Lakes found in semi-arid and arid regions are crucial to the sustenance of water resources and local climate regulation, along with the facilitation of ecological processes. In addition to being reservoirs for domestic and agricultural purposes, lakes are vital and sensitive evaluators of environmental and climate changes (Lyons & Shengm, 2017; Bai & Wang, 2023). Currently, these systems are threatened by climate change and human activities, including reduced precipitation, rising temperatures, salinization, pollution, and land-use shifts. All of these likely accelerate the deterioration of ecosystems and scarcity of water (Laounia et al., 2017; Chávezet al., 2009). In arid basins, artificial recharge, diversions, and flood control implemented as management interventions further alter hydrological systems, thereby leading to complex trade-offs for availability of water and ecosystem services as well as the ability of these systems to sustain themselves in the long term (Chen et al., 2025; Zhang et al., 2024).

The impact of climate change, coupled with increased desertification and hydrological instability, can turn once permanent lakes into seasonal or even ephemeral ones (Huang et al., 2022; Alsumaieiet al., 2025). Human activities such as the construction of upstream dams and withdrawals for irrigation have resulted in worldwide shrinkage of major inland water bodies (Oroud, 2019). Salinization has further deteriorated water quality and aquatic ecosystems, and directly impacted biodiversity, water security, agriculture, and the regional economy. (Bastiaanssen et al., 1998; Wanget al., 2015). The combined effects of climate change and human activities on these water bodies have increased the need to monitor lakes in arid and semi-arid regions and to develop effective management strategies.

Lake surface temperature (LST) influences both its physical and ecological aspects by governing the vertical and horizontal distribution of water in the lake, its biological productivity, and even the rate of water loss through evaporation (Abiret al., 2024; Utami, & Kim, 2023). Previous investigations have revealed that the surface of lakes, especially shallow lakes, experience more rapid warming than the air temperature, which consequently enhances the evaporation rates of lakes in regions with prolonged drought (Woolwayet al., 2018; Zhouet al., 2021). In these areas, evaporation is typically the primary mechanism of water loss from the lake, and is often greater than the total annual precipitation, thereby leading to considerable reductions in lake volume (Li et al., 2009; Limaet al., 2020). Recent studies have indicated that mechanisms of water loss are likely to vary as one moves between different dryland areas. This indicates the importance of evaluating the surface temperature and evaporation patterns of lakes accurately (Malekiet al., 2024).

These challenges are particularly evident in western Iraq, which has an arid to semi-arid climate characterized by low precipitation and high evaporation (Al-Daghistani, 2003; Ali et al., 2021). Over the years, this region has experienced increased temperatures coupled with decreased precipitation which has resulted in lower river flow and severe drought and water stress (Micijevic et al., 2019). Furthermore, restrictions placed on the flow of the Tigris and Euphrates rivers, along with climate change, have created additional challenges for lakes and reservoirs that are critical to sustain irrigation, flood control, and ecosystem services (Sharma et al., 2024; Singhet al., 2024). Notably, remote sensing studies demonstrate that the surface area and storage volume of many Iraqi lakes are shrinking, thus, illustrating the extreme impacts of climate change and anthropogenic impacts (Ahmood et al., 2024; Holman et al., 2023).

In Iraq, satellite-based remote sensing has significantly contributed to the temporal and spatial monitoring of inland water bodies. However, previous hydrological studies of Lake Tharthar have primarily focused on surface area dynamics and volumetric fluctuations (Forootan, 2019; Fadel & Najiban, 2024). Although some recent attempts have attempted to estimate evaporation rates using remote sensing (Forootan, 2019; Biggset al., 2016). these studies have largely relied on simplified empirical equations, low-resolution climate networks, or general land-temperature products without accurate correlation to actual water surface temperatures (Ahmood et al., 2024). Consequently, the lack of continuous and high-resolution integration between satellite-derived lake surface temperatures and local terrestrial climate observations of Lake Tharthar remains a critical research gap (Matta et al., 2022).

To bridge this gap, to the best of our knowledge, this study is the first to develop a precise and highly decomposed framework that integrates multi-source thermal remote sensing with local terrestrial meteorological data. This approach directly adapts to microclimatic feedback and the thermal hysteresis of Lake Tharthar, thereby providing more dynamic and accurate evaporation estimates compared to previous evaporation models.

This study aimed to analyze the surface temperature of Tharthar Lake in the west of Iraq and discover the spatiotemporal evaporation. Landsat 9 thermal datasets and ground-based data were used to obtain this data. Subsequently, the METRIC surface energy balance model was adopted for further investigation. This model enabled evaluation of the lake surface evaporation by constructing a physically based evaporation model. Furthermore, it verified the satellite data of evaporation by using machine learning-based validation. This offered valuable insight into future evaporation trends in relation to the management of the data-poor and climate-affected national water resources. Importantly, this study creates a framework that can be adopted for measuring lake evaporation in dry climates worldwide.

