Due to their unique geographical and climatic conditions, Central Asia and China’s Xinjiang region face serious soil salinization issues (Jiang et al., 2019). The study area selected for this thesis includes the five Central Asian countries and the Xinjiang Province of China, covering latitudes from 55°27′40″N to 34°20′10″N and longitudes from 46°29′30″E to 96°23′35″E, as shown in Figure 1 . Due to their unique geographical and climatic conditions, Central Asia and the Xinjiang region face serious soil salinization issues. Central Asia, including Kazakhstan, Uzbekistan, Turkmenistan, Tajikistan, and Kyrgyzstan, features a typical continental climate with vast plains and deserts. This results in higher evaporation than precipitation, high groundwater levels, and significant salt accumulation. The region predominantly comprises extensive plains and deserts, with the plains mainly located in the Syr Darya and Amu Darya river basins, and the deserts spread across the southern and eastern parts of Central Asia (Hamidov et al., 2016). Xinjiang, located in northwestern China, has an arid climate and complex terrain. The natural conditions, including the Tianshan and Kunlun mountain ranges and the vast Taklamakan Desert, further exacerbate soil salinization (Zhuang et al., 2021). Xinjiang is characterized by mountains, basins, and deserts, with the Tianshan and Kunlun Mountain ranges stretching from east to west. The Tarim Basin lies in the central part, which contains the vast Taklamakan Desert (Duan et al., 2022b). Irrational irrigation and drainage systems have exacerbated this issue, leading to reduced crop yields, soil structure degradation, and deterioration of the ecological environment, thereby affecting regional economic and sustainable agricultural development (Wang et al., 2019). Therefore, researching the causes, current status, and prevention measures of soil salinization in these areas is of significant importance for enhancing agricultural productivity and the quality of the ecological environment (Peng et al., 2019).

When vegetation is subjected to soil salinity stress, it is generally believed that the plant’s photosynthetic capacity decreases, and the SIF value correspondingly decreases (Song et al., 2018). However, the corresponding SIF value may not necessarily show significant differences when soil salinity changes significantly over time. Phenological changes, vegetation type, climatic conditions such as precipitation and temperature, and soil characteristics such as moisture and texture are also important drivers of SIF variation (Hamidov et al., 2016). Therefore, it is necessary to filter the samples to eliminate the influence of other factors on SIF and ensure that soil salinity is the dominant factor in SIF observations. Specifically, to avoid the impact of land use changes on SIF, we excluded samples where land use types changed during the study period (2000-2020). Next, we collected samples at 0.05-degree intervals and extracted the corresponding SIF observation values. Finally, we used multiple comparisons to test whether there were significant differences (p< 0.05) in SIF observations under different salinity conditions.

Du et al. (2024) developed a method to standardize SIF observations based on time-series data (2000-2020) and probability distribution functions. The standardized Solar-Induced Chlorophyll Fluorescence Index (SIFI) was calculated from the time series of SIFI at different temporal scales. This SIFI was then used to develop a soil salinization model. This standardization method was initially used to normalize precipitation indices to characterize meteorological droughts. In this study, the improved standardized Solar-Induced Chlorophyll Fluorescence Index (SIFI) was calculated from the SIF observation time series at a monthly scale from January to December. The calculation method for the improved SIFI is as follows:

First, the SIF time-series data is fitted to a probability distribution, in this case, a Gamma distribution, to account for its non-negativity and skewness. The probability density function of the Gamma distribution is defined as:

Here, x represents the SIF value, while α and β are the shape and scale parameters, respectively, and Γ is the Gamma function. The parameters α and β are obtained through maximum likelihood estimation (MLE) from the data. After estimating the parameters, the cumulative distribution function (CDF) of the Gamma distribution is calculated, yielding the cumulative probability G(x) for each SIF observation:

Next, the cumulative probability G(x) is transformed into the corresponding quantile of the standard normal distribution, which gives the SIFI value:

The subscript i represents the time scale of SIFI. For example, SIFI1, SIFI2, and SIFI6 correspond to SIFI calculated from SIF observations over time scales of one month, two months, and six months, respectively.

