Assessing Flood Vulnerability of Landfills in Southern New Jersey: Incorporating Climate Change and Extreme Weather Impacts

Abstract

Southern New Jersey faces increasing flood risk due to several factors including rapid development, climate change, and aging infrastructure. This study evaluated the flood vulnerability of two municipal solid waste landfills located in Gloucester and Cumberland Counties. These sites are located near rural communities that rely on shallow groundwater for drinking water, which may be contaminated by floods. To assess these challenges, this research applies a hydrologic–hydraulic model to evaluate future flood vulnerability at the Cumberland County Improvement Authority (CCIA) landfill and the Gloucester County Solid Waste Complex (GCSWC) landfill. The method uses HEC-HMS and HEC-RAS 2D model simulations with climate-adjusted precipitation data derived from global climate models. Model performance was evaluated using Hurricane Ida (31 August–2 September 2021) by comparing HEC-RAS-simulated inundation extents with independently derived Sentinel-1 SAR flood maps generated in Google Earth Engine. Climate forcing was developed by deriving climate-adjusted 24 h precipitation–frequency (PF) design depths for 50-year and 100-year design storms under the Shared Socioeconomic Pathway (SSP) emissions pathways SSP2-4.5 (moderate) and SSP5-8.5 (high) for mid-century (2025–2050) and late-century (2070–2100) periods. These PF storm totals were converted to rainfall hyetographs using a fixed alternating variability method (AVM) temporal pattern within the coupled HEC-HMS/HEC-RAS modeling chain. Hazard amplification was primarily expressed through lateral inundation expansion and longer persistence of shallow flooding in low-relief operational zones, rather than uniform increases in peak depth across landfill interiors. Across both facilities, the landfill toe and adjacent access corridors were consistently identified as the most sensitive operational areas.

1. Introduction

Flooding is among the most destructive natural hazards, and its frequency and intensity are projected to rise with climate change, increasing exposure and damages across the United States [1]. Flood waters can also mobilize contaminants from waste facilities; empirical work shows many landfills lie in flood hazard zones and that flooding can generate leachate releases and erosion of waste, necessitating explicit risk screening for such sites [2,3]. In southern New Jersey, compound coastal and inland flood hazards are increasing due to sea level rise and changing tropical cyclone climatology, with pronounced mid-Atlantic impacts documented for late-century scenarios [4]. Reliance on shallow Coastal Plain aquifers and ongoing relative sea level rise or subsidence further heighten vulnerability of water supply and low-lying communities during floods [5]. Over the past two decades, the region has experienced severe storms (e.g., Hurricane Sandy in 2012 and Hurricane Ida in 2021) that produced widespread flooding, overwhelmed stormwater infrastructure, and disrupted critical services, underscoring systemic fragilities [6]. These impacts tend to be especially acute in rural and under-resourced places, where limited emergency capacity and uneven resilience planning exacerbate recovery challenges [7]. Flood-related landfill contamination is not unique to New Jersey. During the 2002 Elbe River flood (Germany), contaminant-laden sediments were widely mobilized; subsequent analyses documented organic toxicants and trace metals in post-flood sediments, while the geotechnical literature highlights how barrier performance and failure modes govern releases under thermal/hydraulic loads [8,9,10]. In the United States, the Norman (Oklahoma) municipal landfill sited on the Canadian River floodplain experienced flooding in 1986 that eroded protection, penetrated the clay cap, and exposed contents, with USGS documenting the resulting leachate plume and long-term groundwater risks [11,12]. The 2017 Hurricane Harvey event damaged the protective cap at the San Jacinto River Waste Pits Superfund site, exposing dioxin-bearing wastes and prompting immediate EPA-directed actions and a long-term removal remedy [13]. In New Jersey, Hurricane Irene (2011) was associated with benzene releases beyond protective barriers at the American Cyanamid Superfund site; since then, EPA has required 1-in-500-year flood-resilient designs, which performed effectively when Ida (2021) produced even higher water levels [14].

Although integrated hydrologic–hydraulic modeling is widely used for flood hazard assessment, it is rarely applied at the facility scale for landfill even though flood-driven leachate generation, erosional scouring, and barrier performance are site specific. Existing modeling efforts often privilege urban drainage systems, large basins, or ecosystem scale processes [15], and large-scale hazard products still face limitations for local calibration in ungauged or sparsely monitored catchments. For these reasons, this study’s contribution is a facility-scale, climate-informed flood hazard assessment for landfills in southern New Jersey, and a rationale for a modeling framework suited to limited monitoring data while directly addressing contaminant release pathways under future floods [15].

This study advances landfill-focused flood hazard assessment by implementing a coupled hydrologic–hydraulic modeling framework that is spatially explicit and incorporates monitoring points, functional landfill zones, and depth-based hazard classes. The framework extends prior modeling practice by integrating zone-specific interpretations of inundation, enhancing operational relevance, and aligning with spatial vulnerability principles emphasized in established vulnerability frameworks [16]. A further methodological contribution lies in adopting depth classification thresholds consistent with commonly applied flood hazard categories and infrastructure impact criteria, including flood depth grid products of FEMA [17] and shallow flooding classifications [18]. These depth categories also align with transportation impact studies demonstrating depth disruption relationships for roadway networks [19]. The study also contributes a data-driven validation strategy for hydraulic model inundation patterns through satellite-derived flood extent comparison, reflecting established practice in SAR-based flood evaluation [20]. In addition, the structured analysis of climate-adjusted design storms across baseline, mid-century, and late-century periods provides site-specific evidence of evolving flood extent, depth, and spatial connectivity under intensified rainfall forcing [21,22]. These results complement research showing projected increases in nonstationary and shifts in return periods for extreme precipitation in CMIP6 models [23]. Finally, the study offers operationally meaningful insights by interpreting inundation across functional landfill zones and identifying depth-sensitive hotspots such as landfill toe areas, thus linking hydrodynamic behavior directly to facility operations and long-term climate resilience planning.

The selected modeling framework supports the objective of simulating design storm flood hazards with outputs interpretable at the facility scale. The combined use of HEC-HMS and HEC-RAS, integrated with terrain data, is well established in regional floodplain mapping and hydrologic–hydraulic studies [24,25]. Within this workflow, HEC-HMS transforms design storm rainfall into watershed runoff hydrographs, which are subsequently routed through channel and floodplain geometries in HEC-RAS to produce spatially distributed flood depths and inundation extents [26,27]. HEC-HMS is particularly well suited for design storm runoff simulation, with model parameters representing watershed characteristics such as land cover, soils, and basin morphology, an advantage in ungauged basins where streamflow data for calibration are limited [24,25]. For hydraulic routing, HEC-RAS unsteady flow simulation is used to propagate the time-varying hydrographs generated by HEC-HMS through the river network, capturing water-surface dynamics governed by upstream discharge magnitude and timing, channel conveyance, and floodplain storage [28,29]. In addition to hydrograph boundary conditions, the study applies a precipitation boundary condition directly onto the 2D computational mesh in HEC-RAS, enabling rainfall or rainfall excess to be spatially distributed and routed across the terrain surface. This approach is supported by the HEC-RAS meteorological framework, which allows gridded or gage precipitation to drive overland flow routing [30].

Including a rainfall boundary enhances representation of pluvial flooding, particularly in low-relief floodplains where facilities situated near river corridors may experience combined riverine and pluvial inundation. This combined mechanism risk has been emphasized in hydrologic–hydraulic studies addressing compound flood drivers and localized runoff dynamics [30,31].

Several alternative hydrologic and integrated modeling systems were evaluated to ensure alignment between model capabilities and the study’s design storm focused objectives. The Soil and Water Assessment Tool (SWAT) is widely used for long-term, continuous watershed simulation and for evaluating hydrology and water quality responses to land management practices [32,33]. Because SWAT is designed around daily to multi-decadal water-balance processes rather than short-duration event simulation, it is less suited for repeated design storm applications required for facility scale flood mapping under baseline and mid- to late-century climate scenarios. MIKE SHE, a physically based, distributed, integrated hydrologic modeling system, represents surface and subsurface processes in a fully coupled manner [34]. While MIKE SHE is powerful, its extensive data requirements, calibration demands, and computational cost reduce its practicality for repeated design storm simulations over multiple climate horizons.

