The Mahakam Lakes are located in the Central Mahakam Sub-watershed, East Kalimantan Province, Indonesia (Figure 1). Lake Jempang is hydrologically isolated from Lake Melintang and Lake Semayang by the Mahakam River. During the wet-season inundation, Lakes Melintang and Semayang coalesce into a single water body, with the Pela River serving as the primary outlet. More information provided in Supplementary Table S1.

Lake shorelines were delineated from cloud-free Landsat imagery collected from five specific years (1990, 1996, 2000, 2010, 2020), acquired from the United States Geological Survey (USGS) and the Indonesian National Research and Innovation Agency (BRIN). Image specifications: Landsat 5 TM (bands 7-4-2) for 1990 and 1996; Landsat 7 ETM+ (bands 7-4-2) for 2000; Landsat 8 OLI (bands 7-5-3) for 2010 and 2020. All Landsat scenes were obtained from the Collection-2 Level-2 products, which include radiometric calibration, atmospheric correction, and cloud masking. Remote sensing has become an essential tool for lake research (Palmer et al., 2015; Dörnhöfer and Oppelt, 2016). Challenges in remote sensing of inland waters include cloud cover and water turbidity (Palmer et al., 2015).

Bathymetric data were obtained from the General Bathymetric Chart of the Oceans (GEBCO) 15 arc-second grid (∼450 m resolution). Despite its coarse resolution, GEBCO was selected because it provides the only temporally consistent bathymetric coverage for the study period (1990–2020) across all three lakes. High-resolution hydrographic surveys were unavailable for the historical time steps, and satellite altimetry data (e.g., ICESat-2) do not extend back to the early years of this analysis. Thus, GEBCO represents the most feasible dataset for reconstructing decadal morphometric trends, a common constraint in tropical lake studies (Lo et al., 2022). MODIS satellite imagery has been applied for detailed lake morphometry in systems with large water level fluctuations (Ovakoglou et al., 2016), further illustrating the diversity of available satellite-based approaches, though we prioritized temporal consistency over higher resolution for this multi-decadal analysis. We acknowledge that a direct comparison between GEBCO and higher-resolution altimetry (e.g., ICESat-2) for the overlapping years (2018–2020) would help quantify systematic biases, but such an analysis is beyond the scope of this study. Future work should prioritize this comparison to refine volume and depth estimates.

Spatial measurements were performed using QGIS version 3.34.10. Surface area (A₀) was calculated via digital planimetry; shoreline length (SL) was measured as the total perimeter. Maximum depth (Z_max) for 2010 and 2020 was extracted from GEBCO-interpolated bathymetric surfaces using ordinary kriging. For years without direct bathymetric observations (1990–2000), Z_max was held constant based on historical surveys (see Section 4.3).

Volume was calculated using the raster surface volume tool in QGIS (version 3.34.10) by integrating the difference between lake bottom elevation (extracted from GEBCO) and mean water surface elevation, assuming a constant water level at the time of satellite overpass. The original GEBCO grid at 450 m resolution was used without resampling. The integration sums the difference values over all raster cells within each lake polygon (cut-fill method), producing a total volume estimate. No additional interpolation or resampling was applied to the GEBCO data for volume calculation. We acknowledge that inter-annual water level fluctuations are not accounted for due to the lack of continuous gauge data, which introduces additional uncertainty (see Section 2.4). Ordinary kriging was applied only for generating continuous bathymetric surfaces in Figure 2 visualization; statistical analyses (e.g., VD calculations) were performed directly on the original GEBCO grid values without additional interpolation (Li et al., 2023).

Volume Development (VD) and Shoreline Development Index (SDI) was calculated following Haakanson (1981) and Johansson et al. (2007) based on the original formulation by Mortimer (1959): V D = 3 Z ^ Z max (1) S D I = S L 2 π A 0 (2)

Shoreline Development Index (SDI) was calculated using Equation 2. Where Z ^ = mean depth (V/A₀) and V = lake volume derived from bathymetric interpolation. SDI >1 indicates increasing shoreline irregularity relative to a circular lake.

Volume and depth estimates derived from GEBCO’s 450 m resolution carry inherent uncertainties. Following approaches in recent bathymetric studies (Li et al., 2023), a potential error of ±10%–20% for volume calculations is estimated based on the coarse grid. Additional uncertainty arises from shoreline digitization: manual delineation of lake boundaries from Landsat imagery typically introduces ±1%–2% error in lake area (Palmer et al., 2015). For maximum depth (Z_max), a variation of ±1 grid cell (∼450 m) was considered, translating to approximately ±0.5 m of uncertainty for shallow lake environments. Propagating these sources of uncertainty (volume ±10%–20%, area ±1%–2%, Z_max ±0.5 m) yields a total potential error of approximately ±20%–30% for volume change estimates. Therefore, only decadal changes exceeding ±20%–30% are considered meaningful; smaller changes may fall within the methodological noise. These error ranges are used in the interpretation of morphometric changes but did not alter the overall trend patterns observed.