We used satellite-based Solar-Induced Chlorophyll Fluorescence (CSIF) observation data to perform sensitivity analysis on all sample points. Figure 3 shows the sensitivity of SIF values to soil salinity in the five Central Asian countries and the Xinjiang region of China. The results show that 44.7% of the sensitive points were typical sample points in Kazakhstan, indicating many susceptible sample points in this region, followed by Xinjiang, which accounted for 29.4%. In contrast, the proportion of typical sample points in Turkmenistan, Uzbekistan, Kyrgyzstan, and Tajikistan was 12.9%, 8%, 3.5%, and 1.5%, respectively. Non-Typical Sample Points were primarily distributed in Kazakhstan (53.4%) and Xinjiang (28.6%). Spatial distribution analysis indicates that typical sample points in Kazakhstan are predominantly concentrated in its southern region, while those in China’s Xinjiang are primarily located in the northern areas. This distribution pattern may be associated with the spatial extent of core irrigated agricultural zones in both regions: Southern Kazakhstan and Northern Xinjiang represent significant agricultural hubs within their respective territories. Large-scale irrigation under arid/semi-arid climatic conditions may contribute to the development of secondary soil salinization, potentially resulting in heightened sensitivity of vegetation photosynthesis to soil salinity changes in these areas. Turkmenistan exhibits a significantly higher number of typical sample points compared to non-typical points, suggesting a potentially strong association between SIF values and soil salinity in this region, with the model demonstrating relatively favorable performance here. This observation may reflect Turkmenistan’s distinctive environmental context—characterized by extreme aridity intense evapotranspiration, and an irrigated oasis agriculture system. These factors collectively may drive widespread and severe soil salinization. Within relatively sparse vegetation and potentially less complex environmental stress regimes in desert/saline-alkali landscapes, SIF signals may experience reduced interference from confounding biophysical factors. Consequently, the SIF response to soil salinity—as a key stressor—might appear more discernible, potentially enhancing model performance. Conversely, Tajikistan shows a higher proportion of non-typical sample points, and the statistical correlation between SIF values and soil salinity is markedly weaker than in other study regions. This indicates that the SIF response to soil salinity variations may be relatively weak or non-significant in this area. Potential underlying reasons for this discrepancy may lie in Tajikistan’s complex mountainous terrain: Fragmented topography drives substantial heterogeneity in hydrothermal conditions and diverse local microclimates. Vegetation growth and associated SIF signals may be predominantly influenced by, or interact strongly with, topography-controlled water availability, temperature stress, variations in vegetation community types, or other environmental factors not fully quantified. The presence of these complex factors may partially explain the weaker-than-expected association between observed SIF values and soil salinity, posing challenges to model applicability in this region.

