Long-Term Land Use/Land Cover Change and Climate-Driven Projection of Soil Organic Carbon Stocks and Sequestration Using the RothC Model in the Northern Nile Delta, Egypt

Abstract

Soil organic carbon (SOC) is a major component of the global carbon cycle. This study aimed to: (i) monitor five decades’ land use/land cover (LULC) changes in the northern Nile delta using Landsat imagery; (ii) quantify baseline SOC stocks (SOCs) in 2021; (iii) project SOCs and potential SOC sequestration (PSOCS) to 2100 under four SSP2-4.5 climate scenarios using RothC model; and (iv) evaluate uncertainty in SOCs and PSOCS projections using the Monte Carlo approach. Sixty soil samples were collected during the winter and summer seasons of 2018/2019 (30 per season). Agricultural land expanded from 12% in 1972 to 35% in 2021, while fish farms, established in the 1990s, accounted for 24% of the area by 2021. SOCs varied across LULC types and seasons. Between 13 and 28% of agricultural land exceeding 7 Mg C ha⁻¹ in summer and winter, respectively. Barren land and sabkha were characterized by low SOCs (<3 Mg C ha⁻¹). Model predictions indicate that mean SOCs will increase from 5.83 (2021) to 6.16 (mid-century), followed by a decline to 5.96 Mg C ha⁻¹ by 2100. Estimated PSOCS range from 0.13 to 0.32 Mg C ha⁻¹. Monte Carlo uncertainty analysis yielded median SOCs between 6.01 and 6.27 Mg C ha⁻¹ and median PSOCS between 0.18 and 0.44 Mg C ha⁻¹, reflecting moderate projection uncertainty.

1. Introduction

Soil organic matter (SOM), composed of plant and animal residues, microbial biomass, and stable humified substances [1], is fundamental to soil health despite constituting a small fraction of mineral soils. Its principal component, soil organic carbon (SOC), regulates nutrient cycling, water retention, and biological activity [2]. SOC also mitigates climate change by sequestering atmospheric carbon dioxide (CO₂) into the soil carbon (C) pool [3]. This sequestration is driven mainly by plant photosynthesis and can be enhanced by agricultural management practices [4,5,6]. Because SOC stocks (SOCs) are strongly influenced by land use and management practices, maintaining a positive C balance is essential for sustaining soil quality [7].

Land use/land cover (LULC) change and agricultural practices within the agriculture, forestry and other land use (AFOLU) sector are major sources of greenhouse-gas (GHG) emissions. Between 2007 and 2016, AFOLU contributed about 13% of global CO₂ emissions and 23% of net anthropogenic GHG emissions. Simultaneously, terrestrial ecosystems functioned as a net CO₂ sink of around 6.0 ± 3.7 Gt CO₂ yr⁻¹(Gigatonne = 10⁹ t) [8]. Improved cropland management could sequester 0.9–1.85 PgC yr⁻¹ [9]. However, conversion of natural ecosystems to agriculture typically reduces initial SOC stocks by 30–50% within 50 years in temperate climates [10].

Soil constitutes the largest terrestrial reservoir of organic C. They store more C than the atmosphere (around 800 Pg C) and terrestrial biomass (around 500 Pg C) combined [11,12]. Of the estimated 2500 Pg C in soils, approximately 20% resides in topsoil [13,14]. Recent assessments estimate 1500–2400 Pg C within the upper 1.0 m and up to 3000 Pg C when deeper layers are included [15,16]. Consequently, even small SOC changes induced by LULC conversion, soil disturbance, or altered residue inputs can significantly influence global C dynamics [12,17,18]. SOC sequestration occurs when C inputs exceed C losses [19,20]. This balance underscores the importance of sustainable land management for climate mitigation [11,21].

Understanding long-term SOC dynamics requires process-based modeling. The Rothamsted Carbon model (RothC) is widely applied to simulate monthly SOC turnover in non-waterlogged topsoil over decadal to centennial timescales [22,23]. RothC partitions SOC into one inert organic matter (IOM) pool, and four active C pools; decomposable plant material (DPM), resistant plant material (RPM), microbial biomass (BIO), and humified organic matter (HUM) [23]. Although originally developed for temperate arable soils, RothC has been successfully applied across diverse ecosystems [15,24,25,26,27,28,29,30]. Recent refinements have improved its representation in conservation practices and enhanced predictive performance [31,32].

The Nile delta is among the world’s most climate-vulnerable deltas. It is threatened by sea-level rise (SLR), land subsidence, salinization, and rapid LULC change [33]. Projections indicate that 1500–2600 km² may be exposed to flooding by 2100 under moderate SLR scenarios [34,35]. In the Kafr El-Sheikh Governorate, studies documented wetland loss, agricultural land conversion, and expansion of fish farms and urban areas [36,37,38,39,40,41]. SOC stocks in the region vary considerably with soil texture, land use, and management history. Higher SOC stocks are typically associated with fine-textured and long-cultivated soils [42,43,44,45]. However, most regional studies rely on static SOC measurements and short-term LULC assessments. Few integrated long-term LULC dynamics, process-based modeling, and uncertainty analysis extend to 2100.

The central scientific question of this study is how will long-term LULC and climate scenarios influence SOCs in the northern Nile delta, and what is the magnitude of uncertainty in projected SOC sequestration to 2100? To address this question, we integrate spatial analysis, field data, modeling, and uncertainty quantification within a unified framework. Specifically, we: (1) monitor five decades’ LULC change (1970s–2010s) in the northern Nile delta using Landsat imagery, (2) quantify and map 2021 topsoil SOCs and its seasonal variability as modeling baseline, (3) simulate future SOCs and potential SOC sequestration (PSOCS) to 2100 under four SSP2-4.5 climate scenarios using RothC model, and (4) quantify uncertainty in SOCs and PSOCS projections using Monte Carlo simulation.

This integrated approach provides the longest LULC reconstruction for the region, the first multi-scenario RothC-based SOCs projections to 2100, and the first quantitative assessment of PSOCS uncertainty in the Nile delta. The findings deliver high-resolution, policy-relevant insights for soil C accounting, climate mitigation, and land-use planning.

2. Materials and Methods

2.1. Study Area

2.1.1. Location and Soils

The study area is located in the northern Nile delta within the Kafr El-Sheikh Governorate, Egypt (Figure 1A). It covers approximately 1480 km² and extends between longitudes 30°20′ and 31°10′ E and latitudes 31°15′ and 31°37′ N. The region is characterized by intensive agricultural activity and substantial land-use pressures. It includes Lake Burullus and its surrounding areas. Lake Burullus, Egypt’s second largest coastal lake, has experienced substantial anthropogenic-driven shrinkage [46]. Guirguis [47] reported losses of 8.6 km² yr⁻¹ between 1983 and 1991. Hossen and Negm [48] detected lake a reduction in lake area from approximately 32,000 to 18,000 ha between 1984 and 2015. Over a similar period, Khalil [49] reported a decline from 503 to 410 km². Keshta [39] observed a decrease in marsh and wetland areas from 336 to 185 km² between 1985 and 2020 due to agricultural expansion and fish farm development. Additionally, water bodies and floating vegetation declined by 16% and 52%, respectively, between 1984 and 2019 [50].