Bathymetric interpolation using GEBCO geospatial data revealed substantial morphometric reorganization across the three lakes between 2010 and 2020 (Figure 2). Lake Melintang exhibited the most pronounced transformation: its bathymetric cross-section decreased considerably to shallower levels, with the 5 m isobath shifting toward the lake center (Figures 2C,D), indicating progressive basin infilling (Lo et al., 2022). Such shifts in isobaths are commonly used as indicators of sediment accumulation in tropical floodplain lakes (McGlue et al., 2015). In contrast, Lake Semayang retained deeper bathymetric zones despite a reduction in surface area, suggesting that deeper channels persisted or even deepened (Figures 2E,F; Håkanson, 2005).

Volume Development (VD) values calculated from bathymetric data for 2010 and 2020, revealed starkly contrasting lake trends (Figure 3). Lake Melintang exhibited the most pronounced decline: VD dropped from 0.94 in 2010 to 0.52 in 2020 (ΔVD = −0.42)—a 46% reduction in basin concavity, suggesting rapid sedimentary infilling. This magnitude of change over a single decade represents one of the most severe infilling cases documented in an Indonesian priority lake to date.

Lake Semayang showed a contrasting trend: VD increased from 0.06 in 2010 to 0.24 in 2020 (ΔVD = +0.18) despite concurrent surface area reduction. However, this VD increase occurred alongside a substantial decrease in maximum depth (from 7.5 m to 1.5 m) as derived from GEBCO. While the absolute Z_max values carry large uncertainty due to the coarse grid resolution (see Section 4.3), the shallowing trend itself is consistent with the observed volume loss and the lake’s overall degradation. However, without independent validation, these Z_max values should be treated as preliminary indicators of shallowing rather than as confirmed measurements. This indicates that the VD rise may partly reflect a mathematical response to rapid shallowing rather than an unambiguous ecological recovery (Table 1). Thus, concluding ecological improvement from VD alone would be misleading, as the index’s increase is driven primarily by a shrinking denominator (Z_max) rather than by true basin deepening. Similar behavior has been observed in other shallow tropical lakes where extreme changes in maximum depth disproportionately affect VD (Johansson et al., 2007). This highlights that area loss does not necessarily equate to volume loss—a critical distinction for monitoring programs relying solely on satellite-derived surface area (Håkanson, 2005).

Lake Jempang remained mostly stable throughout the study period, with VD values ranging from 0.63 to 0.71 (ΔVD = −0.08 between 2010 and 2020). This stability suggests Lake Jempang functions as a through-flow system with efficient sediment export, contrasting sharply with Lake Melintang’s sediment-retentive geometry.

The 2010–2020 decade captured pronounced volumetric dynamics (Table 1). Lake Melintang experienced a substantial volume decreease: from 579.9 × 10⁶ m³ in 2010 to 116.6 × 10⁶ m³ in 2020—a reduction of 463.3 × 10⁶ m³ (80%) over a single decade. Using this volume loss and an average lake area of approximately 200 km², the corresponding theoretical sediment thickness would be roughly 2.3 m, implying a very rough sedimentation rate of about 23 cm yr^-1 if the entire volume loss were due to sediment accumulation. This value is two orders of magnitude higher than the global average of 4.5 mm yr^-1 for tropical floodplain lakes (Lo et al., 2022), strongly suggesting that the estimate is dominated by methodological limitations (e.g., GEBCO’s coarse resolution) rather than reflecting actual sediment accumulation. Hence, this number should be viewed as a crude approximation only and is not used as a primary quantitative conclusion. Over the same decade, regional temperatures in East Kalimantan increased by approximately 0.3 °C per decade (BMKG, 2020), which may have contributed to higher evaporation rates and further volume loss. However, given the large uncertainties, the relative contributions of sedimentation versus evaporation cannot be reliably separated. This decrease indicates that Lake Melintang is undergoing rapid shallowing, though the exact infilling rate remains uncertain.

Lake Jempang showed a moderate volume decrease from 342.6 × 10⁶ m³ to 269.0 × 10⁶ m³ (Δ = −73.6 × 10⁶ m³; −21%), despite relatively stable surface area and VD. This suggests that Jempang is experiencing gradual sediment accumulation rather than the large infilling observed in Melintang.