Given these considerations, the HEC-HMS / HEC-RAS modeling framework was selected because it is computationally efficient, transparent, and repeatable, while remaining well validated in regional flood studies [24,25,26,27]. This framework supports consistent comparative assessment across baseline, 2025–2050, and 2070–2100 design storm periods. Hydrologic model development was completed in a companion study described in the hydrologic model development chapter, which established the HEC-HMS configuration and generated the runoff hydrographs used in this analysis [35]. The present study integrates those HMS-derived hydrographs with HEC-RAS geometry and terrain to quantify flood depths and inundation extents resulting from both pluvial runoff accumulation and riverine overbank flooding under baseline, mid-century, and late-century climate-influenced conditions [36].

2. Study Area

2.1. Location of Landfills

The Cumberland County Improvement Authority landfill is in Cumberland County within the Maurice River watershed and is being illustrated in Figure 1. The peak elevation reaches 64.6 m above mean sea level. The CCIA landfill overlies the Kirkwood Cohansey Aquifer, a major potable water source for southern New Jersey that is highly susceptible to contamination because of strong groundwater and surface water connectivity. Construction of the CCIA landfill progressed in phases under NJDEP permitting prior to operation [37]. NJDEP records indicate that the facility was included in the Cumberland County Solid Waste Management Plan on 15 March 1984, received a permit to operate on 30 December 1985, and was to be constructed and operated in phases [38].

The Gloucester County Solid Waste Complex landfill, located in Gloucester County (see Figure 1), is situated at elevations ranging from approximately 30.5 to 45.7 m above sea level. However, its northern edge lies adjacent to low-elevation, flood-prone areas near Raccoon Creek. Terrain slope analysis conducted as part of this study showed surface drainage patterns oriented toward adjacent agricultural areas and water bodies. Construction and early facility development of the GCSWC are documented in NJDEP permit records from the mid- to late 1980s, including an Engineering Design Report dated June 1986, a revised Engineering Design Report dated May 1988, and an Operation and Maintenance Manual dated September 1986 [39].

2.2. Flood Hazard Mapping of Federal Emergency Management Agency

Both CCIA and GCSWC landfills are situated in areas identified by FEMA’s Preliminary Flood Insurance Rate Map (FIRM; see Figure 2) [40,41]. The surrounding zones include the 1% annual chance floodplain, commonly referred to as the 100-year floodplain, and the 0.2% annual chance floodplain, commonly referred to as the 500-year floodplain.

3. Methodological Approach

The approach combines multiple geospatial and hydrologic inputs to calculate the extent and flood depth associated with extreme weather conditions. The overall methodological approach is summarized in Figure 3.

3.1. Hydrologic–Hydraulic Model

An integrated hydrologic and hydraulic workflow was used to simulate flood depth and extent. The hydrologic foundation for the Maurice River and Raccoon Creek watersheds was established in the companion study [35]. The study utilized peak flow records to calibrate peak magnitude and timing. The model employed the Deficit and Constant loss method, Clark Unit Hydrograph transformation, Muskingum routing, and a simple monthly baseflow representation. Calibration performance was assessed using Root Mean Square Error (RMSE), Normalized Root Mean Square Error (NRMSE), Nash–Sutcliffe Efficiency (NSE), and peak to peak comparisons. Building on that foundation, the present study focuses on coupling the Hydrologic Engineering Center–Hydrologic Modeling System (HEC-HMS) inflow hydrographs with two-dimensional Hydrologic Engineering Center–River Analysis System (HEC-RAS) simulations for spatial inundation mapping and facility scale depth estimation. This study provides the details of hydraulic analysis and avoids model development steps already documented [35].

3.2. Hydrodynamic Modeling in Two-Dimensional HEC-RAS

The hydraulic simulations were designed to represent both fluvial and pluvial processes within a unified two-dimensional framework. Fluvial forcing was introduced through upstream inflow hydrographs generated by the calibrated HEC-HMS rainfall–runoff model described in the companion study. For each event, HEC-HMS produced an hourly discharge hydrograph at the outlet of the sub-basin draining toward the hydraulic domain. The hydrograph was imported directly into HEC-RAS as an unsteady upstream boundary, preserving its original timing and peak characteristics. This practice is consistent with established guidance for unsteady boundary specification in HEC-RAS and is also aligned with documented applications of 2D unsteady HEC-RAS inflow boundaries in peer-reviewed hydraulic studies [42,43].

Pluvial forcing was represented using the precipitation boundary condition within HEC-RAS, applying spatially distributed rainfall as an hourly hyetograph. Incremental precipitation was mapped directly to the 2D computational mesh during the unsteady solution process. The configuration supports simulation of rainfall-driven runoff, shallow overland flow, and depressional storage, consistent with HEC-RAS documentation.

A downstream unsteady boundary condition was assigned based on channel slope and control characteristics. Sensitivity testing confirmed that the boundary did not induce backwater artifacts or constrain drainage, consistent with best practices for 2D unsteady hydraulic modeling reported in the literature [43].

3.2.1. Terrain Processing

Model setup began in RAS Mapper by inserting a hydraulic terrain at approximately 3 m resolution. Before mesh generation, hill shade visualization was used to confirm that major drainage gradients and floodplain structure were physically reasonable. Flow direction grids were inspected to ensure that overland flow pathways were coherent and continuous. These checks reduce the likelihood of spurious sinks, artificial barriers, and other DEM artifacts that can distort drainage patterns or produce unrealistic ponding in two-dimensional simulations [44,45,46]. Only clearly nonphysical artifacts were corrected, while natural depressions were preserved because they influence rainfall-driven storage. The verified terrain was used consistently throughout the workflow to maintain a common topographic basis.

3.2.2. Two-Dimensional Computational Mesh Design and Local Refinement

A two-dimensional computational domain was defined to include the active floodplain, overbank storage areas, shallow interfluve pathways, and the CCIA and GCSWC landfill facilities. A base mesh resolution of 10 m was selected to maintain computational feasibility at watershed scale while preserving dominant topographic controls, consistent with recommended practices for two-dimensional hydraulic modeling [43]. Mesh refinement to 5 m was applied in floodplain corridors and around the landfill facilities where higher spatial detail was required to represent localized topographic variability. Peer-reviewed hydrodynamic analyses show that finer mesh and DEM resolutions substantially improve representation of flow paths, velocities, and depth patterns, particularly in areas with complex terrain [47].

Facility footprints and adjacent operational areas were represented by multiple computational cells to capture elevation variability and local flow pathways within the site. Using multiple cells reduces discretization bias and produces more stable and realistic depth estimates in decision relevant zones, a finding consistent with recent high-resolution hydraulic modeling research [47]. Break lines were incorporated along channels, road embankments, and facility perimeter features to preserve hydraulic controls and prevent mesh smoothing from distorting key flow pathways, following guidance in the HEC-RAS User’s Manual [43].

3.2.3. Roughness and Infiltration Parameterization

Spatially variable Manning’s roughness coefficients were assigned based on land-cover classes from the National Land Cover Database (NLCD), which provides 30 m land-cover products produced by the Multi-Resolution Land Characteristics Consortium (MRLC). Manning’s n values were selected within commonly used ranges, with developed surfaces typically between 0.02 and 0.04 and vegetated floodplain areas between 0.05 and 0.15. Infiltration losses were represented using a curve number-based rainfall-loss method. Curve numbers were derived from NLCD land cover and NRCS hydrologic soil groups following standard SCS guidance, which serves as the foundational methodology for CN runoff estimation [48]. The resulting curve-number layer was applied through the HEC-RAS precipitation-loss configuration so that rainfall was partitioned into infiltration and excess runoff in a spatially distributed manner. This approach is appropriate for event-based rainfall-driven flooding in low-relief terrain, where contrasts between impervious, compacted, and vegetated surfaces strongly influence runoff generation and localized ponding. Numerous studies emphasize the sensitivity of flood inundation modeling to surface characteristics and runoff generation dynamics under varying land cover and hydrologic conditions [49,50].