The Nile delta is predominantly composed of Nile alluvial deposits. According to previous studies [51,52,53], three dominant depositional environments occur in the region (Figure 1B): flood plain (fluvial) deposits, lacustrine deposits, and the quaternary marine deposits. Lacustrine deposits occupy approximately one-third of the area, while flood plain and marine deposits together account for another third.

A total of sixty topsoil samples were collected during the winter and summer seasons of 2018/2019 (30 per season) (Figure 1B). The sampling strategy was designed to capture the spatial variability across the three main depositional units within the study area (Figure 1B). Sampling depth followed the IPCC guidelines for SOCs estimation [54]. Soil texture ranges from sandy to clayey across the region. According to the USDA Natural Resources Conservation Service Keys of Soil Taxonomy [55], soils were classified into two orders: Entisols and Aridisols.

2.1.2. Climate

The study area is characterized by a Mediterranean climate, with mild winters, hot summers, and low annual precipitation. Meteorological data for the period 1981–2021 were obtained for two locations on the eastern (a) and western (b) margins of Burullus Lake (Figure 1B). Data were sourced from the NASA Prediction of Worldwide Energy Resources (POWER) (https://power.larc.nasa.gov) database. The dataset includes daily air temperature (maximum, minimum, and mean at 2 m, °C), wind speed (maximum, minimum, and mean at 2 m, m s⁻¹), precipitation (daily and cumulative, mm), and relative humidity (%). Additionally, four soil-related parameters were derived from the climatic records: earth skin temperature (°C), and three soil wetness indices representing moisture condition at the surface, root-zone, and deep-profile. Monthly means were calculated for each site (Table S1 in Supplementary Materials) and averaged over the full 40-year time series (Table S1 and Figure 2).

The eastern site exhibited higher maximum temperatures, with an annual mean approximately 1.5 °C greater than the western site over the 40-year period (Figure 2 and Table S1). Conversely, minimum temperatures were slightly higher in the west by around 0.7 °C (Figure 2 and Table S1). Mean air temperatures were comparable between the two locations, ranging from around 15 °C in winter (January and February) to around 27 °C in summer (July and August) (Figure 2 and Table S1). Wind speed was relatively uniform across the area, with an average of around 4 m s⁻¹ annually (Table S1). Relative humidity was marginally higher at location b (67%) than at location a (66%). This pattern corresponds to greater annual precipitation in the west (139 mm) compared to the east (103 mm) (Table S1). Precipitation exhibited strong seasonality, with minimal rainfall during summer and peak values in January (Figure 2 and Table S1). Earth skin temperature followed a clear seasonal cycle, peaking in summer and reaching a minimum in winter (Figure 2 and Table S1). Soil wetness indices (0–1) were consistently higher at the western site, reflecting higher precipitation and greater soil moisture across all soil depths (Table S1).

2.2. Methodological Framework

As illustrated in Figure 3, the methodological framework comprised four principal components. First, LULC dynamics were monitored using Landsat imagery of 1972, 1978, 1984, 1990, 1996, 2002, 2008, 2018, and 2021. Second, SOC stocks were quantified and spatially mapped based on soil depth, SOC content, and bulk density. Third, predicted SOCs and PSOCS were simulated through to 2100 using climate data. Finally, the uncertainty associated with projected SOCs and PSOCS was assessed using a Monte Carlo approach.

2.2.1. Land Use/Land Cover Mapping

Nine cloud-free Landsat images (1972, 1978, 1984, 1990, 1996, 2002, 2008, 2018, and 2021) were acquired (

Table 1. Detailed information on the Landsat satellite images utilized in this study.

Imagery Date	Spatial Resolution	Spacecraft ID	Sensor Identifier *	Path/Row	No. of Bands	Band Combination (R-G-B)
1972	60 m	Landsat-1	MSS	190/038	4	4-2-1
1978	60 m	Landsat-3	MSS	190/038	4	4-2-1
1984	60 m	Landsat-5	MSS	177/038	4	4-2-1
1990	30 m	Landsat-4	TM	177/038	7	4-7-3
1996	30 m	Landsat-5	TM	177/038	7	4-7-3
2002	30 m	Landsat-5	TM	177/038	8	4-7-3
2008	30 m	Landsat-5	TM	177/038	8	4-7-3
2018	30 m	Landsat-8	OLI/TIRS	177/038	11	7-5-4
2021	30 m	Landsat-8	OLI/TIRS	177/038	11	7-5-4

* MSS: multispectral scanner, TM: thematic mapper; OLI: operational land imager; TIRS: thermal infrared sensor.

) from USGS EarthExplorer (https://earthexplorer.usgs.gov/). All images were geometrically corrected to the UTM zone 36N projection with the WGS84 datum. Image processing and analysis were conducted using ENVI 5.1 (Exelis Visual Information Solutions, Boulder, Colorado) and ArcMap 10.4 [56], applying appropriate band combinations (Table 1 and Figure 4). LULC classification was conducted using a supervised maximum likelihood classifier (MLC). This method was selected for its robustness, proven accuracy, and relative simplicity compared to more complex machine-learning algorithms [57,58,59,60]. Training samples were established for each LULC class and validated using the Google Earth imagery and ground truth points (GTPs). A spectral signature file was generated, and final LULC maps were produced using MLC decision rule.

The accuracy of the classified LULC maps was assessed using a confusion matrix approach. Overall accuracy, the Kappa coefficient, producer’s accuracy, and user’s accuracy were calculated to evaluate both overall and class-specific performance. The complete confusion matrix is provided in Supplementary Table S2.

2.2.2. Laboratory Analyses and Soil Organic Carbon Stock Calculations

Soil samples were air-dried, sieved to <2 mm, and stored for subsequent physiochemical analyses. Particle size distribution was determined following the method of Gee and Or [61]. SOC content was measured using the Walkley–Black method [62]. Bulk density (BD) was determined according to Grossman and Reinsch [63].

Although soil samples were collected in winter 2018 and summer 2019, laboratory analyses were completed in 2021 due to COVID-19-related delays. To ensure temporal consistency, LULC dynamics between 2019 (completion of sampling) and 2021 were checked using Landsat imagery. The analysis indicated no statistically significant LULC changes within the study area during this period. Major LULC transitions occurred primarily before or during 2018–2019. Given this stability, SOC measurements from 2018/2019 were considered representative of 2021 conditions and were integrated with the 2021 LULC map without numerical adjustment. Accordingly, the term “2021 baseline” refers to the year of finalized dataset validation and spatial integration rather than the sampling year. SOCs, expressed as Mega gram C per hectare, Mg C ha⁻¹, was calculated for topsoil using the widely adopted equation proposed by Gue and Gifford [64], and later formalized by IPCC [65]. This approach is consistent with international guidelines for SOCs estimation [66] and support most field-level SOC assessments:

SOCs = D × OC × BD × 10,

(1)

where SOCs is the soil organic carbon stock (Mg C ha⁻¹), D is the depth of sampling (cm), OC is the organic carbon (%), BD is the soil bulk density (g cm⁻³), and 10 is the unit conversion factor. This equation applies to fine-earth fraction (<2 mm). For soils containing coarse fragments, a correction factor of (1- coarse fragments volume/100) was applied as recommended by IPCC [65]. This equation has been widely validated across divers regions. Similar formulations were used by Wairiu and Lal [67] in the Solomon Islands, Yigini and Panagos [68] and Chen et al. [69] in Europe, Zeng et al. [70] and Li et al. [71] in China, Sharma et al. [72] in India, and Abu-hashim et al. [43] and Abd-Elmabod [73] in the Nile delta, Egypt.