Lake Semayang exhibited a contrasting trend: volume decreased from 27.3 × 10⁶ m³ to 18.4 × 10⁶ m³ (Δ = −8.9 × 10⁶ m³; −33%). However, this volume loss occurred alongside increasing VD (from 0.06 to 0.24). The very low mean depth values derived from GEBCO for Semayang (0.12–0.14 m) are likely artifacts of the 450 m grid resolution, which systematically underestimates water depth in shallow, morphologically complex floodplain lakes when compared to historical depth ranges (3–6.5 m; (Muchlis et al., 2022)). Therefore, absolute depth and volume estimates from GEBCO should be interpreted as relative indicators of shallowing trends rather than as precise bathymetric measurements. With this caveat, the observed VD increase is largely driven by a sharp decline in maximum depth (from 7.5 m to 1.5 m in the GEBCO grid) rather than an actual increase in basin concavity. This apparent paradox is consistent with differential shallowing: nearshore areas infill while deeper channels may maintain or deepen their bathymetry—a pattern observed in other river-connected floodplain lakes where flow concentration maintains deeper channels despite overall shallowing (Hamilton, 1999). Nevertheless, the 33% volume loss indicates ongoing degradation, and the VD increase should be interpreted cautiously as a morphometric artifact (Håkanson, 2005; Johansson et al., 2007).

SDI values ranged from 1.12 (Semayang 2020) to 1.33 (Jempang 2000) (Supplementary Table S2). Lake Jempang consistently exhibited the most complex shoreline (SDI ∼1.3), reflecting its geologically heterogeneous margins. Notably, shoreline complexity did not correlate strongly with volumetric trends, suggesting that SDI responds to different forcings (e.g., local erosion/deposition) than basin-scale infilling captured by VD (Johansson et al., 2007). This decoupling between shoreline complexity and volumetric change has also been observed in other tropical lake systems where morphological processes operate at different spatial scales (McGlue et al., 2015).

The starkly contrasting VD trajectories of three hydrologically connected lakes within the same river system provide a natural experiment demonstrating that morphometric change is not uniform (Håkanson, 2005). Rather, it reflects local-scale processes superimposed on regional drivers—such as differential sediment supply, hydrological connectivity, and basin morphology (Pham et al., 2016; Lo et al., 2022). Lake Melintang’s VD decline of −0.42 between 2010 and 2020 reveals a concerning and remarkably high rate of infilling, which appears to be among the highest reported for an Indonesian priority lake.

This finding aligns with the global meta-analysis by Lo et al. (2022), which documented exponentially increasing sediment accumulation rates in tropical floodplain lakes over the past 50 years, primarily in response to population growth and deforestation in topographically steep catchments with pronounced seasonal rainfall. Their study demonstrated that tropical floodplain lakes accumulate sediment more rapidly than many extratropical lakes on centennial timescales, with some lakes projected to fill completely within centuries. Our findings are consistent with sediment transport modeling in the Mahakam continuum (Pham et al., 2016). Comparable sediment deposition patterns occur in African floodplain systems (Omengo et al., 2016). Geochemical characteristics of Indonesian tropical lake sediments provide context for interpreting infilling (Nomosatryo et al., 2021).

A key distinction of our study lies in the magnitude of change observed within a single decade. Melintang’s VD decline of 0.42 over ten years indicates that infilling processes are occurring far more rapidly than the global averages reported by Lo et al. (2022) for channel-type lakes (0.73 mm yr^-1 long-term vs. 4.51 mm yr^-1 short-term). This fluctuation suggests that sediment trapping efficiency can increase dramatically in tropical flood-pulsed lakes during low-water phases, when sediment-laden water from surface runoff and river inflows plunges directly into the remaining hypolimnion (Hamilton, 1999; McGlue et al., 2015).

Our findings further strengthen the argument that VD change (ΔVD) serves as a more sensitive early indicator compared to other morphometric parameters. As illustrated by (Li et al., 2023) in their study of lake volume estimation on the Tibetan Plateau, low-cost approaches integrating remote sensing data with limited surveys can yield reliable volume estimates (mean volume bias of approximately 15%). In our context, ΔVD successfully captured infilling signals before volume loss became detectable through surface area monitoring alone.