3.2.4. Numerical Framework for Two-Dimensional Hydrodynamic Simulation and Output Generation

Water surface elevations and depth-averaged velocities were computed across the 2D domain using HEC-RAS. The model was run in full two-dimensional unsteady flow mode, allowing temporal changes in water surface elevation and velocity fields to be resolved across the computational mesh, particularly during rapidly varying storm-driven conditions. A 5 s computational time step was selected to maintain numerical stability during intense rainfall and rapidly varying flow transitions, especially in refined mesh regions. Outputs were stored at hourly intervals to match the temporal resolution of hydrologic forcing. For each computational cell, the model computed water surface elevation, depth, depth-averaged velocity magnitude, and maximum depth, consistent with standard HEC-RAS 2D hydrodynamic output variables [43]. Flood depth was calculated as the difference between simulated water surface elevation and terrain elevation. Depth grids and time series were clipped to the CCIA and GCSWC landfill footprints and adjacent operational areas to generate facility-scale flood-exposure metrics.

3.3. Hydraulic Model Validation Against SAR-Derived Flood Extents

Validation of the modeled flood extent was performed using observations from Hurricane Ida, which produced intense rainfall and widespread flooding in southern New Jersey from 31 August to 2 September 2021. The validation simulation used upstream inflow hydrographs generated by the calibrated HEC-HMS model together with event-specific rainfall from Ida, allowing both watershed runoff and direct rainfall contributions to be represented. Modeled flood extent was obtained from the HEC-RAS water-depth raster, which was converted from feet to meters before thresholding and comparison. Observed flood extents were mapped using Sentinel-1 Synthetic Aperture Radar imagery processed in Google Earth Engine. GRD scenes were filtered to Interferometric Wide mode and VV polarization, and both ascending and descending passes were included to maximize availability. A multi-week pre-event baseline from 15 July to 30 August 2021 was summarized using the pixel-wise median to reduce noise and speckle. Post-event conditions were represented using the 5 September 2021 acquisition. Both the baseline and post-event composites were smoothed using a 1-pixel focal mean filter following established SAR flood mapping practices [51]. This workflow is also consistent with algorithms that leverage multi-temporal Sentinel-1 statistics and Google Earth Engine for rapid flood detection [52].

Flooded pixels were identified using a change detection ratio between pre-event and post-event backscatter. A ratio threshold of 1.10 was selected following sensitivity testing within ranges commonly used in SAR flood mapping algorithms [51,52]. A connected pixel filter requiring at least 8 connected pixels was applied to reduce speckle-related artifacts and to remove isolated misclassifications, consistent with standard SAR post-processing workflows [53]. To compare modeled and observed flood extents, a set of minimum depth thresholds ranging from 0.30 m to 0.91 m was applied to the HEC-RAS depth raster. This range brackets typical shallow inundation thresholds used in hydraulic model comparison studies and includes moderate depths relevant for floodplain hazard interpretation. For each candidate threshold value, the HEC-RAS raster was converted to a binary inundation map and compared with the SAR mask to compute precision, recall, and the F1 score. The optimal threshold t* was defined as the value that produced the highest F1 score. This sensitivity-based tuning approach follows established practices in hydraulic model evaluation where SAR-derived flood extents serve as independent reference data [54,55].

Agreement between modeled and observed flood extents was quantified using area-based confusion accounting. True positive, false positive, false negative, and true negative areas were computed by summing pixel areas. Precision and recall were calculated using standard formulations, where Equation (1) defines precision as the ratio of correctly detected flooded area to all area classified as flooded, and Equation (2) defines recall as the ratio of correctly detected flooded area to the total observed flooded area. The F1 score, the harmonic mean of precision and recall, was then used to summarize overall agreement.

$$\mathrm{Precision}={\displaystyle \frac{{\mathit{A}}_{\mathit{T}\mathit{P}}}{{\mathit{A}}_{\mathit{T}\mathit{P}}+{\mathit{A}}_{\mathit{F}\mathit{P}}}}$$

(1)

$$\mathrm{Recall}={\displaystyle \frac{{\mathit{A}}_{\mathit{T}\mathit{P}}}{{\mathit{A}}_{\mathit{T}\mathit{P}}+{\mathit{A}}_{\mathit{F}\mathit{N}}}}$$

(2)

where ${\mathit{A}}_{\mathit{T}\mathit{P}}$ is the area correctly identified as flooded by both SAR and HEC, ${\mathit{A}}_{\mathit{F}\mathit{P}}$ is the area flagged as flooded by SAR but not by HEC, and ${\mathit{A}}_{\mathit{F}\mathit{N}}$ is the area flooded in HEC but missed by SAR. F1 scores were emphasized because overall accuracy can be inflated when non-flooded area dominates, which is a common characteristic of flood extent comparisons [51].

3.4. Climate Forcing and Scenario Development

3.4.1. Climate Scenario Selection and GCM Preprocessing

To incorporate climate change impacts into design storm estimation, two Shared Socioeconomic Pathways were evaluated, including SSP2 4.5 (intermediate emissions) and SSP5 8.5 (high emissions). These scenarios were analyzed over a mid-century window (2025 to 2050) and a late-century window (2070 to 2100) to represent near-term and long-term changes in precipitation extremes relevant to infrastructure planning [56,57]. The selection of these two planning horizons follows the standard time windows commonly used in major climate impact assessments, which supports their validity for long-range engineering and adaptation analyses [58]. Daily precipitation projections were obtained from ACCESS-CM2 (CMIP6) accessed via the Earth System Grid Federation. The Global Climate Model (GCM) precipitation variable pr (kg m⁻² s⁻¹, equivalent to mm s⁻¹) was converted to daily depth by multiplying by 86,400 s. Negative artifacts were removed during preprocessing. ACCESS-CM2 was selected based on its historical performance, evaluated through comparison of modeled historical precipitation with the observational dataset used as the historical reference in the study domain, a procedure consistent with recommended practice for CMIP6 model evaluation [58].

Given that raw GCM precipitation often departs systematically from observed climatology, a bias correction step was necessary to ensure that the adjusted projections were suitable for use in design storm estimation. Quantile mapping was therefore applied using Daymet [59], a gridded daily surface meteorology dataset derived from ground-based station observations and spatial interpolation across North America. Quantile-mapping bias correction is widely used to reduce systematic GCM precipitation errors and is supported by extensive evaluation across multiple statistical approaches [59,60].

3.4.2. Delta Factor Climate Adjustment

Because daily precipitation simulated by global climate models can exhibit systematic biases in magnitude and occurrence at local scales, projected changes were incorporated using a delta factor (change factor) approach rather than using absolute model precipitation values directly [61]. In a delta approach, the climate change signal is expressed as a ratio between model simulated future and historical conditions and then transferred onto an observational baseline used for design storm estimation [58]. In this study, delta factors were estimated from the bias-corrected ACCESS-CM2 daily precipitation series for each emissions pathway and each planning window $\{2025 \mathrm{to} 2050, 2070 \mathrm{to} 2100\}$ [56]. A baseline precipitation–frequency relationship was first developed from the observational record. Let $D$ denote duration [h] and $T$ denote return period [years]. Here, $D=24$ h. An annual maximum series (AMS) was constructed by extracting the largest 24 h precipitation total in each year, and a GEV distribution was fitted to the AMS [62,63]. The baseline design depth $h_{\left(D,T\right)}^{base}$ [mm] was computed as Equation (3):

$$h_{(D,T)}^{\mathit{base}}=F_{\mathit{GEV}}^{-1},(1-\frac{1}{T})$$

(3)

where $F_{GEV}$ is the fitted cumulative distribution function of the annual maxima and $F_{GEV}^{-1}$ is its inverse (quantile) function. To define days with measurable rainfall and to avoid screening and quantile estimates being influenced by zero or trace precipitation days, a wet day threshold of $P_0=0.10$ in/day [2.54 mm/day] was applied consistently to the observational record and to the bias-corrected ACCESS-CM2 series. This threshold is consistent with stormwater practice that treats rainfall events greater than 0.1 in as qualifying events. Let $P_t$ denote daily precipitation on day $t$ [mm]. Wet days were defined as those satisfying $P_t\ge P_0$, and the wet day sample for any period was $\left\{P_t:P_t\ge P_0\right\}$.