The SOCs calculations (Equation (1)) were conducted using soil samples corresponding to four LULC classes, including agricultural land, barren land, dry sabkha, and wet sabkha based on 2021 LULC classified map. Of the 30 samples per season, of which 23 were located within these LULC classes and were therefore included in the SOCs analysis. The remaining seven samples, situated in fish farm facilities and urban surrounding areas, were primarily used as GTPs for classification validation. SOCs values were spatially interpolated and integrated with the 2021 LULC map to generate a spatial SOCs distribution map. These values serve as a baseline for projecting SOCs changes through to 2100.

The descriptive statistics and a Wilcoxon signed-rank test were performed to assess seasonal significant difference in soil characteristics at a 95% confidence interval (α = 0.05). Seasonal SOCs values were also statistically examined. All statistical analyses were achieved using SPSS 26 software [74].

2.2.3. RothC Model for Estimating SOCs and PSOCS Until 2100

Future SOCs and PSOCS (predicted SOCs minus SOCs in 2021) were simulated to 2100 using RothC model under spatially explicit climate projection. Simulations were conducted under four medium-forcing scenarios of Shared Socioeconomic Pathway (SSP2-4.5, “middle-of-the road”) conditions [75,76].

By 2100, projected changes in monthly mean temperature, precipitation summation, and potential evapotranspiration (PET; Penman–Monteith; mm) were as follows; +1.5 °C, −5%, +3% (scenario 1, Sce.1), +1.8 °C, −10%, +5% (scenario 2, Sce.2), +2.2 °C, −15%, +8% (scenario 3, Sce.3), and +2.5 °C, −20%, +12% (scenario 4, Sce.4), respectively. These projections are consistent with reported values for Northern Africa, the Eastern Mediterranean, and Egypt under SSP2-4.5 [8,33,77,78,79,80].

The RothC model was implemented in RStudio 2025.09.2 [81] under a business as usual (BAU) assumption. No changes in land use, management, or cultivation practices were assumed relative to the 2021 baseline [15,29,82]. Soil organic matter dynamics were simulated using the “SoilR” package [83], supported by “deSolve” for solving differential equations [84], “read” package [85], and “dplyr” package [86] for data processing.

The model inputs included (i) projected climate data (2022–2100; temperature, precipitation, and PET); (ii) soil properties (initial SOCs in 2021, C inputs to DPM, RPM, BIO, HUM, and IOM, clay content, and soil depth), with a default DPM/RPM ratio of 1.44 for croplands; and (iii) LULC derived from Landsat imagery of 2021, coded as 1.0 for cultivated and 0.0 for non-cultivated. Parameter calculations followed established RothC model formulations that account for plant litter quality and SOC turnover [29,87,88]. Details are provided in Supplementary Table S3.

2.2.4. Monte Carlo Uncertainty Approach

Uncertainty in projected SOCs and PSOCS was quantified using a Monte Carlo approach [89]. Random realizations of model inputs and parameters were generated from predefined probability distribution functions (PDFs). This approach propagates uncertainty associated with input data, parameterization, and initial conditions through RothC simulations [15,26,90]. The analysis focused on parametric and input-data uncertainty. Considered parameters included annual C inputs, clay content, initial SOC stocks, and climatic variables. Prior to distributional assumptions, these parameters were tested for normality using the Shapiro–Wilk test. The resulting p-values 0.129, 0.398, and 0.192 for annual C inputs, clay content, and initial SOCs, respectively, indicated no significant deviation from normality (α = 0.05). Therefore, normal distributions were assumed, with means equal to parameter estimates and standard deviations derived from measured standard errors [15]. Because these variables are physically bounded, truncated normal distributions were applied. Lower bounds were set at zero for SOC and C inputs, and clay content was constrained to the range 0–100%. Non-physical draws were rejected and resampled. This procedure captures uncertainty arising from measurement error, parameter estimation, and initial conditions.

For each SSP2-4.5 climate realization, 1000 simulations were performed. The RothC model was run at a monthly time step from 2022 to 2100 using the SoilR and dplyr packages in RStudio. The Monte Carlo procedure was applied independently to each climate realization.

This framework quantifies uncertainty associated with input data and parameter estimates within a fixed RothC structure. Structural uncertainty, including alternative process representations or SOCs model formulations, was not evaluated. Scenario-related uncertainty was partially addressed through multiple SSP2-4.5 realizations. Uncertainty associated with alternative climate–socioeconomic pathways or future land-use transitions was beyond the study scope.

Model outputs were summarized as median SOCs and PSOCS trajectory, accompanied by the 5th and 95th percentile representing uncertainty bounds. This probabilistic framework characterizes long-term SOC dynamics under parameter and input uncertainty. It also enables direct comparison between parametric uncertainty and climate-scenario uncertainty [15].

3. Results

3.1. Land Use/Land Cover Classes

Nine LULC classification maps were produced using the MLC procedure for the years 1972, 1978, 1984, 1990, 1996, 2002, 2008, 2018, and 2021. The spatial extent and proportional coverage of each class are presented in Figure 5 and

Table 2. Area coverage by square kilometer (km²) and percentage (%) of land use/land cover (LULC) classes in the study area from 1972 to 2021.

LULC		1972	1978	1984	1990	1996	2002	2008	2018	2021
Agricultural Land	km2	180.63	187.59	209.68	256.66	304.7	368.57	407.89	479.68	510.58
	%	12.21	12.68	14.18	17.35	20.6	24.92	27.58	32.43	34.52
Fish Farm	km2	0.00	0.00	0.00	109.94	102.26	105.03	181.62	306.66	347.73
	%	0.00	0.00	0.00	7.43	6.91	7.10	12.28	20.73	23.51
Urban Areas	km2	0.66	0.93	0.78	4.55	10.73	12.70	30.25	64.48	63.55
	%	0.04	0.06	0.05	0.31	0.73	0.86	2.05	4.36	4.30
Water Bodies	km2	460.87	337.49	276.10	315.71	329.55	297.02	283.38	231.73	257.1
	%	31.16	22.82	18.67	21.35	22.28	20.08	19.16	15.67	17.38
Natural Vegetation	km2	110.66	233.39	246.98	230.4	193.7	160.9	149.43	153.55	189.91
	%	7.48	15.78	16.70	15.58	13.10	10.88	10.10	10.38	12.84
Barren Land	km2	250.89	215.78	195.88	70.45	65.89	62.20	59.32	54.38	27.86
	%	16.96	14.59	13.24	4.76	4.46	4.21	4.01	3.68	1.88
Dry Sabkha	km2	315.74	350.75	398.89	355.6	326.91	327.09	235.76	113.81	28.34
	%	21.35	23.72	26.97	24.04	22.10	22.12	15.94	7.70	1.92
Wet Sabkha	km2	159.56	153.08	150.70	135.70	145.27	145.50	131.36	74.72	53.94
	%	10.79	10.35	10.19	9.18	9.82	9.84	8.88	5.05	3.65

, respectively. The identified LULC classes comprise agricultural land, urban/built-up areas, water bodies (mainly Burullus Lake), fish farms, natural vegetation, wet sabkha, dry sabkha, and barren land.