Figure 1 illustrates the overall methodological framework used to estimate and evaluate lake evaporation using an internally calibrated surface energy balance model. The first stage of the methodology involves the acquisition and preprocessing of three primary data sources: Landsat 9 satellite data, specifically the optical bands of the OLI-2 sensor and the thermal band 10 of the TIRS-2 sensor, spatially adjusted to a resolution of 30 meters. Hourly climate readings from the Tikrit ground station, including downward solar radiation, air temperature, relative humidity, and wind speed, are also utilized, coupled with historical records of basin evaporation and climate data obtained from the Tikrit station for evaluation and validation purposes.

To ensure that the open water area is evaluated exclusively, without influence from the adjacent land, a dynamic water mask was constructed based on the Modified Water Differences Index (MNDWI) with a threshold value exceeding 0.10, combined with a Quality of Data (QA_PIXEL) band to mask clouds and shadows. Thereafter, atmospheric correction of the spectral radiation recorded in the thermal band was performed using radiative transport data based on the MODTRAN model and NCEP reanalysis data. A constant emissivity of 0.990 for pure water was assigned to derive the surface temperature of the lake via Planck's inversion.

The next stage of the methodology involved calculating the fundamental components of the surface energy budget, namely latent evaporation. Net radiation was calculated based on a hydroalpido-coefficient between 0.05 and 0.06. To facilitate the handling of the high heat capacity of water, hydrothermal flux was calculated as a dynamically variable ratio of net radiation between 0.30 and 0.50. Further, aerodynamic impedance was designed to calculate the perceived heat flux, while assuming an open-water momentum angle roughness coefficient of 0.0003 m.

This study employed an internal calibration technique by establishing two reference points in each spatial snapshot to automatically eliminate thermal and atmospheric biases. The first sample comprised a "cold pixel" taken in deep, clear water areas where perceptible heat flow was almost nonexistent and evaporation is at its maximum. The second sample consisted of a "hot pixel" collected in the dry, bare land surrounding the lake where evaporation was negligible. This enabled the extraction of linear temperature gradient features and supported the model calculations.

During the temporal scaling phase, the instantaneous evaporation ratio was calculated relative to the reference evaporation, which was estimated using the standard Penman-Monteath equation (ASCE). Assuming that this ratio is stable throughout the hours of the day, the calculations were extended to extract the total daily evaporation of the lake in millimeters per day. The methodology concluded with an uncertainty and conformity assessment, in which 80% of the data were used for training while the remaining 20% for testing, with accuracy evaluated using artificial intelligence algorithms (EEDNN, Random Forest, SAM, and LGBM) based on the statistical measures of mean squared error (MSE), root mean squared error (RMSE), and Nash-Soutcliffe efficiency (NSE).

The findings of this study demonstrate that atmospheric forcing is the primary control mechanism for lake evaporation in arid contexts. Lake Tharthar is a lucid example of the interaction between temperature, humidity, and wind, and their joint effect on energy transfer from the surface to the atmosphere. Seasonal patterns indicate that evaporation is impacted by climatic drivers in a non-linear, episodic manner. Evaporation is strongest when the conditions are extremely hot and dry, and is weakest when the conditions are cool and humid. This aligns with the established and well-supported theoretical and empirical frameworks of the evaporation processes in arid and semi-arid contexts, where the demand for atmospheric moisture is often greater than the supply (Siadatmousavi & Seyedalipour, 2019; Lobos-Roco et al., 2021).

Moreover, the results of this study reveal the non-uniform nature of evaporation across Lake Tharthar. Although each spatial pattern example is unique, they are all considerably affected by lake morphology, thermal inertia, and exposure to the atmosphere. There is a clear evaporation preference in the shallow, southern, and peripheral parts of the lake compared to the deeper, central parts of the lake. This is likely due to shorter evaporation rates, susceptibilities, and the heat retention capacity of the lake. Across various lakes, studies have been demonstrated that shallow water bodies (especially the peripheral regions of lakes) serve as evaporation hotspots. This is attributed to rapid evaporation from the lake surface due to differences in heating, synthesis of energy, and its interaction with the atmosphere (Shao et al., 2020; Holman et al., 2023). The findings of the present study are consistent with this pattern and further support previous observations that shallow lake margins can function as evaporative hotspots.

The impact of wind on evaporation is spatially differentiated and accentuated by wind fetch. In areas of high fetch, wind promotes turbulence, which increases the evaporation rate from the lake surface. It has been demonstrated that the evaporation rate from the lake may differ due to the varying exposure of the lake to wind, even under a relatively homogeneous climate in the region. Sugita et al. (2020) and Farooq et al. (2022) suggest that evaporation may be enhanced in a localized area due to differing vapor pressure, surface roughness, and uneven roughness of the lake surface. It is evident from these findings that the evaporation processes of the lake are best captured through modeling the surface of the lake. Simplified methods, which represent a unit of space as a single point or use coarse aggregate values, fundamentally misrepresent the evaporation from large lakes that exhibit a high degree of morphological and spatial variability.