Figure 4 reveals significant temporal differences in the sensitivity of SIF to soil salinity across the five Central Asian countries and Xinjiang, China, which are closely related to seasonal climate conditions and land-use activities in each region. We found that April is the most sensitive month for all countries. This suggests that in spring, the response of land-use types to changes in soil salinity and model accuracy is most pronounced because increasing temperatures and moisture availability may enhance soil salinity effects, making them easier to detect. Additionally, spring is typically a period of increased agricultural activity, with environmental changes driven by both natural and human-induced factors, making SIF data more sensitive to salinity changes during this month. April was the most sensitive month in Xinjiang, Turkmenistan, and Tajikistan, demonstrating that spring plays a crucial role in land-use types in these regions. In contrast, Kazakhstan was most sensitive in December, Uzbekistan in June, and Kyrgyzstan in November. These differences reflect each country’s specific land-use patterns and climate condition changes during these months. For example, the sensitivity in Uzbekistan in June may be related to higher temperatures and soil moisture variations in early summer, while the high sensitivity in Kazakhstan in December may be linked to cold-season moisture retention and salinity accumulation before deep winter. The distribution of sensitive months also showed significant differences across various terrain types. In Kazakhstan, rainfed cropland, forest, and grassland were most sensitive in November, while herbaceous cover, irrigated cropland, and shrubland showed the highest sensitivity in December, and wetlands peaked in September. This illustrates the different response characteristics of terrain types to seasonal changes, reflecting their varying adaptability to environmental conditions. In Xinjiang, most terrain types, including rainfed cropland, herbaceous cover, irrigated cropland, shrubland, grassland, sparse vegetation, and wetlands, showed peak sensitivity in April, while only forests were most sensitive in December. This consistent sensitivity pattern suggests that April is a critical transition period for soil moisture and salinity effects in Xinjiang. Rainfed cropland in Uzbekistan was most sensitive in June, while herbaceous cover peaked in May, irrigated cropland in October, and forest in April. Shrubland and grassland exhibited peak sensitivity in April and June, respectively, while sparse vegetation and wetlands were most sensitive in November and February. These variations indicate that seasonal changes in soil moisture and vegetation growth cycles significantly impact salinity sensitivity. In Kyrgyzstan, the overall peak sensitivity occurred in November. However, different land cover types exhibited varied sensitivity peaks: rainfed cropland, irrigated cropland, and forest were most sensitive in January, herbaceous cover in June, and shrubland, grassland, sparse vegetation, and wetlands in November. These patterns suggest that the interactions between soil salinity and environmental factors are strongly influenced by winter conditions and early growing season changes. In Turkmenistan, the sensitivity peak in April was particularly notable in rainfed cropland, herbaceous cover, and forest, while irrigated cropland was most sensitive in December, shrubland in October, grassland in November, sparse vegetation in January, and wetlands in March. These results indicate that salinity fluctuations in different land cover types are influenced by distinct seasonal processes, including irrigation patterns, evaporation rates, and precipitation regimes. In Tajikistan, most land cover types, including rainfed cropland, irrigated cropland, forest, shrubland, and grassland, exhibited peak sensitivity in April. In contrast, herbaceous cover, sparse vegetation, and wetlands showed the highest sensitivity in May. These findings suggest that spring is a critical period for soil salinity dynamics in Tajikistan, likely due to increased precipitation, snowmelt, and early agricultural activities.

( Figure 5a ) shows the distribution of classification accuracy in different regions for 2016 based on Solar-Induced Chlorophyll Fluorescence (SIF) data. The study found that in most regions, the classification models exhibited high accuracy, further validating the effectiveness and reliability of SIF technology in large-scale soil salinity monitoring. Regarding classification accuracy in typical regions, the model performed well in Tajikistan, which had the highest model accuracy, followed by Kyrgyzstan and Xinjiang, both of which also demonstrated high accuracy. This indicates that SIF data can effectively distinguish the spatial variability of soil salinity in these regions, and the models in these areas have predictive solid capabilities when dealing with typical samples. In contrast, the classification model accuracy in typical regions of Uzbekistan and Turkmenistan was relatively lower. This is related to the diversity of soil types, vegetation cover, and the complexity of land-use patterns in these regions, which resulted in more significant classification errors. In non-typical regions, Kyrgyzstan and Tajikistan again demonstrated high classification accuracy, particularly when handling complex environmental samples, showcasing their models’ strong adaptability and reliability. Xinjiang also performed well, indicating that SIF data in non-typical regions of this area had high monitoring capabilities. However, Turkmenistan’s classification accuracy in non-typical regions was significantly lower, indicating that the model in this region has apparent deficiencies when addressing complex environmental changes, requiring further parameter optimization or the integration of other data sources to improve accuracy. ( Figure 5b ) shows the 3D scatter plot distribution of overall accuracy in various regions. The study found that the spatial distribution of accuracy exhibits some geographical variability. Overall, areas with higher classification accuracy were mainly concentrated in Xinjiang, Tajikistan, and Kyrgyzstan, while regions with lower accuracy were primarily distributed in Turkmenistan and Uzbekistan.