The classified maps demonstrated high performance. The overall accuracy and kappa coefficient were approximately 0.90. The detailed confusion matrix and class-specific accuracy metrics are provided in Supplementary Table S2. Overall classification performance was consistently strong across all years, with a mean overall accuracy of approximately 90%. The lowest accuracy was recorded for the 2008 map (85%), whereas the highest was achieved in 2021 (95%). Class-specific evaluation indicates that agricultural land attained high producer’s and user’s accuracies (approximately 90%), confirming the robustness and reliability of the classification results.

Between 1972 and 2021, agricultural lands and urban areas expanded markedly and consistently. Agricultural lands increased from 181 km² (12%) in 1972 to 511 km² (35%) in 2021. Urban areas expanded from 0.66 km² (0.04%) in 1972 to 64 km² (4%) in 2021 (Table 2). These trends are clearly illustrated in Figure 5, highlighting sustained agricultural reclamation and urban growth over time.

Fish farms emerged as a dominant land-use class following their establishment in the late 1980s and early 1990s. By 2021, fish farms occupied approximately 348 km² (24%) of the study area. Their extent increased more than threefold between 2002 and 2021 (Figure 5 and Table 2), underscoring the rapid expansion of aquaculture activities.

The water class exhibited notable temporal fluctuations. Its areal coverage declined from 461 km² (31%) in 1972 to 276 km² (19%) in 1984, increased to 330 km² (22%) in 1996, and subsequently decreased to 232 km² (16%) by 2018 (Table 2). Natural vegetation associated mainly with the Burullus Lake surface increased from 111 km² (7%) in 1972 to approximately 16% during the period 1978–1990. Thereafter, it declined to around 9% (130 km²) in 2008 before recovering to 190 km² (13%) by 2021 (Table 2).

Non-productive LULC classes—including barren land, dry sabkha, and wet sabkha—dominated the landscape during the early study period. Collectively, they accounted for nearly 50% of the total area between 1972 and 1984 (Table 2 and Figure 5). Their combined extent declined to around 36% by 2002 and further decreased to approximately 29% by 2008. By 2021, these classes had dropped sharply to around 7.5% (Figure 5). This substantial decrease reflects extensive LULC conversion over the study period, particularly during the most recent decade.

3.2. Soil Characteristics

Descriptive statistics for laboratory-measured soil properties during winter and summer 2018/2019 are presented in

Table 3. Descriptive statistical parameters of measured soil properties (number of soil samples = 30).

Soil Properties		Soil Organic Carbon		Sand		Silt		Clay		Bulk Density
		%								g cm−3
		Summer	Winter	Summer	Winter	Summer	Winter	Summer	Winter	Summer	Winter
Minimum		0.14	0.34	2.85	7.40	3.15	3.00	3.45	3.00	0.96	1.06
Maximum		1.97	2.13	93.40	94.00	47.49	47.20	65.25	63.90	1.58	1.68
Mean		1.30	1.47	30.11	31.64	25.92	25.54	43.97	42.84	1.28	1.38
Std. Dev. 1		0.52	0.51	28.79	28.57	12.52	12.58	19.40	19.14	0.14	0.13
Std. Err. 2		0.09	0.09	5.26	5.22	2.29	2.30	3.54	3.49	0.02	0.02
Percentile	25%	1.02	1.21	10.26	11.10	16.85	17.10	35.00	34.30	1.23	1.33
	50%	1.44	1.61	16.40	18.10	26.40	25.85	50.78	49.65	1.30	1.40
	75%	1.66	1.83	44.00	44.70	35.50	35.10	59.00	57.46	1.35	1.45
Skewness		−0.86	−0.83	1.31	1.32	−0.22	−0.23	−1.10	−1.10	−0.53	−0.53
Kurtosis		−0.15	−0.19	0.32	0.32	−0.76	−0.71	0.01	−0.01	0.73	0.79
p-value		<0.001 *		<0.001 *		0.008 *		<0.001 *		<0.001 *

¹ Std. Dev., standard deviation; ² Std. Err., standard error; * p-value is significant at significance level of α = 0.05.

. The SOC contents range from 0.14 to 1.97% in summer and 0.34 to 2.13% in winter, with an overall mean of around 1.4% (Table 3).

Soil texture fractions exhibited limited seasonal variability. Sand content showed a slight increase in winter (mean = 31.64%) compared to summer (30.11%). Silt and clay contents remained relatively stable across seasons. Mean clay content was approximately 44% in summer and 43% in winter.

The BD values ranged from 0.96 to 1.58 g cm⁻³ in summer and 1.06 to 1.68 g cm⁻³ in winter. The mean BD increased modestly from 1.28 to 1.38 g cm⁻³ (Table 3). SOC displayed moderate variability. In contrast, sand and clay fractions exhibited greater dispersion. Skewness and kurtosis values indicated near-normal distributions for most soil properties. Seasonal differences in soil characteristics were statistically significant (p < 0.001), confirming a strong seasonal signal (Table 3).

Pearson’s correlation analysis revealed strong and consistent relationships among SOC, soil texture fractions, and BD, across both summer and winter seasons (

Table 4. Pearson’s correlation matrix of measured soil properties.

Soil Properties		Soil Organic Carbon		Sand		Silt		Clay		Bulk Density
		%								g cm−3
		Summer	Winter	Summer	Winter	Summer	Winter	Summer	Winter	Summer	Winter
SOC	Summer	1
	Winter	0.996 **	1
Sand	Summer	−0.557 **	−0.564 **	1
	Winter	−0.556 **	−0.564 **	0.999 **	1
Silt	Summer	0.469 *	0.475 **	−0.845 **	−0.845 **	1
	Winter	0.470 *	0.476 **	−0.843 **	−0.846 **	0.996 **	1
Clay	Summer	0.523 **	0.531 **	−0.939 **	−0.937 **	0.609 **	0.608 **	1
	Winter	0.521 **	0.529 **	−0.938 **	−0.937 **	0.607 **	0.606 **	1.000 **	1
BD	Summer	−0.285	−0.274	0.287	0.290	−0.194	−0.207	−0.300	−0.297	1
	Winter	−0.274	−0.263	0.278	0.278	−0.194	−0.201	−0.287	−0.284	0.997 **	1
−1–−0.8	−0.8–−0.6	−0.6–−0.4		−0.4–−0.2	−0.2–0	0–0.2	0.2–0.4	0.4–0.6		0.6–0.8	0.8–1
Correlation is significant				Correlation is not significant				Correlation is significant

** Correlation is significant at the significance level of α = 0.01, * Correlation is significant at the significance level of α = 0.05.

). SOC measured in summer and winter was perfectly correlated (r = 0.996, p < 0.01), indicating high temporal stability in spatial SOC patterns despite seasonal variation. SOC exhibited a significant negative correlation with sand content in both seasons (summer: r = −0.557; winter: r = −0.564; p < 0.01). In contrast, positive correlations were observed between SOC with silt (r ≈ 0.47; p < 0.05–0.01) and clay content (r ≈ 0.52–0.53; p < 0.01) (Table 4). Sand content was strongly and negatively correlated with both silt and clay (r = −0.84 to −0.94; p < 0.01). BD exhibited weak and non-significant correlations with SOC in both seasons (−0.26 to −0.29), although BD was highly stable seasonally (r = 0.997; p < 0.01) (Table 4).