Return period-specific scaling was selected because changes in extreme precipitation can be quantile dependent and may vary with rarity [62,63]. Let $Q_{\left(D,T\right)}^{\left(hist,BC\right)}$ denote the bias-corrected ACCESS-CM2 historical return level for duration $D$ and return period $T$ [mm], estimated from the historical reference window using the same AMS and GEV framework, and let $Q_{\left(D,T,s,w\right)}^{\left(fut,BC\right)}$ denote the corresponding bias-corrected future return level [mm] for scenario $s$ and window $w$. The return period-specific delta factor was defined as Equation (4):

$$\begin{array}{cccc}& {\Delta }_{\left(D,T,s,w\right)}={\displaystyle \frac{Q_{\left(D,T,s,w\right)}^{\left(fut,BC\right)}}{Q_{\left(D,T\right)}^{\left(hist,BC\right)}}}& & \end{array}$$

(4)

where ${\Delta }_{\left(D,T,s,w\right)}$ is dimensionless. The climate-adjusted 24 h design depth was then obtained by scaling the observational baseline design depth using Equation (5):

$$\begin{array}{cccc}& h_{\left(D,T,s,w\right)}^{fut}={\Delta }_{\left(D,T,s,w\right)} h_{\left(D,T\right)}^{base}& & \end{array}$$

(5)

where $h_{\left(D,T,s,w\right)}^{fut}$ is the future design storm depth [mm] for duration $D$ and return period $T$. This change factor formulation preserves the baseline return period definition while modifying storm magnitudes in proportion to modeled changes in extreme precipitation, and it retains the observational baseline as the local anchor for design storm estimation. Design Storm-specific [30,64] delta factors ${\Delta }_{\left(D,T,s,w\right)}$ for both CCIA and GCSWC are summarized in

Table 1. Return period-specific delta factors for IDF rainfall depths at GCSWC and CCIA landfill.

	CCIA				GCSWC
Return Period	SSP2 4.5 Mid-Century	SSP2 4.5 Late Century	SSP5 8.5 Mid-Century	SSP5 8.5 Late Century	SSP2 4.5 Mid-Century	SSP2 4.5 Late Century	SSP5 8.5 Mid-Century	SSP5 8.5 Late Century
T2	1.268722	1.405286	1.277533	1.581498	1.278261	1.413043	1.286957	1.591304
T5	1.272446	1.405573	1.278638	1.585139	1.272727	1.409091	1.281818	1.587879
T10	1.270073	1.403893	1.277372	1.581509	1.277778	1.413043	1.285024	1.589372
T25	1.271095	1.40754	1.280072	1.585278	1.276364	1.410909	1.283636	1.589091
T50	1.269615	1.405136	1.278174	1.582026	1.274742	1.409158	1.282127	1.587888
T100	1.271591	1.405682	1.279545	1.584091	1.27503	1.410133	1.282268	1.588661

. For CCIA, mid-century (2025 to 2050) delta factors range from approximately 1.27 to 1.28 under SSP2-4.5 and remain close to 1.28 under SSP5-8.5 across return periods $T_2$ through $T_{100}$. By late century (2070 to 2100), the delta factors increase to about 1.40 to 1.41 under SSP2-4.5 and to about 1.58 to 1.59 under SSP5-8.5, reflecting the stronger forcing under the higher emissions pathway. For GCSWC, the same pattern is observed. During the mid-century window, delta factors fall between 1.27 and 1.29, while late-century factors reach approximately 1.41 under SSP2-4.5 and about 1.59 under SSP5-8.5.

Across both landfills, differences across return periods remain small within each emissions scenario and each planning window, indicating near-uniform scaling of baseline 24 h IDF depths. This behavior is consistent with CMIP6 SSP scenario framing and with guidance on updating IDF relationships under climate change [56,58]. These delta factors quantify the modeled relative change in 24 h extreme rainfall for each return period and provide the scaling terms used to update baseline design storm depths.

Table 2. Baseline and climate-adjusted 24 h design rainfall depths for 2-, 5-, 10-, 25-, 50-, and 100-year design storm under SSP2-4.5 and SSP5-8.5 for mid- and late-century at CCIA.

Scenario	T2 (mm)	T5 (mm)	T10 (mm)	T25 (mm)	T50 (mm)	T100 (mm)
Baseline (HIST 1985-2020)	57.658	82.042	104.394	141.478	178.054	223.52
SSP2-4.5 Mid (2025–2050)	73.152	104.394	132.588	179.832	226.06	284.226
SSP2-4.5 Late (2070–2100)	81.026	115.316	146.558	199.136	250.19	314.198
SSP5-8.5 Mid (2025–2050)	73.66	104.902	133.35	181.102	227.584	286.004
SSP5-8.5 Late (2070–2100)	91.186	130.048	165.1	224.282	281.686	354.076

and

Table 3. Baseline and climate-adjusted 24 h design rainfall depths for 2-, 5-, 10-, 25-, 50-, and 100-year design storm under SSP2-4.5 and SSP5-8.5 for mid- and late-century at GCSWC.

Scenario	T2 (mm)	T5 (mm)	T10 (mm)	T25 (mm)	T50 (mm)	T100 (mm)
Baseline (HIST 1985–2020)	58.42	83.82	105.156	139.7	171.958	210.566
SSP2-4.5 Mid (2025–2050)	74.676	106.68	134.366	178.308	219.202	268.478
SSP2-4.5 Late (2070–2100)	82.55	118.11	148.59	197.104	242.316	296.926
SSP5-8.5 Mid (2025–2050)	75.184	107.442	135.128	179.324	220.472	270.002
SSP5-8.5 Late (2070–2100)	92.964	133.096	167.132	221.996	273.05	334.518

then apply the corresponding return period-specific delta factors to the observed baseline 24 h design depths to obtain climate-adjusted design rainfall depths for mid-century and late century under SSP2 4.5 and SSP5 8.5. Together the delta factor table and the design depth tables provide a transparent link between the modeled change signal and the final climate-adjusted IDF values used in the subsequent analyses

3.5. Temporal Structure of Design Storms and Application of the Alternating Variability Method

Design storms were constructed for a duration D = 24 h. The analysis evaluates the T = 50-year and T = 100-year design storm as these design levels are widely used in engineering flood hazard assessment and facility planning to represent moderate- and high-hazard conditions and to support consistent scenario comparison [65,66]. For each scenario and planning window, the total 24 h design depth $h_{\left(D,T\right)}$ [mm] reported in Table 2 (Design rainfall depths for the location at CCIA landfill) and Table 3 (Design rainfall depths for the location at GCSWC landfill) was converted to an hourly rainfall hyetograph using the alternating variability method, a variant of the alternating block procedure commonly used in design hydrology [67,68]. The temporal discretization used a uniform simulation time step of Δt = 1 h, yielding 24 incremental rainfall values for the 24 h storm. This hourly structure was selected to match the temporal resolution used to apply rainfall forcing in the hydrologic and hydraulic simulations.