This study evaluated the accuracy of classification models for typical and non-typical regions in the five Central Asian countries (Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, and Uzbekistan) and the Xinjiang region of China. By analyzing the 2016 regional divisions and model accuracy data, the median classification accuracy for each region was calculated, allowing for a comparison of model performance across the countries. Due to the imbalance in salinity categories during the modeling process, the estimated models produced more conservative predictions. Most regional models exhibited high classification accuracy, indicating that SIF technology can effectively monitor changes in soil salinity. This result suggests that SIF can be a reliable remote sensing tool for large-scale soil salinity monitoring. The classification accuracy rankings for typical regions were as follows: Tajikistan > Kyrgyzstan > Xinjiang > Kazakhstan > Turkmenistan > Uzbekistan. This indicates that the classification models for Tajikistan, Kyrgyzstan, and Xinjiang performed well in typical regions, exhibiting high accuracy, while the classification performance in Uzbekistan and Turkmenistan needs improvement. The classification accuracy rankings for non-typical regions were Kyrgyzstan > Tajikistan > Xinjiang > Kazakhstan > Uzbekistan > Turkmenistan. Kyrgyzstan and Tajikistan performed exceptionally well in the non-typical areas, showing that their models have high accuracy when dealing with complex environments. In contrast, the classification accuracy in the non-typical Turkmenistan regions was significantly lower, indicating substantial deficiencies in the model’s classification performance in this region. We further analyzed the model’s performance across typical and atypical regions. A comparative analysis of Tables 2 and 3 reveals that the soil salinity classification model based on solar-induced chlorophyll fluorescence (SIF) demonstrates robust performance in typical sample regions. In particular, the F1-scores for the “Non” and “Slightly” saline categories reached 0.95 and 0.88, respectively, with an overall accuracy of 0.85. These results suggest that the model can effectively capture the nonlinear relationship between SIF signals and soil salinity levels in areas with stable spectral responses and sufficient vegetation cover. In contrast, the classification performance declines notably in atypical sample regions, where the overall accuracy drops to 0.75, and the F1-scores for the “Moderately,” “Highly,” and “Extremely” saline classes are significantly lower. This performance degradation can be attributed to the complex environmental conditions in atypical areas, including sparse vegetation, heterogeneous soil backgrounds, and frequent fluctuations in soil water and salt dynamics. Moreover, the number of training samples for the “Highly” and “Extremely” saline classes is considerably limited in these regions, which constrains the model’s ability to learn their discriminative features effectively and leads to insufficient recognition accuracy for these minority classes. Figure 6 provides a visual comparison of the classification results between typical and atypical regions through confusion matrices. As shown in ( Figure 6a ), the majority of samples in the typical region are correctly classified, especially in the “Non” and “Slightly” categories, which exhibit minimal misclassification. In contrast, ( Figure 6b ) illustrates increased confusion among the moderate to extreme salinity classes in the atypical region, confirming the model’s limited ability to generalize under less stable and underrepresented conditions. Nevertheless, the performance gap between typical and atypical regions provides meaningful insights into the model’s reliability and stability. The high accuracy achieved in typical samples indicates that the proposed SIF-based feature framework is capable of delivering accurate predictions in well-structured and representative regions. Meanwhile, the intentional construction of atypical sample sets and the observed performance drop therein serve as a valuable approach to assess the model’s generalization boundaries and application scope. This validation strategy, which incorporates sample representativeness as a key factor, enhances the interpretability of model performance and offers a practical framework for evaluating classification models under complex ecological conditions.