3.3. Soil Organic Carbon Stock (SOCs)

The SOCs calculations were conducted using the 23 soil samples corresponding to four LULC classes—including agricultural land, barren land, dry sabkha, and wet sabkha—based on 2021 LULC classified map (Figure 6). The four LULC classes collectively cover approximately 42% of the study area, excluding fish farms, urban areas, water bodies, and natural vegetation classes in 2021 classified map (Table 2). The spatial distribution of these SOCs is shown in Figure 6, and the proportional coverage of each SOCs class is summarized in

Table 5. Area coverage by square kilometer (km²) and percentage (%) of soil organic carbon stock (SOCs) across four 2021 LULC classes in summer and winter seasons.

SOCs (Mg C ha−1)	Land Use/Land Cover Classes
	Agricultural Land		Barren Land		Dry Sabkha		Wet Sabkha
	km2	%	km2	%	km2	%	km2	%
	Summer
<3	31.00	6.09	10.23	37.08	9.02	31.84	26.18	48.65
3–5	263.69	51.79	9.40	34.09	11.03	38.97	19.78	36.76
5–7	146.89	28.85	7.95	28.83	8.26	29.19	7.85	14.59
7–9	46.84	9.20	---	---	---	---	---	---
9–12	20.72	4.07	---	---	---	---	---	---
	Winter
<3	4.92	0.97	6.16	22.32	5.67	20.03	17.24	32.04
3–5	139.84	27.47	11.20	40.59	10.37	36.41	18.02	33.47
5–7	220.24	43.26	9.40	34.09	11.16	39.41	18.26	33.92
7–9	94.58	18.58	0.83	3.00	1.11	3.92	0.31	0.57
9–12	19.42	3.81	---	---	---	---	---	---
12–14	30.09	5.91	---	---	---	---	---	---
Total	509.10	100.00	27.59	100.00	28.31	100.00	53.82	100.00

. This distribution provides representative coverage of the LULC classes influencing regional SOCs dynamics. Although the overall sampling density is moderate relative to the total study area, the stratified coverage across geomorphological units supports the representativeness of SOCs estimates and the broader applicability of the methodological workflow to similar deltaic and fragmented coastal environments.

Based on the spatial distribution of SOCs across the four LULC classes (Figure 6), agricultural land dominates the eastern and western parts of the study area, whereas barren land and sabkha are confined to the northern region. SOCs varied markedly among LULC classes and between summer and winter seasons (Figure 6 and Table 5).

In summer, agricultural land exhibited the highest spatial extent of moderate SOCs; the largest proportion (51.8%) of area fell in the 3–5 Mg C ha⁻¹ class. Lower SOCs classes (<3 Mg C ha⁻¹) accounted for 6.1% of agricultural land, while higher SOCs classes (7–9 and 9–12 Mg C ha⁻¹) together comprised only 13.3% of the area. Conversely, barren land and sabkha were dominated by low SOCs classes (<3 and 3–5 Mg C ha⁻¹). For barren land in summer, 37.1% of area was in the <3 Mg C ha⁻¹ class and 34.1% in the 3–5 class, with no representation above 7 Mg C ha⁻¹ (Figure 6 and Table 5). Both dry and wet sabkha showed similar distributions, with the majority of their areas concentrated in the lower classes: for dry sabkha, 31.8% of area had <3 Mg C ha⁻¹ and 38.9% had 35 Mg C ha⁻¹, while wet sabkha had 48.7% and 36.8% in the same classes, respectively.

In winter, SOCs shifted toward higher classes across all LULC types (Figure 6 and Table 5). Agricultural land showed a notable increase in areas within the 5–7 Mg C ha⁻¹ category (43.3%), and the emergence of higher classes (7–9 and 9–12 Mg C ha⁻¹) totaling >22% of its area. The highest SOCs class (12–14 Mg C ha⁻¹) is only observed in the winter season and represents 5.9% of total agricultural land (Figure 6 and Table 5). Barren land and sabkha also exhibited increases in the 5–7 Mg C ha⁻¹ category (34.1–39.4%) compared with summer and show representation in the 7–9 Mg C ha⁻¹ class, albeit at lower proportions (<4%). The lowest class (<3 Mg C ha⁻¹) markedly decreased in all LULC types in winter compared to summer (Figure 6 and Table 5). Total mapped areas were 509 km² for agricultural land, 27.6 km² for barren land, 28.3 km² for dry sabkha, and 53.8 km² for wet sabkha.

SOCs exhibited clear seasonal differences between summer and winter (

Table 6. Descriptive statistical parameters for the calculated SOCs (Mg C ha⁻¹) (number of soil samples= 23).

Parameters		Season
		Summer	Winter
Minimum		0.61	1.60
Maximum		11.36	13.75
Mean		4.78	5.83
Standard Deviation		2.80	3.17
Standard Error		0.58	0.66
Percentile	25%	2.83	3.65
	50%	4.28	5.30
	75%	6.38	7.79
Skewness		0.83	1.04
Kurtosis		0.65	1.01
p-value		<0.001 *
Pearson Correlation		0.997 **

* p-value is significant at significance level of α = 0.05, ** correlation is significant at the significance level of α = 0.01.

). SOCs ranged from 0.61 to 11.36 Mg C ha⁻¹ in summer and from 1.60 to 13.75 Mg C ha⁻¹ in winter, indicating a wider and generally higher range of C storage during the winter season. Mean SOCs increased from 4.78 Mg C ha⁻¹ in summer to 5.83 Mg C ha⁻¹ in winter, representing a statistically significant seasonal difference (p < 0.001). The SOCs distribution showed moderate variability, with standard deviations of 2.80 Mg C ha⁻¹ (summer) and 3.17 Mg C ha⁻¹ (winter). Median values (50^th percentile) increased from 4.28 Mg C ha⁻¹ in summer to 5.30 Mg C ha⁻¹ in winter, while upper quartile values (75th percentile) rose from 6.38 to 7.79 Mg C ha⁻¹, further highlighting enhanced SOC storage during winter.

Both seasonal datasets were positively skewed (skewness = 0.83–1.04) with slightly positive kurtosis (0.65–1.01), indicating a right-tailed distribution with a small number of high SOCs values. Despite the significant seasonal difference in magnitude, SOC stocks measured in summer and winter were almost perfectly correlated (r = 0.997, p < 0.01), suggesting strong temporal stability in the spatial ranking of SOCs across sampling sites.

3.4. Estimated SOCs and PSOCS till 2100

SOCs and PSOCs were predicted to 2100 under four SSP2-4.5 climate scenarios, reflecting progressive increases in temperature (+1.5 to +2.5 °C) and PET (+3% to +12%) and decreasing precipitation (−5% to −20%). The baseline is a calculated SOCs in 2021. The minimum, maximum, and mean values of projected SOCs (Mg C ha⁻¹) and PSOCS (Mg C ha⁻¹) at 10-year intervals are presented in

Table 7. Minimum, maximum, and mean of projected SOCs (Mg C ha⁻¹) and PSOCS (Mg C ha⁻¹) with 10-year intervals up to 2100 under four SSP2-4.5 climate scenarios (SSP_Sce1, SSP_Sce2, SSP_Sce3, and SSP_Sce4).