The alternating variability method applies to an alternating block ordering, in which rainfall increments are arranged to concentrate intensity near the storm center and distribute remaining increments outward in a structured, balanced form. In this study, the largest incremental depth was placed near the storm mid-point and the remaining increments were arranged in descending order by alternation around the peak, consistent with the frequency storm and alternating block procedure described in HEC-HMS guidance and in external hydrology references illustrating ABM construction [67]. This ordering yields a standardized nested temporal structure in which shorter duration intensities are embedded within the 24 h storm, supporting design comparison under a consistent storm pattern [30]. The incremental rainfall depth assigned to hour i was computed as:

$$p_i=f_i\cdot h_{\left(D,T\right)}$$

where $p_i$ is the rainfall depth during hour i [mm], $f_i$ is the dimensionless fraction applied to hour i, and ${\sum }_{i=1}^{24}{f}_i=1$. The same fraction pattern was retained across baseline, mid-century, and late-century conditions so that differences among scenarios reflect changes in storm magnitude rather than changes in storm temporal structure. These hyetographs define the rainfall forcing patterns applied consistently across all simulations. Figure 4 illustrates the alternating variability method-derived 24 h design storm hyetographs for CCIA under baseline, mid-century, and late-century climate scenarios for the 50-year and 100-year storms.

Figure 5 illustrates the alternating variability method-derived 24 h design storm hyetographs for GCSWC under baseline, mid-century, and late-century climate scenarios for the 50-year and 100-year storms.

In this study, uncertainty in the climate-adjusted design depths is addressed through scenario choice, planning horizon, and climate model considerations. Scenario uncertainty is represented by reporting results for both SSP2 4.5 and SSP5 8.5, which reflect contrasting forcing pathways used in CMIP6-based assessments [56,58]. Planning horizon uncertainty is represented by evaluating mid-century (2025 to 2050) and late-century (2070 to 2100) conditions, recognizing that climate change signals generally strengthen with time, while internal climate variability can influence extreme rainfall estimates within finite windows [69]. Climate model structural uncertainty is acknowledged, but it was reduced through historical performance screening and bias correction. Bias correction using quantile mapping is widely applied in hydrologic climate impact studies to reduce systematic precipitation bias and improve agreement with observed distributions [70,71]. Model screening based on historical skill is also commonly used to support model selection when time and computational constraints limit large ensembles [72]. After applying the same preprocessing and quantile mapping procedure to each model, the alternative CMIP6 models evaluated produced future period peak daily precipitation intensities that differed from ACCESS-CM2 by approximately 6.1% and 11.1% at the CCIA site and 4.2% and 7.6% at the GCSWC site, while the three models differed from ACCESS-CM2 by approximately 5.7% and 3.9%, respectively. These differences, computed within this study, indicate modest inter-model spread for the study locations and support use of the selected model as a defensible basis for scenario and horizon comparisons, while recognizing that some structural uncertainty remains.

4. Results and Discussion

4.1. Spatial Framework for Hydrodynamic Evaluation: Monitoring Points, Functional Zones, and Depth Classification

Flood depth was obtained as the difference between the simulated water surface elevation and the underlying terrain elevation. Flood depths for the CCIA and GCSWC landfills were evaluated for two hazard levels: the 50-year design storm and 100-year design storm events following standard hydrologic frequency-analysis conventions commonly applied in hydraulic modeling and flood-risk assessment studies [65].

Although both SSP2-4.5 and SSP5-8.5 scenarios were used in the design storm development, the results section of this paper focuses primarily on SSP5-8.5 as this pathway produces the strongest forcing signal and therefore provides a clearer upper bound assessment of future hazard conditions. This approach is widely used in climate impact studies, where the high emissions pathway serves as a stress test scenario to reveal the maximum plausible amplification of extreme rainfall and flood depth risk in long-term planning contexts [58]. Each facility was evaluated by subdividing the flood analysis extent into three functionally relevant assessment zones: (i) Main Landfill Cell, (ii) Administrative and Operational Support Area, and (iii) Nearby Transportation and Access Area (see Figure 6). This zoning was used to interpret modeled flood hazards in terms of operational function and exposure rather than relying on a single facility-wide summary.

Representative monitoring points were selected for each facility (see

Table 4. Monitoring points at CCIA and GCSWC landfills.

Landfill	Broad Category	Representing Area	Latitude (WGS84)	Longitude (WGS84)
CCIA	Administrative and operational support area	Administrative area	39.44992	75.0965
		Parking area	39.45057	75.0953
	Main landfill cell	Landfill toe area	39.44756	75.1018
		Center of landfill	39.45179	75.1006
	Nearby transportation and access area	Roads	39.45331	75.0926
GCSWC	Administrative and operational support area	Administrative area	39.71256	75.2822
		Parking area	39.71146	75.2812
	Main landfill cell	Main landfill cell	39.71329	75.2855
		Landfill toe area	39.7078	75.2847
	Nearby transportation and access area	Roads	39.71526	75.2802

) and grouped into broad operational zones to support a consistent comparison of modeled flood depths. Specifically, the main landfill cell zone includes points within the primary disposal footprint and along the landfill toe; the administrative and operational support area includes points located within administrative and parking areas; and the nearby transportation and access area includes points situated along access roads. Structuring the results across these site-specific zones improves the interpretability of modeled flood depths by linking hydraulic conditions to operationally meaningful facility areas. This approach aligns with established guidance emphasizing that results are more decision-relevant when organized according to exposure-relevant spatial units rather than aggregated across an entire site [73]. It also reflects spatial analysis frameworks which underscore the importance of evaluating spatial heterogeneity within a system to better distinguish differences across sub-areas where functional roles and potential flooding impacts vary [74].

This depth-focused categorization is also consistent with broader flood hazard assessment practice reviewed in the recent literature, where discrete depth classes are commonly used to communicate hazard severity and to support comparison across methods and scenarios [75]. In this study, flood depths were grouped into four classification bands: very shallow (0–0.30 m), shallow (>0.30–0.91 m), moderate (>0.91–1.83 m), and deep (≥1.83 m). These thresholds align with long-standing mapping practices for shallow and ponding-type flood conditions [17,18].

4.2. Findings of Hydraulic Model Validation Against SAR-Derived Flood Extents

Under the CCIA landfill domain, the depth cutoff sensitivity analysis indicates that the strongest overall agreement occurs at t = 0.31 m. At this cutoff, precision = 0.9456, recall = 0.5708, and F1 = 0.7119, indicating that the flood map is reliable within this depth threshold range, with relatively few false positives and moderate completeness relative to the modeled flood extent [76]. The comparison between the SAR-derived flood extent and the HEC-RAS flood extent at t = 0.31 m for the CCIA domain is illustrated in Figure 7, which also shows how precision, recall, and F1 score vary as functions of the applied depth threshold.

This pattern suggests that agreement is maximized when shallow flood depth values are retained in the binary HEC-RAS flood mask rather than applying a more restrictive minimum depth threshold. Similar sensitivity to minimum depth cutoffs has been reported in flood extent comparison studies where small thresholds, including those on the order of 0.05 m, are commonly tested and can meaningfully influence mapped extents and overlap-based performance metrics [76]. However, when the cutoff is increased to t = 0.6 m, precision remains high at about 0.86, while recall and F1 decrease into a moderate range. This behavior is consistent with expectations because a higher minimum depth threshold removes shallow floodplain wetting and small hydraulically disconnected low-depth areas. Studies have shown that excluding these shallow regions reduces spatial overlap with independently derived flood maps and increases omission errors, particularly for SAR-based flood delineation where shallow water commonly contributes to the observed flood extent. Removing these areas reduces mapped extent alignment and increases omission relative to the SAR water extent. As a result, recall and F1 decline even when precision remains high at t = 0.91 m, recall becomes low and F1 decreases substantially, indicating that a near 1 m depth requirement is too restrictive for extent comparison for this event and location. This decline suggests that a substantial portion of the flood extent around the CCIA domain is shallow to moderate in depth, and excluding these depths removes many true flooded pixels.

