are each created from multiple ATL03 photon data points, reducing error (Neuenschwander and Pitts, 2019). We also ensured that vegetation is a minimal issue in water surface elevation detection by using the Dadap et al. (2021) dataset that relies on optical identification of open water surfaces, visible from space as they are not covered by tree canopy or floating vegetation.

The focus on industrial plantations further reduces the likelihood of canals being covered by tree canopy (as plantations often maintain tree-free zones along canals for access) and of ground vegetation reducing LiDAR penetration (as oil palm plantations especially often remove weeds by spraying and sometimes controlled fire). Pronk et al. (2024) found that the accuracy of the DeltaDTM elevation model, which uses ICESat-2 ATL08 data to correct Copernicus DEM, is lowest where dense ground vegetation occurs in burnt tropical peatlands (classed as ‘grassland’ following international standards but in fact consisting of ferns), but moderate in SE Asian croplands and forest. This is consistent with unpublished assessments, where we found ATL08 accuracy to be higher in plantations than in forest and other landcover types.

Representativeness of satellite LiDAR-derived CWD values

When considering the findings reported here, inherent limitations of both satellite LiDAR CWD measurements and published water table depths (WTD) should be noted.

Our CWD dataset consists of measurements at intersections between ICESat-2 flight lines and canals. Dadap et al. (2021) reported a high accuracy of the canal map in their validation and test areas, but noted that a lack of contrast between canals and surrounding vegetation may have contributed to not identifying certain canals. We indeed noted some data gaps in the canal maps. Moreover, the CWD analysis constraint to angles between the flight line and canal being over 30° excludes canals with more or less N-S orientation (as ICESat-2 orbit inclination is 92°). However, the angle constraint is not expected to have introduced bias, considering that the distribution of canal orientation across elevation zones, industrial plantations, soil types, and regions is more or less uniform (Supplementary Table S5). The combined data gaps may leave some local areas with limited data, but overall data density suffices for CWD characterization at the landscape scale, as demonstrated for East Sumatra in Figure 6.

The limitations of published WTD records include an understandable but rarely noted preference amongst plantation managers to allow research in areas that are considered better managed, i.e., with water levels that do not deviate far from targets. This will sometimes result in records that, even while accurate for the study areas, are not representative of general conditions in a region. This may be the reason for the average position above 8 m of nearly all plantations for which records are published (Table 1), i.e., excluding the lower-lying areas where flooding issues and some other water management complications are more likely.

Cause and extent of lowland plantation areas with likely high flood risk

The very limited water table depth often found for industrial plantations in the 1–3 m elevation range, yielding an overall mean CWD value of ∼ -0.2 m, is likely explained by the limited gradient between water tables in such areas and the nearby tidal range, with high tide levels in some cases approaching soil surface levels. With only gravity drainage available, i.e., in the absence of large pumping stations, such conditions prevent effective water discharge to the sea, especially in periods of high rainfall.

Internal plantation water management protocols in the region usually consider water table depths > -0.4 m to be ‘flood risk’ level (‘banjir’) and to be avoided. It is therefore conservative to assume that water table depths > -0.2 m indicates conditions of near-flooding (waterlogging) or actual flooding.

The area of 984,700 ha (9,847 km 2) of industrial oil palm and Acacia plantation in the 1–3 m elevation range (Table 3) is a minimum estimate of the lowland area in Sumatra and Borneo that likely has frequent flooding occurrence. This area makes up a quarter (∼25%) of the lowland industrial plantation extent of 3,969,300 ha (39,693 km 2).

Besides large-scale industrial oil palm and Acacia plantations, an even larger area of 4,741,200 ha (47,412 km 2) of other tree plantation types exists in the 1–9 m elevation range (Table 3). These are mostly smaller-scale developments, often smallholders, usually growing oil palm but also rubber, coconut and sago. Of this, a large proportion of 39%, i.e., 1,855,400 ha (18,554 km 2), is in the lowest range of 1–3 m. Noting that smallholders face even greater challenges than industrial plantation managers in achieving controlled water levels (Uda et al., 2017; Khasanah and Van Noordwijk, 2019; Roundtable on Sustainable Palm Oil RSPO, 2019; Supriatna et al., 2024), this could bring the total area of highly flood-prone tree plantation land in the 1–3 m elevation range to 2,840,100 ha (28,401 km 2), of which 37% (10,554 km 2) is on peat.