Lake Semayang’s VD increase (+0.18) alongside a 33% volume loss illustrates that VD trends must be interpreted within the context of all morphometric parameters. The VD rise is largely driven by a sharp decline in Z_max (from 7.5 m to 1.5 m), which reduces the denominator in the VD equation (VD = 3Ẑ/Z_max). Consequently, the index may increase even when the lake becomes shallower overall. This mathematical caveat does not represent ecological recovery; rather, Semayang continues to experience substantial volume and depth reduction. The apparent “concavity” suggested by VD alone thus warrants cautious interpretation. Numerous studies have demonstrated that relying solely on satellite-derived surface area can be misleading when assessing lake health, as volume changes may not correlate with area changes (Håkanson, 2005; Li et al., 2023; Fu et al., 2024).

The increased VD observed in Lake Semayang suggests that the underlying recovery mechanism requires further investigation. Several hypotheses can be advanced based on the literature: (1) sediment flushing during extreme flood events, as documented in Amazon floodplain lakes (Räsänen et al., 1987; Armijos et al., 2020); (2) altered flow distribution following channel avulsion in the Pela River outlet, a process commonly observed in river-connected floodplain systems (Hamilton, 1999); or (3) differential compaction of organic-rich lakebed sediments, which can significantly alter basin morphology over decadal timescales (McGlue et al., 2015; Bertrand et al., 2024). Coggins and Anas (2019) demonstrated that sonar technology deployed on remotely operated vehicles (ROVs) can provide substantially higher spatial resolution bathymetry with significantly reduced profiling time, and has been applied to over 400 water bodies in Australia. Similar approaches are recommended for hypothesis validation in Semayang.

The case of Lake Semayang underscores a fundamental limitation of VD when used as a standalone indicator. As shown in Equation 1, VD is mathematically sensitive to changes in maximum depth. A sharp decline in Z_max—even if accompanied by net shallowing and volume loss—can artificially increase VD, as observed in Semayang. This mathematical caveat does not reflect ecological or hydrological recovery. Consequently, VD should never be interpreted in isolation. Reliable assessment of lake infilling and degradation requires a multi-parameter approach that integrates VD with other morphometric metrics, including volume change, maximum depth, surface area, and shoreline complexity (SDI), alongside direct bathymetric validation where available. This study demonstrates that while VD can serve as a useful early warning indicator, its application must be accompanied by complementary data to avoid misinterpretation—particularly in shallow tropical flood-pulsed lakes where rapid and spatially heterogeneous depth changes are common.

Lake Jempang’s morphometric stability (ΔVD = −0.08) suggests that this lake functions as a through-flow system with efficient sediment export, contrasting sharply with Melintang’s broader, shallower geometry that promotes sediment retention. Similar morphometric patterns have been reported for other Indonesian lakes, where basin morphology strongly influences sediment trapping efficiency and water quality dynamics (Muhtadi et al., 2024; Atmaja and Jung, 2025). This underscores the need for lake-specific rather than uniform watershed-based management strategies—a principle increasingly recognized in lake restoration literature (FutureLakes, 2026).

Several limitations warrant consideration in interpreting these findings:

First, the spatial resolution of GEBCO (15 arc-second, ∼450 m at the equator) cannot resolve fine-scale bathymetric features. This resolution is inherently mismatched with the scale of shallow, morphologically complex floodplain lakes. Consequently, the volume and depth estimates presented here should be viewed as minimum bounds, and actual volumetric changes may be larger. The use of GEBCO, while necessary for historical consistency, introduces uncertainty that we have quantified in Section 2.4. Comprehensive reviews emphasize that resolution and accuracy are crucial for constructing reliable bathymetric digital elevation models and highlight the potential and limitations of remote sensing for shallow bathymetry (Li et al., 2023; He et al., 2024). Environmental factors such as water turbidity, surface waves, and lakebed composition can cause signal attenuation, scattering, or refraction, leading to data inaccuracies (Gao, 2009; Ma et al., 2020). Our volume estimates therefore represent minimum bounds, and actual volumetric changes may be larger. A specific illustration of this bias is Lake Semayang, where GEBCO yields mean depths of only 0.12–0.14 m, while historical records indicate natural depths of 3–6.5 m (Muchlis et al., 2022). Hence, absolute depth and volume values from GEBCO are systematically underestimated and should be treated as relative indicators of shallowing direction rather than as precise measurements. The extreme decline in maximum depth for Lake Semayang (from 7.5 m to 1.5 m) lacks independent validation from field surveys or satellite altimetry (e.g., ICESat-2). Hence, all Z_max values presented here are preliminary, a finding consistent with documented large discrepancies between GEBCO and higher-resolution altimetry in shallow waters elsewhere (Giribabu et al., 2024). Thus, without independent validation, all depth and volume estimates in this study should be considered exploratory and subject to future confirmation.