This is also consistent with the conceptual difference between the two products. SAR flood mapping primarily delineates exposed surface water at the satellite overpass time, and its performance can be reduced in urban and vegetated settings due to complex backscatter behavior and partial visibility. In contrast, the hydraulic model represents depth wetting that can remain well below 1 m across broad areas [20]. To harmonize the two products and develop agreement metrics, permanent and recurrent water bodies were removed from both datasets using a long-term surface water occurrence mask derived from Landsat observations. Under the GCSWC landfill domain, the depth cutoff sensitivity analysis indicates that the strongest overall agreement occurs at t = 0.31 m. At this cutoff, precision = 0.8793, recall = 0.54048, and F1 = 0.6543, indicating reliable mapped inundation with moderate completeness relative to the modeled inundation extent. The GCSWC threshold sensitivity results and the selected cutoff are summarized in Figure 8, which reports precision, recall, and F1 as functions of the depth threshold.

The moderate recall at the best cutoff can be interpreted as a combination of SAR omission mechanisms and event sampling limitations. SAR inundation mapping is known to under-detect flooding in vegetated flood plains, rough surfaces, and mixed pixels where open water signatures are partially obscured, which increases omission error relative to hydraulic model extents. In addition, the post-event composite was based on a single available Sentinel-1 acquisition, which might have reduced the stability of change detection and missed detections in marginal areas. This is also a practical limitation of event-based SAR mapping because suitable acquisitions are not always available immediately after a major event and the closest overpass can occur after peak inundation. These limitations are widely discussed in Sentinel-1 flood mapping and model support studies [52,55].

The objective of this validation step was not to achieve perfect pixel-level agreement across all depth thresholds but to identify a reasonable minimum depth cutoff for converting the modeled depth field into a binary flood mask suitable for comparison with a SAR-based water extent product [77,78]. In addition, this approach was adopted because no streamflow gauges exist within the study region to validate hydrodynamic model outputs, making satellite-based flood extent comparison a practical event-specific evaluation method under data-scarce conditions [79,80]. Within this context, F1 = 0.71 for the CCIA domain and F1 = 0.65 for the GCSWC domain, together with precision values of 0.95 and 0.88 respectively indicate useful event-scale agreement for footprint evaluation.

4.3. Climate-Adjusted Design Storms for Flood Exposure Assessment

The modeled flood responses under baseline and climate-adjusted design storms are presented in this section. It summarizes how projected inundation translates into operationally meaningful hazard levels across key functional areas of the CCIA and GCSWC landfills. Results are structured to first establish spatial evidence of inundation expansion and then translate these patterns into area-based operational comparisons and finally synthesize outcomes into hazard categories for near-term, mid-century, and late-century planning under SSP2-4.5 and SSP5-8.5.

To visually demonstrate the spatial expansion of flood hazard, HEC-RAS inundation outputs are compiled into a three-panel comparison for each landfill. The sequence focuses on the most decision-relevant contrast: Baseline vs. SSP5-8.5 mid-century vs. SSP5-8.5 late-century. This layout highlights how the high-emissions pathway progressively expands inundation into operational areas that remain only minimally affected under historical design conditions. For the 50-year event (see Figure 9), baseline inundation appears localized and fragmented, characterized by discrete shallow patches distributed along low-gradient areas. Under SSP5-8.5 mid-century forcing, these patches expand laterally and begin to coalesce, producing a more continuous inundation footprint. Late-century conditions further amplify this pattern with inundated areas becoming both wider and more spatially connected, indicating an increased likelihood of disruption across functionally relevant low-elevation zones.

The 100-year event (see Figure 10) exhibits a similar but more pronounced evolution. Even under baseline conditions, inundation covers a broader area than in the 50-year case with fewer isolated patches. Transitioning to SSP5-8.5 mid-century conditions, the inundation footprint expands further and becomes increasingly continuous, reflecting enhanced flood connectivity along low-gradient corridors. Under late-century forcing, inundation reaches its maximum spatial extent, forming large and contiguous flooded areas with minimal fragmentation. Compared to the baseline, this represents a clear intensification of flood extent and persistence rather than a fundamental shift in the spatial pattern of flooding.

Like the CCIA results, the dominant climate signal is reflected through changes in inundation extent and spatial continuity rather than a fundamental redistribution of flooding into previously unaffected interior areas. For the 50-year event (see Figure 11), baseline conditions show inundation that is comparatively localized and patchy with shallow to moderate flooding occurring in discrete low-gradient areas. Under SSP5-8.5 mid-century forcing. These inundated patches expand outward and begin to merge, producing a more laterally continuous flood footprint. Late-century conditions further intensify this response, with inundation becoming more spatially connected and occupying a larger contiguous area.

The 100-year event (see Figure 12) exhibits a similar progression but with a more extensive baseline footprint. Compared to the 50-year case, baseline inundation already shows greater spatial reach and reduced fragmentation. Under SSP5-8.5 mid-century conditions, inundation expands further and displays increased continuity across adjacent low-lying areas. Late-century forcing produces the most extensive and connected inundation pattern with large continuous flooded areas and minimal patch separation, indicating enhanced persistence and spatial coherence of flooding under extreme forcing.

4.4. Inundated Area Growth Across Baseline, Mid-Century, and Late-Century Periods Around the Landfill Area

Accordingly, the analysis consolidates zone-wise inundation depths across key functional locations for both landfills covering the main landfill cell, landfill toe zone, roads administrative facilities, and parking areas so that the mapped inundation patterns can be translated into operationally meaningful indicators of flood exposure relevant to access, management, and continuity of site operations.

Table 5. Total inundated area (km²) by functional zone for CCIA and GCSWC under 50-year and 100-year design storms across baseline (1980–2020), mid-century (2025–2050), and late-century (2070–2100) periods (SSP5-8.5).

Landfill	Zone	Total Inundated Area (km2) by Functional Zone for CCIA and GCSWC
		50-Year Design Storm			100-Year Design Storm
		Baseline	Mid- Century	Late- Century	Baseline	Mid- Century	Late- Century
CCIA	Administrative area	0.0144	0.0148	0.0152	0.0148	0.0151	0.0154
	Parking area	0.0113	0.012	0.0129	0.012	0.0129	0.0137
	Landfill toe area	0.0588	0.0689	0.0767	0.0682	0.077	0.0843
	Center of landfill	0.0095	0.0101	0.0106	0.01	0.0106	0.0116
	Roads	0.0279	0.0301	0.0322	0.03	0.0323	0.0344
GCSWC	Administrative area	0.014	0.0151	0.0157	0.0147	0.0156	0.0164
	Parking area	0.0089	0.0095	0.0102	0.0095	0.0101	0.0106
	Landfill toe area	0.0327	0.037	0.0402	0.0363	0.04	0.0434
	Center of landfill	0.0653	0.071	0.0771	0.0701	0.0763	0.0854
	Roads	0.0693	0.0754	0.081	0.0743	0.0807	0.0872

reports the total area experiencing any positive flood depth (depth > 0 m), representing the sum of all four depth classes for the two landfills’ surrounding area.

Table 5 reports the total inundated area (km²) by functional zone for CCIA and GCSWC for the 50-year and 100-year design storms across the baseline (1980–2020), mid-century (2025–2050), and late-century (2070–2100) periods under SSP5-8.5. Across both sites, the inundated area increases consistently from baseline to late century in every zone, and the 100-year event produces larger inundated footprints than the 50-year event. At CCIA, the largest absolute inundation occurs in the landfill toe area for both return periods, increasing from 0.0588 km² in the baseline 50-year case to 0.0843 km² in the late-century 100-year case. At GCSWC, inundation is largest in the roads and the center of the landfill, with late-century 100-year footprints of 0.0872 km² and 0.0854 km², respectively.