Another perspective is to consider that half (50%) of all 15,575 CWD measurements across all regions, elevation zones, industrial plantations and soil types are in the −0.2 to 0 m range (Table 2). While the CWD measurement distribution is not entirely proportional to the distribution of plantations, this could yield a tentative estimate of 1.98 million hectares of industrial plantations (half of 39,693 km 2; Table 3) and 2.37 million hectares of other plantations that would currently have very high water levels, i.e., 4.35 million hectares overall.

On this basis, we estimate that the total area of lowland plantations at high current risk of frequent flooding will likely reach 2.8 million hectares and could potentially exceed 4 million hectares.

It is worth noting that industrial plantations cover only 134 km 2 (13,400 ha) below 1 m, i.e., below the elevation range studied. Over the 9–30 m elevation range outside our study area, a more substantial area of industrial plantations on peatland exists, at 566,900 ha (5,669 km 2), of which 78% (444,200 ha) is oil palm and 22% (122,700 ha) is Acacia. Overall, our analysis focusing on the 1–9 m elevation range includes 78% and 86% of oil palm and Acacia plantations below 30 m, respectively.

Cause and extent of peatland plantation areas with low water tables

Oil palm plantations in the 5–9 m range have a mean CWD of −0.6 m, which is within most stated management targets although below the Indonesian policy target of −0.4 m (Government of Indonesia, 2016).

Mean CWD in Acacia plantations on peatland at 5–9 m is found to be around −0.85 m, which is well below most stated water management targets and will cause high subsidence rates, fire risk, and carbon emissions (Hooijer et al., 2012; Jauhiainen et al., 2012). The extent of such land is 445,600 ha (4,456 km 2). Acacia plantations at elevations above 9 m will likely have similarly low water tables, but that was not investigated in this study.

A CWD difference between oil palm and Acacia of around 0.2 m is consistent with field reports (Table 1), as earlier noted by Carlson et al. (2015), who report an overall average of −0.67 ± 0.20 m, and has several explanations. First, oil palm is more tolerant of occasional flooding, so it can be managed at somewhat higher mean water tables. Second, individual oil palm plantation concessions are typically considerably smaller than Acacia concessions, and have smaller internal production districts, which allows a more fine-grained water management approach. Finally, oil palm is planted on a rotation cycle of ∼25 years, and harvests are transported over roads or in small boats, whereas Acacia pulp wood has a 5-year rotation cycle with harvests being transported in large barges that require removal of water level control infrastructure from canals for extended periods. As a result, canals in oil palm can have mature water level control infrastructure that may be fine-tuned over years to manage small water steps over numerous gates, which is harder to achieve in Acacia plantations.

Referencing to local mean sea level instead of global MSL.

The current analysis applies elevation data referenced to global MSL (EGM2008), as no full regional elevation model referenced to local MSL currently exists. It is known that in SE Asia, the EGM2008 reference results in land surfaces being from 0.5 to 1 m higher above actual local MSL (Vernimmen et al., 2019; Seeger and Minderhoud, 2026). If local MSL were applied and the difference would indeed be 0.75 on average, our study would likely cover an elevation range of ∼0.25–∼8.25 m above MSL instead of 1–9 m. When assessing the findings presented here in view of local conditions, this 0.25–8.25 m elevation range is best assumed, until more precise datasets for local MSL re-referencing become available. We do not expect such future re-referencing to result in very different estimates of the area at risk of flooding.

Peat soils vs. mineral soils

The maps used to separate peatlands from mineral soils in this study are not considered highly accurate (Hooijer and Vernimmen, 2013), but no better maps are generally accepted or available in the public domain at present. They are based on limited field data and without considering topography, as we found necessary for accurate peat mapping (Vernimmen et al., 2020a). Moreover, the field data used is at least 20 years old; hence the maps do not reflect the considerable peat loss that has since occurred in drained peatlands, in some drained areas already resulting in the complete loss of several meters of peat since the 1990s (Vernimmen et al., 2020a). Therefore, it is possible that the peat maps used overestimate current peat extent, especially in some lowest-lying areas where the peat layer was thin to begin with. However, since CWD results for peatlands and mineral soils are similar, especially at the lowest elevations, this uncertainty does not affect results.