These results validate METRIC for modeling lake evaporation in different environments, including arid, semi-arid, and agricultural regions (Danodia et al., 2024; Iqbal et al., 2024). Particularly, data-scarce countries benefit from the internal calibration method using hot and cold pixels as it does not depend on extensive meteorological networks, while also providing high accuracy (Allen et al., 2007; Acharya et al., 2020). Unlike other energy balance approaches, METRIC can provide reliable and accurate estimates for evaporation in agricultural and semi-arid areas with a patchy land cover (Khand et al., 2021; Mondal et al., 2022; Rahimi et al., 2021).

The integration of METRIC and machine learning–based validation within the methodological framework strengthens the design by allowing the uncertainty and performance to be expressed comprehensively across various modeling approaches. Moreover, this framework advances the interpretability of the modeling results and provides a valid framework for using satellite-based estimates of evaporation in the planning and management of water resources where in situ measurements are sparse or lacking.

Therefore, the framework described here for the estimation of evaporation from Lake Tharthar in Iraq is innovative and extremely useful. However, despite its strengths, the methodology also has a few limitations. The first and most notable limitation is the use of meteorological data from a single weather station. This is likely to overlook the microclimatic variability within the area of Lake Tharthar, and particularly within the considerable area of this region. The challenge remains to better represent the spatial atmospheric gradients by using reanalysis products or by extending ground-based observation networks.

Second, the 16-day revisit cycle for Landsat data is likely to mask the evaporative response from synoptic weather events. Moreover, the spatial and atmospheric resolution of the data may be improved further by using MODIS or Sentinel data. Data fusion methodologies have exhibited promising results in combining high spatial resolution with high temporal data.

Third, drawing representative hot and cold pixels adds some level of subjectivity that could impact model results, even though METRIC internal calibration lessens the need for field data collection. Thus, future studies should focus on hybrid calibration techniques to diminish uncertainties further. Additionally, the inclusion of eddy covariance or scintillometry data to enhance validation efforts would add more focused bounds to energy flux estimates. This multi-faceted approach will optimize lake evaporation estimation and enable the application of adaptive water management techniques to react to evolving climatic conditions.

To compare these results with their regional and broader context, the evaporation dynamics of Lake Tharthar were evaluated in comparison with previous studies conducted in water bodies with extremely arid and semi-arid environmental, and in geographical conditions in the Middle East and North Africa. The pronounced seasonal variation in LST, ranging from below 15 °C in winter to above 40 °C in shallow areas in summer, is consistent with findings reported for regional lakes such as Lake Urmia in Iran (Tasumi, 2019; Javadian et al., 2019) and the Dead Sea (Oroud, 2020), where high solar irradiance and severe vapor pressure deficits contribute to extreme evaporative responses (Farooq et al., 2025). In contrast to reservoirs in semi-arid Mediterranean climates, such as Lake Qaraoun in Lebanon (Mhawej et al., 2020) and Lake Oubira in Algeria (Rezzag Bara et al., 2019), where coastal humidity limits water loss (Zhao et al., 2022), Lake Tharthar experiences accelerated evaporation due to strong dry winds, reaching 31.7 m/s, and summer temperatures reaching up to 47 °C. The observed spatial pattern, in which shallow lake margins form evaporation hotspots relative to the deeper central basin, can be attributed to morphometric differences in water heat capacity and thermal inertia. These findings are consistent with the physical models proposed by Shao et al. (2020) and Zhang et al. (2024).

More broadly, the findings of this study are directly linked to several United Nations Sustainable Development Goals. The precise quantification of annual evaporation losses (exceeding 2.93 billion cubic meters) contributes to Goal 6: Clean Water and Sanitation (sub-goal 6.4) by providing accurate data to improve water use efficiency and manage water scarcity in the strategic reservoirs of Iraq. Furthermore, identifying the high climate sensitivity of the lake supports Goal 13: Climate Action (sub-goal 13.1) by offering a practical remote sensing tool to enhance adaptation and strengthen climate resilience against desertification and drought. Finally, identifying the shallow shorelines most vulnerable to rapid degradation provides a scientific baseline for wetland protection under Goal 15: Life on Land (sub-goal 15.3), thereby supporting efforts to combat land degradation and maintain ecosystems in arid environments.

This study applied the METRIC model using thermal data from the Landsat 9 satellite and combined it with terrestrial meteorological observations to estimate evaporation rates and assess heat exchange between the lake and atmosphere across seasons. By leveraging advanced satellite thermal remote sensing techniques, this study provides a robust framework for measuring heat fluxes in data-poor arid environments, such as Iraq.

These results emphasize the important role of evaporation in water loss from Lake Tharthar and its implications for water resource management in Iraq. Accurately quantifying evaporation is necessary to ensure the operation of reservoirs, plan flood control, and allocate water over the long term under increasing climatic stress. This technique generates spatially distinct information that can be used to support targeted management strategies, such as prioritizing monitoring efforts in high-loss zones and assessing the feasibility of evaporation-reduction measures.

To better represent spatial atmospheric gradients, future research should aim to integrate higher-temporal-resolution satellite data in order to capture short-term variability in evaporation, expand ground-based meteorological networks, and explore data-fusion approaches that combine physical models with advanced machine learning techniques. These advancements will enhance adaptive water management strategies and facilitate accurate evaporation monitoring in arid and semi-arid regions experiencing growing water scarcity.