When the inundated footprints are evaluated in the context of the total mapped landfill areas, which are approximately 1.229 km² for CCIA and 1.066 km² for GCSWC, the extent of climate-driven expansion becomes more evident. At CCIA, the most substantial changes occur within the landfill toe area, which consistently represents the largest share of the inundated extent and shows the strongest growth across all time periods and return period events. At GCSWC, the largest increases appear in the road network, where late-century, inundated footprints exceed eight percent of the total mapped domain. Although any degree of flooding within waste management facilities is operationally and environmentally undesirable, particularly given the importance of maintaining containment integrity and access routes, the results highlight that the specific zones of greatest sensitivity vary between the two landfills.

Taken together, the depth-class results clarify not only how much of each landfill becomes inundated but also how that inundation is distributed across severity levels, revealing which operational zones are most susceptible to deeper water and which remain dominated by shallow ponding. These patterns establish a clearer understanding of hazard intensity across CCIA and GCSWC and highlight where climate-driven changes translate into meaningful shifts in flood behavior. With this depth-based perspective in place, it becomes important to examine the maximum flood depths experienced within each zone, since peak depth often determines the upper limit of structural loading, access disruption, and operational risk. Thus far, the depth class analysis has shown that the landfill toe consistently emerges as the most depth-sensitive hotspot, while the central landfill areas at both sites remain largely confined to shallow, ponding-type inundation. The following section therefore builds on the percentage-based depth distributions by evaluating how the maximum point-based depths respond to different design storm scenarios, providing a final layer of insight before transitioning to the scenario comparison framework.

4.5. Depth Class Redistribution Across Landfill Operational Zones Under Future Climate Scenarios

Percentage composition across the four standardized depth classes 0 to 0.30 m, >0.30 to 0.91 m, >0.91 to 1.83 m, and ≥1.83 m was calculated for each operational zone under the 50-year and 100-year design storm events for SSP5-8.5. These depth categories follow widely used thresholds in flood hazard mapping and infrastructure impact assessments, where depth-based groupings are applied to distinguish shallow, nuisance level inundation from deeper water capable of producing more severe operational challenges [18]. In Figure 13 and Figure 14, the depth distribution for each scenario is expressed as a percentage of the inundated area falling within each depth class, allowing direct comparison of how much of each zone is occupied. This representation clarifies not only whether a zone floods, but also how the severity of flooding shifts across scenarios, highlighting whether projected future conditions increase the share of deeper hazard classes or primarily expand shallow ponding.

For the Cumberland County Improvement Authority, the depth results indicate that most operational zones remained dominated by very shallow to shallow flooding, while the clearest progression toward deeper inundation occurred at the landfill toe area [LTA]. The administrative area [AA] was strongly dominated by the very shallow flooding class, accounting for about 90 to 95% of the inundated area under the 50-year design storm and about 87 to 92% under the 100-year design storm, with most of the remainder in the shallow flooding class and essentially no meaningful share in the deeper classes. The parking area [PA] also remained shallow dominated, with about 58 to 63% in the very shallow class and about 35 to 39% in the shallow class, while only a small fraction, about 3 to 5%, fell within the moderate-to-deep flooding class. A similar pattern was observed for the center of landfill [COL], where about 67 to 71% of the inundated area remained in the very shallow class and about 22 to 27% in the shallow class, with only about 6 to 7% in the moderate-to-deep class and little-to-no very deep flooding. The roads [R] remained among the least severe zones, with about 80 to 90% of the area in the very shallow flooding class and most of the remainder in the shallow class.

In contrast, the landfill toe area [LTA] showed the broadest depth distribution and the strongest temporal increase in flood severity. Under the 50-year design storm, the combined very shallow and shallow classes declined from about 65% at baseline to about 45% by late century, while the combined deeper classes increased from about 35% to about 55%. This increase was driven by growth in both the moderate-to-deep flooding class, which rose from about 27% to about 33%, and the very deep flooding class, which increased from about 9% to about 23%. Under the 100-year design storm, the same pattern became more pronounced. The combined shallow classes decreased from about 58% at baseline to about 40% by late century, whereas the combined deeper classes increased from about 42% to about 60%. In this case, the very deep flooding share increased from roughly 15% at baseline to about 29% by late century, while the moderate-to-deep flooding class remained in the upper 20% to low 30% range. Overall, the landfill toe area was the primary depth-sensitive hotspot within the CCIA domain, while the other zones remained predominantly shallow flooded.

For the GCSWC landfill, the depth distribution indicates that future change is expressed primarily as a redistribution within the shallow flooding classes across most operational zones, while the most notable increase in flood severity occurs at the landfill toe area [LTA]. This distinction is important because the operational implications of flooding are strongly influenced by water depth, with deeper inundation generally associated with greater disruption to access, infrastructure performance, and recovery. The administrative area [AA] does not remain entirely within the shallow classes, as all four depth classes are represented under both the 50-year and 100-year design storms across baseline, mid-century, and late-century conditions. In contrast, the parking area [PA], center of landfill [COL], and roads [R] remain predominantly shallow flooded. For PA, the inundated area is almost entirely confined to the very shallow flooding [0 to 0.30 m] and shallow flooding [0.30 to 0.91 m] classes for both storm events and all time horizons, with essentially no meaningful area in the deeper classes. A similarly stable pattern is observed for R, where the inundated area remains concentrated within the shallow classes throughout the study period. The COL also remains shallow dominated, with the very shallow class accounting for most of the inundated area and only a small proportion occurring within the moderate-to-deep flooding [0.91 to 1.83 m] class, while very deep flooding [at least 1.83 m] is absent.

The landfill toe area [LTA] was identified as the principal flood transition zone within the GCSWC domain. However, flooding in this zone remained primarily concentrated within the shallow classes. Under the 50-year design storm, the combined very shallow and shallow flooding classes decreased from about 71.7% at baseline to about 58.0% by late century, indicating that shallow flooding still represented the dominant areal condition. Under the 100-year design storm, the same general pattern was observed, with the combined shallow classes decreasing from about 64.4% at baseline to about 48.2% by late century. Although deeper flooding became more evident over time, this increase should be interpreted as a secondary shift superimposed on a flood regime that was still largely shallow. The increase in deeper flooding was driven by growth in both the moderate-to-deep flooding class, which rose from about 29.7% to about 34.8%, and the very deep flooding class, which increased from about 8.0% to about 19.2%.

Taken together, the depth class results clarify not only how much of each landfill becomes inundated but also how that inundation is distributed across severity levels, revealing which operational zones are most susceptible to deeper water and which remain dominated by shallow ponding. These patterns establish a clearer understanding of hazard intensity across CCIA and GCSWC and highlight where climate-driven changes translate into meaningful shifts in flood behavior. With this depth-based perspective in place, it becomes important to examine the maximum flood depths experienced within each zone, since peak depth often determines the upper limit of structural loading, access disruption, and operational risk. Thus far, the depth class analysis has shown that the landfill toe consistently emerges as the most depth-sensitive hotspot, while the central landfill areas at both sites remain largely confined to shallow, ponding-type inundation. The following section therefore builds how the maximum point-based depths respond to different design storm scenarios.

4.6. Maximum Flood Depth Sensitivity to Climate-Adjusted Design Storms by Operational Zone

Across the assessed operational zones at CCIA, maximum flood depths fall into distinct and repeatable ranges that align closely with the adopted hazard classes. Flooding at the landfill core remains consistently below 0.31 m, which corresponds to the very shallow hazard class and indicates minimal localized inundation at the central disposal footprint. In contrast, the access road and paved operational surfaces typically experience maximum flood depths of the order of 0.3–1.5 m, which fall within the moderate–deep hazard category, implying persistent disruption potential for circulation, site access, and routine operations even when peak depths do not escalate sharply. When comparing across baseline versus climate-adjusted design storm scenarios, the results show that increases in storm forcing do not produce substantial increases in maximum flood depth at most monitored operational locations. Instead, flood depths remain broadly stable within the same hazard class bands across scenarios. This indicates that the primary expression of changing flood risk is not a large vertical increase in peak depth across the site but rather the consistency with which certain operational areas remain within disruptive hazard classes.