Water table depths in canals and in the soil

While CWD is a convenient and often-used parameter for water table depth in drained systems, it is not necessarily always equal to water table depth (WTD) as measured inside the soil. In tropical peatlands, however, CWD and WTD are found to be similar, as high permeability of the woody peat ensures rapid water flow between peat and canals. Published hydraulic conductivity values for tropical peat are far above those in temperate peatlands (Yonekura et al., 2025), with published values between 3 and 100 m d -1 (Salim et al., 2021; Hooijer et al., 2024; Mahardika et al., 2024). Hooijer et al. (2024) report values around 100 m d -1 over the fibric top 3 m of peat, for a large number of ‘peat pit’ field measurements in Sumatra and Kalimantan. In contrast, published values for temperate peat, generally sphagnum peat, tend to be between 10 −7 to 10 −2 m d -1 (Clymo, 2004; Mustamo et al., 2016; Kaczmarek et al., 2023), i.e., at least two orders of magnitude lower.

A further explanation of the limited difference between water tables in canals and in the soil is the small distance between drainage features (∼500 m between main canals, less than 100 m between field drains) that are common in SE Asian lowlands, in response to the high peak rainfall that is characteristic of the region.

As a result, from field experience in plantation water management, we know that overall, measured across the seasons and at multiple locations, water levels in canals tend to be less than 0.1 m below those in the peat. Hooijer et al. (2024) found that peat water levels were 0.05 m above canal water levels in carefully controlled field measurements in Sumatra. This difference is well within the accuracy range of most ‘dipwell’ field measurements of water table depth in the peat, which are commonly affected by variations in surrounding topography as well as peat surface subsidence relative to the dipwell, factors that are rarely corrected for in our experience.

While hydraulic conductivity in mineral soils in the region is well below that in peat, due to the high clay content, drainage densities in such areas tend to be even higher, with field drain intervals in flood-prone areas often being less than 10 m (one ditch for every one or two rows of oil palm).

We conclude that, unless a local relation between CWD in canal and WTD in the peat or mineral soil is established from local measurements, uncorrected CWD may be used as a proxy for WTD, as is indeed common practice in some plantation water management. If a conservative correction is required, we would advise adding 0.1 m to CWD values, which would yield overall estimates of ∼ -0.5 and ∼-0.75 m for mean water table depth in peat soils at higher elevations, across regions.

Use of findings

Our findings will be of use to researchers wanting to better understand water table depth conditions in drained lowlands in SE Asia, for instance in relation to coastal flood risk projections, land suitability assessments, or carbon emission studies. They may also be considered by practitioners as a way to assess spatial variation in water management requirements specific to crops and land surface elevation.

For specific locations, a higher local satellite LiDAR CWD measurement density can be achieved than is presented in this study. We have applied strict conditions on ICESat-2 data selection to minimize possible error sources and to ensure we compared similar areas across different regions, which may be relaxed when accounting for known local conditions.

Broader implications of findings

We have identified pronounced patterns in CWD that are in large part explained by land surface elevation. It appears that these physical landscape factors are important controls on water management outcomes, and that water level management targets in many areas are likely not achieved. While it was beyond the scope of this study to determine precise challenges in water management, we propose that such landscape water management limitations linked to elevation are considered in planning and design.

Where landscape factors create insurmountable obstacles to achieving optimum water levels for conventional crops, such areas could be prioritized for introducing sustainable production models using flood-tolerant species, or for full forest restoration. Where the observed frequent flooding at lowest elevations would be inevitable and increasing, it may be realistic and cost-effective to aim for flood-tolerant crops at such elevations, or for full restoration to swamp forest (including mangrove at very lowest elevations near sea levels), rather than to continue investments in oil palm and Acacia plantations that are likely to remain unproductive (Hooijer et al., 2015; Sumarga et al., 2016). And if it is indeed too difficult to maintain high water levels in peatlands at higher elevations, maybe plantations in such locations could also be prioritized for restoration to natural forest, contributing to the stated policy aim of capturing water in conservation areas on top of peat domes that will help stabilize water levels at lower elevations (Wetlands International and Tropenbos International, 2016).