Scenario	Parameter		2030	2040	2050	2060	2070	2080	2090	2100
SSP_Sce1	SOCs (Mg C ha−1)	Minimum	1.67	1.68	1.69	1.69	1.68	1.68	1.68	1.67
		Maximum	14.36	14.45	14.49	14.51	14.50	14.48	14.45	14.40
		Mean	6.10	6.13	6.15	6.16	6.15	6.14	6.13	6.11
	PSOCS (Mg C ha−1)	Minimum	0.07	0.08	0.09	0.09	0.08	0.08	0.08	0.07
		Maximum	0.61	0.70	0.74	0.76	0.75	0.73	0.70	0.65
		Mean	0.26	0.30	0.32	0.32	0.32	0.31	0.30	0.28
SSP_Sce2	SOCs (Mg C ha−1)	Minimum	1.67	1.68	1.68	1.68	1.68	1.67	1.67	1.66
		Maximum	14.36	14.44	14.47	14.47	14.45	14.42	14.37	14.30
		Mean	6.09	6.13	6.14	6.14	6.13	6.12	6.09	6.07
	PSOCS (Mg C ha−1)	Minimum	0.07	0.08	0.08	0.08	0.08	0.07	0.07	0.06
		Maximum	0.61	0.69	0.72	0.72	0.70	0.67	0.62	0.55
		Mean	0.26	0.30	0.31	0.31	0.30	0.28	0.26	0.23
SSP_Sce3	SOCs (Mg C ha−1)	Minimum	1.67	1.68	1.68	1.68	1.67	1.66	1.65	1.64
		Maximum	14.36	14.43	14.44	14.43	14.39	14.33	14.25	14.16
		Mean	6.09	6.12	6.13	6.12	6.10	6.08	6.05	6.01
	PSOCS (Mg C ha−1)	Minimum	0.07	0.08	0.08	0.08	0.07	0.06	0.05	0.04
		Maximum	0.61	0.68	0.69	0.68	0.64	0.58	0.50	0.41
		Mean	0.26	0.29	0.29	0.29	0.27	0.25	0.21	0.17
SSP_Sce4	SOCs (Mg C ha−1)	Minimum	1.67	1.68	1.68	1.67	1.66	1.65	1.64	1.63
		Maximum	14.35	14.41	14.42	14.39	14.34	14.27	14.17	14.06
		Mean	6.09	6.12	6.12	6.11	6.08	6.05	6.01	5.96
	PSOCS (Mg C ha−1)	Minimum	0.07	0.08	0.08	0.07	0.06	0.05	0.04	0.03
		Maximum	0.60	0.66	0.67	0.64	0.59	0.52	0.42	0.31
		Mean	0.26	0.28	0.29	0.27	0.25	0.22	0.18	0.13

.

Across the four scenarios, the minimum SOCs values exhibit minimal variation (1.63–1.69 Mg C ha⁻¹), indicating consistent lower bounds across all SSPs and thresholds. Similarly, maximum values vary only slightly, ranging from 14.06 to 14.51 Mg C ha⁻¹. Mean SOCs values remain largely stable across all scenarios and thresholds, with a narrow range of 5.96–6.16 Mg C ha⁻¹ (Table 7). The highest mean values occur under the SSP_Sce1 and SSP_Sce2 scenarios, particularly at the 2050 and 2060 thresholds. The lowest mean is observed under SSP_Sce4 in 2100 (5.96 Mg C ha⁻¹), which may reflect reduced average performance under the highest intensity for SSP_Sce4. A slight declining trend in mean SOCs values is also evident, particularly for SSP_Sce3 and SSP_Sce4 (Table 7). Correspondingly, the minimum PSOCS values display very limited variation (0.03–0.09 Mg C ha⁻¹), whereas maximum values show greater variability, ranging from 0.31 to 0.76 Mg C ha⁻¹. Mean PSOCS values range between 0.13 and 0.32 Mg C ha⁻¹ (Table 7).

Annual SOCs (Figure 7A) and PSOCS (Figure 7B) were simulated until 2100 using four SSP2-4.5 climate scenarios. The baseline starting with calculated SOCs in 2021 with a consistent value of 5.83 Mg C ha⁻¹ across all scenarios. Under all scenarios, SOCs initially increased during the early decades of the century before plateauing and eventually declining toward 2100 (Figure 7A). Scenario 1 (Sce.1), with the mildest climate change, produced the highest SOCs retention and sequestration rates until around 2050, with SOCs increasing from 5.83 to a peak of 6.16 Mg C ha⁻¹ (Figure 7A). Also, PSOCS rising to 0.32 Mg C ha⁻¹ (Figure 7B). After 2050, Sce.1 exhibited a gradual stabilization of SOCs and PSOCS at mid-century before a slight decline toward the end of the century (6.11, and 0.28 Mg C ha⁻¹ by 2100, respectively).

Progressively stronger warming and moisture stress in Scenario 2 (Sce.2) and Scenario 3 (Sce.3) attenuated SOC accumulation. Peak SOCs under Sce.2 reached 6.14 Mg C ha⁻¹ (mid-century) and declined to 6.07 Mg C ha⁻¹ by 2100. The PSOCS peaking at 0.31 Mg C ha⁻¹ at mid-century and reducing to 0.23 Mg C ha⁻¹ by 2100 (Figure 7A,B). Sce.3, driven by greater temperature increases and precipitation reductions, showed a lower peak (6.13 Mg C ha⁻¹) and stronger decline in sequestration potential to 0.18 Mg C ha⁻¹ by century’s end (Figure 7A,B).

Under the most intense perturbation (Scenario 4, Sce.4), initial SOCs was smallest and reversed earlier, with SOC peaking at 6.13 Mg C ha⁻¹ and declining continuously to 5.96 Mg C ha⁻¹ by 2100. Correspondingly, PSOCS peaked near 0.29 Mg C ha⁻¹ and dropped to 0.13 Mg C ha⁻¹ by 2100. Overall, both absolute SOCs and PSOCS were highest under the mildest climate change scenario and increasingly depressed with greater warming and drying, indicating the strong sensitivity of SOCs dynamics to climatic stressors.

3.5. Monte Carlo Approach of Projected SOCs and PSOCS

Monte Carlo outputs represented median SOC and PSOCS trajectory and associated 5th and 95th percentile uncertainty (

Table 8. Descriptive statistical parameters for the Monte Carlo uncertainty approach of projected SOCs using SSP2-4.5 climate scenarios.