Second, for the 1990–2000 period, we assumed constant Z_max due to the absence of historical bathymetric data. This assumption likely underestimates VD change if sedimentation was ongoing throughout the study period. As noted by Lo et al. (2022), bathymetric data limitations represent a common constraint in tropical lake studies, and many lakes in global databases (n = 76) lack published bathymetric maps.

Third, VD assumes a conical basin shape—an approximation that is weakest for complex floodplain lakes such as Semayang, which exhibit irregular morphology. Li et al. (2023) developed a hypsometric curve approach coupled with lake bottom elevation measurements to address this limitation, demonstrating that this method can be applied to ungauged lakes in extreme environments. The geometric cone model assumption has been shown to produce overestimations, particularly in convex waterbodies, which represent the majority of lake basins globally (Stachelek et al., 2022).

Fourth, the attribution of observed changes to specific drivers (coal traffic, land use, climate) remains qualitative. Development of integrated hydrodynamic-sediment models incorporating river discharge, sediment load, and land cover data is required (Liu et al., 2022; Nobert and Shibayama, 2006). Lo et al. (2022) identified that environmental and human factors such as watershed population density and tree cover loss correlate with short-term sediment accumulation rates.

Fifth, the VD increase observed in Lake Semayang may partially reflect a mathematical caveat rather than true ecological improvement. When maximum depth declines more rapidly than mean depth, VD can increase even under net infilling, highlighting the need for multi-parameter morphometric assessments. The influence of lake morphometry on biogeochemical dynamics has been documented in Amazon floodplain lakes (Brito et al., 2017), underscoring that even morphometric artifacts may have real ecological consequences.

Our findings have direct implications for the implementation of Presidential Regulation (60/2021) concerning the Rescue of National Priority Lakes. We propose a preliminary decadal ΔVD threshold of −0.2 as a potential intervention trigger, acknowledging that this value is derived from a limited dataset (three lakes) and has not been statistically validated. This threshold should be tested and refined with additional lake morphometric data before operational use. The −0.2 value was selected as a pragmatic boundary separating stable or recovering lakes from those undergoing critical infilling, aligning with similar threshold-based approaches in lake management frameworks (Istvánovics, 2009; Lo et al., 2022). Based on this criterion. ⁃

Lake Melintang (ΔVD = −0.42) qualifies for immediate prioritization within the national revitalization framework. With 80% volume loss in a single decade, interventions such as targeted dredging and erosion control in the upstream catchment cannot be delayed.

More broadly, this study demonstrates that low-cost geospatial monitoring using publicly available satellite data and accessible bathymetry can provide the quantitative baselines required for evidence-based lake management (Coggins and Anas, 2019; Li et al., 2023). We recommend that Indonesia’s Ministry of Environment and Forestry adopt VD tracking as a standard metric in the National Lake Monitoring Network.

The ecological implications of these morphometric changes extend beyond volume reduction. Sedimentation alters water quality, disrupts pelagic habitats, and shortens the lifespan of tropical floodplain lakes (Scheffer et al., 2001; Lo et al., 2022), while lake morphometry modulates carbon cycling and influences fish habitat niche segregation (Hayden et al., 2014; Casas-Ruiz et al., 2021; Evans, 2022). These pressures reflect broader Anthropocene syndromes in river systems globally (Meybeck, 2003), and are consistent with water vulnerability assessments that have identified similar threats in other regions (Vörösmarty et al., 2000). When aquatic environments shift to alternative states due to climate change or human engineering, ecosystem services may be substantially altered or even lost (Carpenter et al., 2011). The 80% volume loss observed in Lake Melintang thus signals serious threats to supporting services (fisheries, biodiversity) and provisioning services (water supply, irrigation), which are critical for local communities and the new capital city (Vörösmarty et al., 2000).

This morphometric assessment establishes the geometric framework for subsequent investigation of lake hydrodynamics and water quality. Planned research integrates.

Water isotopes and chemistry to trace sediment sources, adapting approaches developed in volcanic and catchment hydrology studies (Kendall and Caldwell, 1998; Carey and Quinton, 2005).

Collaboration with international researchers and adoption of cutting-edge technologies—as recommended in the review by Li et al. (2023) on multi-source data integration and advanced interpolation techniques—will substantially enrich our understanding of flood-pulsed tropical lake morpho-hydrological dynamics. Emerging approaches such as deep learning offer new possibilities for predicting lake bathymetry (Martinsen et al., 2023).