A clear exception to this stability is observed at the landfill toe area. Toe zone depths are consistently the largest among all the areas assessed and show the most noticeable scenario to scenario increase, and progress from already severe baseline conditions to higher late-century values. This pattern indicates that the toe area is the most hydraulically sensitive component of the facility under intensified rainfall forcing, and it is the only zone where changes in forcing translate into a more visible increase in maximum depth.

These patterns are summarized visually in Figure 15, which compares maximum flood depths for the center of the landfill mass and the landfill toe area under baseline SSP5-8.5 mid/late conditions for both 24 h 50-year and 100-year design storms. The bar chart shows that depths at the landfill core remain consistently below 0.3 m across scenarios, indicating minimal sensitivity to scenario forcing at this location. In contrast, the landfill toe area exhibits substantially higher depths (2.1 to 3.33 m), typically exceeding 2.7–3.03 m, and shows the clearest scenario-driven amplification under future climate forcing. Overall, the relative ranking of exposure remains stable across all scenarios with the toe zone consistently exhibiting the highest flood depths and therefore representing the facility’s dominant and most responsive high-hazard hotspot.

The spatial organization of flood depths across CCIA is strongly controlled by site topography. The landfill mass is situated on relatively elevated terrain which promotes rapid shedding of rainfall-generated runoff along engineered gradients and cover materials and limits water accumulation on the landfill crest and central cell. As a result, even under high-intensity storm events, inundation of the landfill mass remains shallow relative to surrounding operational zones as rainfall-driven flooding at facilities is frequently governed by local terrain controls and storage behavior in low-gradient areas rather than by conveyance-dominated flow [81].

In contrast, terrain elevation differences were being explained through the flood response observed at the landfill toe. The toe zone lies at a lower elevation than the landfill crest and represents a slope transition region where surface gradients decrease. Runoff draining from the elevated landfill surface converges in this area, flow velocities diminish, and localized ponding develops. Consequently, the higher flood depths observed at the landfill toe reflect terrain-controlled accumulation driven by elevation contrast and slope transition, rather than backwater effects or channelized inflow.

Taken together, the CCIA results indicate that the limited sensitivity of maximum flood depth to increasing storm magnitude reflects the combined effects of elevated landfill terrain, efficient slope-driven runoff shedding, and storage-dominated ponding within low-gradient operational areas. While future climate projections indicate more intense rainfall, their primary impact at CCIA is expressed through expanded and more persistent shallow inundation rather than substantial increases in localized peak depths. This pattern coincided with change being strongly influenced by connectivity loss, access disruption, and prolonged shallow inundation [82]. The FEMA report [83] emphasized that operational consequences may be substantial even when peak depths remain moderate.

Flooding patterns across the GCSWC demonstrate a stable response under both baseline and climate-adjusted design storm conditions. Across all evaluated scenarios, inundation remains spatially coherent with consistent relative exposure among functional areas reflecting the dominant influence of local topography and engineered grading on surface runoff routing and ponding [84]. The main landfill cell and parking areas consistently fall within the very shallow flood hazard category across all baseline and future scenarios. These areas exhibit minimal sensitivity to increasing storm magnitude or climate forcing, indicating that their elevated position, efficient runoff shedding, and limited surface storage confer a high degree of resilience to intensified rainfall and limits water accumulation on the landfill crest and central cell.

Figure 16 shows that the landfill toe area consistently experiences the highest modeled flood depth across all scenarios, increasing from about 1.97 [m] under the baseline 50-year design storm to about 3.17 [m] under the late-century 100-year design storm. This pattern indicates progressive intensification of flood severity under stronger storm and future climate conditions. Unlike the center of landfill mass, which remains nearly constant at about 0.26 to 0.27 [m], the landfill toe area remains in the very deep flooding class throughout all scenarios. These results confirm that the landfill toe area is the primary flood-sensitive hotspot in the GCSWC landfill.

5. Conclusions

This study establishes a climate-adjusted flood hazard assessment framework for landfill by integrating watershed scale hydrology (HEC-HMS), two-dimensional hydrodynamics in HEC-RAS 2D, and evaluation against Sentinel-1 SAR-derived flood extents processed in Google Earth Engine. HEC-HMS is purpose-built to simulate precipitation–runoff processes at the watershed scale, enabling event-based design hydrographs that can be propagated into hydraulic domains.

A primary outcome is the systematic increase in 24 h design storm magnitudes derived from GCM-informed precipitation–frequency analysis under SSP2-4.5 and SSP5-8.5, indicating that historical design depths correspond to shorter future return periods and therefore represent more frequent events under future climate conditions. This interpretation is consistent with the assessed expectation that heavy precipitation generally becomes more frequent and intense with additional warming, with particularly large frequency changes for rarer events. Although both pathways were evaluated, SSP5-8.5 was emphasized as a high-end stress test scenario to provide an upper bound estimate of climate-forced hazard amplification for long-term planning.

Evaluation using SAR-derived flood extents supported the selection of a minimum depth cutoff of t = 0.31 m (≈1 ft) for converting modeled depth fields into binary inundation masks at both landfill domains. Agreement was characterized by high precision and moderate recall, while higher depth cutoffs reduced recall and F1 by excluding shallow wetting.

When climate-projected storms were propagated through the coupled hydrologic to hydraulic modeling chain, the dominant climate signal was expressed primarily as expansion of inundation extent and increased spatial connectivity from baseline to mid-century (2025–2050) and late-century (2070–2100) conditions, rather than wholesale relocation of flooding to entirely new areas. The inundated area increased consistently from baseline to late century for every operational zone at both facilities, and 100-year storms produced larger footprints than 50-year storms. Exposure hotspots were facility specific: at CCIA, the landfill toe consistently produced the largest inundated areas and the strongest growth across time horizons and storm magnitudes; at GCSWC, inundation was greatest across the road network and the center of the landfill.

Overall, the findings demonstrate that operational vulnerability cannot be represented by a single peak depth or a facility-wide statistic. The combined zoning and monitoring point framework, depth-based hazard interpretation, and SAR-supported inundation threshold provide a defensible basis for identifying where climate-driven changes translate into meaningful operational risk. The results support adaptation prioritization focused on landfill toes and critical access corridors, including targeted drainage and grading improvements and protection of transportation and operational support areas, to maintain facility operability under increasing extreme rainfall intensity across the twenty-first century.

6. Recommendations

Based on the findings of this study, several measures are recommended to reduce flood-related operational risks at municipal solid waste landfill facilities under future climate conditions. Priority should be given to targeted drainage and grading improvements at landfill toe areas, where flooding was shown to concentrate and persist across scenarios. Improving surface slopes and drainage efficiency in these low-relief zones is expected to reduce localized ponding and limit the lateral expansion of inundation.

Reinforcement and, where feasible, elevation of critical access routes should be considered to maintain site operability during extreme rainfall events, as access corridors were identified as highly sensitive to shallow but persistent flooding. In addition, regular inspection and maintenance of conveyance features, including culverts, ditches, and low-gradient flow paths, are recommended to preserve hydraulic connectivity and prevent flow obstructions that can exacerbate surface flooding.

Hazard escalation was driven mainly by expanded inundation, connectivity, and persistence not by uniform peak depth increase. Therefore, future risk management should go beyond point depth metrics and include spatial exposure, access continuity, and functional site zoning.

Finally, the integrated, climate-informed hydrologic–hydraulic framework demonstrated in this study is recommended for broader application to other waste management and critical infrastructure facilities, particularly in data-limited settings. Its use can support proactive adaptation planning by translating projected changes in extreme rainfall into operationally meaningful flood hazard indicators under mid- and late-century climate scenarios.