Soil Organic Carbon Stock (Mg C ha−1)		Scenario 1			Scenario 2			Scenario 3			Scenario 4
		Median	5th	95th	Median	5th	95th	Median	5th	95th	Median	5th	95th
Minimum		6.03	4.06	8.69	6.00	4.02	8.61	6.01	4.05	8.68	6.01	4.03	8.63
Maximum		6.27	4.51	9.10	6.22	4.47	9.00	6.22	4.48	9.05	6.22	4.50	9.04
Mean		6.23	4.40	8.84	6.18	4.36	8.77	6.17	4.35	8.82	6.16	4.36	8.75
Standard Deviation		0.05	0.10	0.13	0.05	0.10	0.12	0.05	0.12	0.12	0.06	0.12	0.10
Standard Error		0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01
Variance		0.00	0.01	0.02	0.00	0.01	0.02	0.00	0.01	0.01	0.00	0.01	0.01
Percentile	25%	6.21	4.32	8.73	6.15	4.30	8.65	6.14	4.26	8.71	6.12	4.27	8.66
	50%	6.24	4.43	8.82	6.19	4.39	8.74	6.19	4.39	8.79	6.18	4.40	8.73
	75%	6.27	4.48	8.95	6.21	4.45	8.88	6.21	4.46	8.93	6.21	4.47	8.82
Skewness		−1.99	−0.98	0.49	−1.54	−1.08	0.37	−1.15	−0.72	0.40	−0.76	−0.69	0.71
Kurtosis		4.06	0.54	−1.13	2.65	0.88	−1.19	0.84	−0.66	−1.29	−0.50	−0.62	−0.16

and

Table 9. The descriptive statistical parameters for the Monte Carlo uncertainty approach of projected PSOCS using SSP2-4.5 climate scenarios.

Potential SOCS (Mg C ha−1)		Scenario 1			Scenario 2			Scenario 3			Scenario 4
		Median	5th	95th	Median	5th	95th	Median	5th	95th	Median	5th	95th
Minimum		0.20	−1.77	2.86	0.17	−1.81	2.78	0.18	−1.79	2.84	0.18	−1.80	2.80
Maximum		0.44	−1.33	3.27	0.39	−1.37	3.17	0.39	−1.36	3.22	0.39	−1.33	3.21
Mean		0.40	−1.44	3.01	0.34	−1.47	2.93	0.34	−1.48	2.99	0.33	−1.47	2.92
Standard Deviation		0.05	0.10	0.13	0.05	0.10	0.12	0.05	0.12	0.12	0.06	0.12	0.10
Standard Error		0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01
Variance		0.00	0.01	0.02	0.00	0.01	0.02	0.00	0.01	0.01	0.00	0.01	0.01
Percentile	25%	0.38	−1.51	2.89	0.32	−1.54	2.82	0.31	−1.57	2.88	0.29	−1.56	2.83
	50%	0.41	−1.40	2.99	0.36	−1.44	2.91	0.35	−1.45	2.95	0.34	−1.44	2.89
	75%	0.44	−1.35	3.12	0.38	−1.38	3.05	0.38	−1.38	3.10	0.38	−1.36	2.99
Skewness		−1.92	−0.98	0.48	−1.54	−1.08	0.37	−1.15	−0.72	0.40	−0.76	−0.69	0.71
Kurtosis		3.73	0.52	−1.12	2.63	0.86	−1.19	0.86	−0.67	−1.29	−0.49	−0.62	−0.17

, respectively) at a monthly time step from 2022 to 2100 per each location (number of soil samples = 23). For each scenario, the model outputs were annually averaged per location and then for the entire area at the whole time step. The final parameters were statistically described in Table 8 and Table 9 for projected SOCs and PSOCS, respectively.

The Monte Carlo uncertainty analysis revealed highly consistent projected SOCs across all four SSP2-4.5 climate scenarios (Table 8). The median average of projected SOCs ranged from 6.16 to 6.23 Mg C ha⁻¹ (Table 8). The 5th–95th percentile uncertainty intervals were broad but overlapping for all scenarios, spanning 4.4–8.8 Mg C ha⁻¹. Minimum and maximum median SOCs values showed limited spread among scenarios, with minimum values around 6.00–6.03 Mg C ha⁻¹ and maximum values between 6.22 and 6.27 Mg C ha⁻¹ (Table 8). Accordingly, standard deviations of the median SOCs were low (0.05 Mg C ha⁻¹), with 0.01 Mg C ha⁻¹ corresponding standard errors (Table 8). There was negative skewness for the median across all scenarios (−0.76 to −1.99), indicating a longer lower tail. These distributional characteristics suggest that median SOCs outcomes are mostly stable.

Projected PSOCS followed patterns similar to SOCs (Table 9). The median average of projected PSOCS ranged from 0.33 to 0.40 Mg C ha⁻¹ across scenarios. Despite positive median values, uncertainty bounds were wide, with negative 5th percentiles (−1.44 to −1.48 Mg C ha⁻¹) and positive 95th percentiles (3.0 Mg C ha⁻¹) (Table 9), indicating a substantial probability of both SOC gains and losses over the projection period. For SOCs, differences among scenarios were small. Standard deviations and variances were around 0.0 across scenarios, with left-skewed distributions (Table 9). Overall, these results demonstrate that parametric uncertainty dominates over climate-scenario uncertainty in determining long-term SOC trajectories and sequestration potential in the modeled system.

4. Discussion

4.1. Soil Properties

Descriptive statistics (Table 3) indicated soil characteristics were not normally distributed. Sand content showed a positive skewness, indicating right-sided asymmetry. However, other characteristics exhibited a negative skewness, reflecting left-sided asymmetry distribution. Similar trends were observed for kurtosis, where sand had positive kurtosis, indicated sharper distribution peaks. Whereas other characteristics showed flatter peaks with negative kurtosis. These deviations from normality are common in environmental datasets [91,92,93]. Means and standard deviations for key soil properties are reported in Table 3 to quantitatively support these observations.

Correlation matrix (Table 4) revealed a significant positive correlation between SOC and fine soil fractions (silt and clay), reflecting the role of mineral-associated OC in SOC stabilization. Fine-textured soils afford greater surface area and physicochemical protection for OM, which enhances SOC retention [94,95]. Conversely, SOC was negatively correlated with sand content, reflecting faster OM turnover and weaker SOC stabilization [96]. Supporting these findings, Abu-Hashim et al. [43,97] reported that the clay loam soil (with saturated hydraulic conductivity ranging from 4.5−7.9 cm d⁻¹ and higher clay content) had the highest SOC pool mean of 7.07 kg C m², whereas the lowest SOC pool of 2.57 kg C m² was observed in bare soil with a sandy clay loam texture. Bakr and El-Ashry [98] also demonstrated strong correlation between SOM and soil texture, mainly sand and clay fractions. In contrast, Onah et al. [99] observed positive correlation between SOM and sand, but negative correlation with silt and clay.

Seasonal analysis indicated minimal variation in soil texture and SOC, emphasizing the inherent resistance of these properties to short-term climatic fluctuations. Their dynamics appear to be governed primarily by long-term LULC and management changes [100]. SOC was negatively correlated with BD, reflecting improved aggregation and porosity in OM-rich soils [101]. Conversely, BD showed a positive association with sand content and significant seasonal variability, likely driven by changes in soil moisture, tillage, and wetting–drying cycles [102].

4.2. Long-Term LULC Change

The expansion of agricultural land and urban areas reflected sustained population growth, land reclamation efforts, and infrastructure development over the past five decades. Similar trends reported across deltaic and coastal agricultural systems in North Africa and the Middle East [38,41]. The conversion of barren land and sabkha into cropland highlighted intensive human intervention in a land-scarce region. Since 1990s, rapid fish farm expansion has marked a major land-use shift driven by increasing food demand, economic incentives, and declining freshwater availability for traditional agriculture [39].

Fluctuations in the Burullus Lake and natural vegetation reflected interactions among hydrological management, land reclamation, and climate variability. Declines in open water areas align with agricultural and aquaculture encroachment [36,39]. The reduction in barren land, dry sabkha, and wet sabkha from approximately 50% of the study area in the 1970s–1980s to only 7% by 2021 reflects substantial landscape transformation. This change likely resulted from a combination of land reclamation, agricultural expansion, and hydrological variability, including fluctuations in surface water extent. While this transition indicates increased land productivity, it may also reduce natural buffering capacity against flooding, salinity intrusion, and coastal erosion [12,103]. These long-term LULC changes strongly influence SOC dynamics. While conversion to cropland and fish farm can initially increase SOC through enhanced biomass inputs, long-term outcomes depend on management, salinity stress, and hydrological conditions [45,104]. The five-decade LULC analysis offered a critical basis for interpreting SOC patterns and improving SOCs projections under climate change, highlighting the need to integrate land-use planning with climate adaptation in vulnerable delta systems [2,105].

4.3. SOCs Impact on LULC

The LULC-based analysis reveals clear seasonal dynamics and land-cover influences on SOCs (Table 5). Changes in LULC and management strongly impact SOCs and topsoil C storage [67,72]. Higher SOCs in winter across all LULC classes is consistent with reduced decomposition rates and increased soil moisture under cooler conditions, which enhance OM preservation [106].

Barren land and sabkha were dominated by low SOC classes (<5 Mg C ha⁻¹). This reflected limited OM inputs and physiochemical constraints imposed by salinity, alkalinity, and shallow soil profiles. These constraints collectively reduce plant productivity and restrict carbon stabilization processes [107,108,109,110]. In contrast, agricultural lands exhibited higher proportions of intermediate and high SOCs classes, driven by greater organic inputs that enhance C stabilization [111]. The pronounced winter increase in the 5–7 Mg C ha⁻¹ class suggests that seasonal moisture and temperature shifts exert a stronger influence on SOCs in cultivated soils than in barren and sabkha. These findings are consistent with observations in Mediterranean and semi-arid agroecosystems [112,113].

Generally, SOCs distributions reflect the combined influences of climate seasonality, vegetation cover, and soil properties. Winter conditions promote SOC retention, whereas higher temperatures in summer enhance SOC mineralization [114,115,116]. Additionally, increased SOCs in winter lead to increases in the dissolved organic carbon concentrations in adjacent water streams (irrigation and drainage canals) due to runoff influence [117]. Positive SOCs skewness indicates spatial heterogeneity, typical of managed and dryland soils [118]. These findings emphasize the need to account for both seasonal variability and texture-based stabilization processes when targeting PSOCS and parameterizing process-based models such as RothC under future climate scenarios [25,92,119].

4.4. Projected SOCs and PSOCS and Uncertainty Analysis

Future SOCs projections were conducted under a BAU assumption in which post-2021 LULC and management practices remain constant, allowing isolation of climate-driven effects on C dynamics. While this approach reduces compounding uncertainties and clarifies the influence of projected climate change, it does not account for potential future LULC transitions that could substantially modify SOCs trajectories. Incorporating dynamic LULC and management scenarios in future studies would provide a more comprehensive assessment of long-term SOCs responses in rapidly changing deltaic environments.

While LULC change exerts a dominant influence on the spatial distribution of current SOCs, the climate projections applied in this study indicate that rising temperatures and changes in precipitation under SSP2-4.5 will increasingly regulate C stability in the study area. Despite the observed slight increase in projected SOCs, rising temperatures under SSP2-4.5 are expected to accelerate microbial decomposition and increase soil respiration rates, thereby enhancing SOC mineralization and potentially offsetting gains associated with land conversion [8]. Experimental and modeling studies consistently show that SOC turnover exhibits strong temperature sensitivity, particularly in moisture-limited or saline systems, where warming can disproportionately destabilize protected C pools [120]. In addition, projected shifts in precipitation patterns may modify soil moisture regimes, further influencing decomposition kinetics and carbon input dynamics [8].

The close agreement of projected SOCs and PSOCS across the four SSP2-4.5 climate scenarios suggests that climate-scenario uncertainty has a relatively minor influence on long-term SOCs compared with uncertainty in model parameters and inputs. This is consistent with previous RothC-based studies showing that C inputs, soil texture, and decomposition parameters exert stronger control on SOCs than moderate climate differences [121,122,123]. The broad and overlapping uncertainty ranges indicate that plausible variations in C inputs, clay content, and turnover rates can outweigh the effects of differing climate trajectories, particularly under constant management assumptions, as also reported in regional and global SOC modeling studies [15,88,116,124]. Although median PSOCS values were positive across all scenarios, negative lower uncertainty bounds indicate a non-negligible risk of SOCs decline, highlighting the asymmetric nature of sequestration potential. This finding aligns with syntheses emphasizing the high uncertainty for future SOC sequestration, especially where C inputs are marginal or decomposition accelerates under warming [32,66,82,125].

Overall, the study demonstrates the importance of Monte Carlo uncertainty analysis for robust SOCs and PSOCS projections. The uncertainty ranges reported in this study reflect variability in input data, parameter estimates, and initial conditions within a fixed RothC model structure. This approach quantifies parametric uncertainty only. It does not account for structural uncertainty arising from differences in model formulation and process representation across models, which can contribute substantially to variation in projected carbon dynamics [126,127]. Scenario-related uncertainty was partially explored using multiple SSP2-4.5 climate realizations. However, uncertainty associated with other SSP pathways or future LULC transitions was not evaluated. This scenario uncertainty is a widely recognized source of projection spread in environmental models [128]. Therefore, the reported uncertainty bounds should be interpreted as conditional to the selected model structure and scenario assumptions, rather than capturing the full range of possible future SOC outcomes.

5. Conclusions

This study provides an integrated assessment of five decades of LULC change, baseline SOCs, and future SOCs and PSOCS dynamics in the northern Nile delta. Analysis of Landsat imagery from 1972 to 2021 revealed substantial agricultural expansion, urban growth, and fish farm development, accompanied by marked declines in barren land and sabkha areas. While LULC conversion largely explains the present spatial distribution of SOCs, future SOCs trajectories are increasingly affected by projected climate change.

RothC simulations under four SSP2-4.5 climate realizations indicate a mid-century increase in SOCs, followed by slight declines toward 2100. This pattern reflects the interacting effects of climate forcing, where rising temperatures may enhance microbial decomposition and C turnover, potentially offsetting gains from improved land productivity. Although projected changes remain slight in magnitude, they signal growing long-term vulnerability of SOCs under sustained warming, particularly in coastal and saline deltaic systems.

The results highlight that SOCs resilience in the Nile delta depends not only on land-use configuration but also on adaptive management strategies capable of buffering climate-driven C losses. Sustainable soil management, salinity control, and climate-informed land planning will therefore be essential to maintain positive PSOCS and long-term carbon stability. By integrating LULC dynamics with climate projections, this study provides policy-relevant insights for regional C management and supports the development of climate-adaptive LULC strategies in vulnerable delta environments.