( Figures 7a, b ) are generated based on the SIFI index and regional models by analyzing the relationship between SIF data and soil salinity, creating soil salinity distribution maps for salt-affected and saline lands. The results show that Kazakhstan has the largest area of salt-affected and saline lands, particularly in the southern part of the country where irrigation-induced secondary salinization dominates, significantly more than other nations. Additionally, northern Xinjiang characterized by endorheic basins amplifying salt accumulation and southern Turkmenistan exhibiting capillary-driven salt efflorescence also exhibit relatively high areas of salt-affected land. Northern Xinjiang and southern Turkmenistan also have relatively large areas of salt-affected land. SIF observation results can identify the distribution of salt-affected lands (EC ≥ 2~4 dS m^-1) in Central Asia and Xinjiang demonstrating moderate-salinity detection capability. According to ( Figure 7c ), Kazakhstan has the largest total area of affected saline land, reaching 56.54 Mkm² reflecting widespread irrigation legacy impacts, indicating that salinization is the most severe in this country. Next are Turkmenistan and Xinjiang, with 25.60 Mkm² tied to intensive cotton monoculture and 24.50 Mkm² concentrated in northern piedmont oases respectively. Uzbekistan has 13.66 Mkm² of saline land primarily in the Fergana Valley, while Kyrgyzstan and Tajikistan have relatively more minor areas, with 2.12 Mkm² limited by mountainous drainage and 1.79 Mkm² restricted by steep topography respectively. This distribution difference indicates that Kazakhstan faces the most severe soil salinization problem, affecting a significant portion of its land, whereas Kyrgyzstan and Tajikistan are relatively less affected. ( Figure 7d ) shows the data for saline land (EC ≥ 4~8 dS m^-1), with Kazakhstan having 4.41 Mkm² of saline land notably near Aral Sea disaster zones, followed by Xinjiang with 1.34 Mkm² associated with paleo-salt deposits and Uzbekistan with 0.38 Mkm². Turkmenistan has 0.23 Mkm² of saline land focused along canal networks, while Kyrgyzstan and Tajikistan have 0.17 Mkm² confined to valley bottoms and 0.15 Mkm² in localized depressions respectively.

The contributions of different SIFI indices to the model vary; therefore, we analyzed the feature importance of each SIFI index across different months to explore its influence over time. This study employed a random forest model to analyze the feature importance of SIFI1-SIFI12 variables across different months, revealing significant differences in predictive capability across temporal scales. As shown in the Figure 8 , some variables, such as SIFI1 and SIFI12, consistently exhibited high importance throughout the year, suggesting that they may carry stable and critical environmental information and contribute significantly to soil salinity prediction. In contrast, other variables, such as SIFI6 and SIFI8, had relatively low importance in certain months, indicating that their influence might be restricted to specific time windows. Additionally, in months like February, June, and September, some variables exhibited considerable fluctuations in importance, implying that the predictive ability of SIFI indices during these periods could be affected by external environmental factors. Overall, the temporal evolution trend indicates substantial variations in the importance of specific variables across different months. In long-term prediction tasks, it is advisable to prioritize stable key variables that maintain high importance throughout the year while employing time-series-based dynamic modeling strategies for variables whose importance fluctuates over time to optimize predictive performance. Furthermore, the temporal variability in feature importance suggests that SIFI indices may be influenced by seasonal climate conditions or other environmental factors. Future studies could integrate external environmental variables, such as precipitation and temperature, to further investigate their impact on SIFI indices.

We used Solar-Induced Chlorophyll Fluorescence (SIF) observation data from all sample points in Central Asia and Xinjiang between 2000 and 2020 to create a soil salinity change map for 2000-2020 ( Figure 9 ). The results show a significant increase in soil salinity in southern Kazakhstan and around Lake Balkhash, while the salinity in northern Kazakhstan has decreased. The soil salinity in western Xinjiang also showed an increasing trend. This is consistent with the findings of Issanova et al. (Issanova et al., 2017). The increase in soil salinity in southern Kazakhstan and southern Central Asia primarily occurs in bare land, shrubland, and irrigated cropland. These land types are sensitive to environmental changes and prone to salinization. This may be related to reduced local precipitation, increased evaporation, and improper irrigation practices (Lobell et al., 2011). In contrast, the decrease in soil salinity in northern Kazakhstan may reflect improvements in agricultural management practices, such as adopting more efficient irrigation technologies, reducing over-cultivation, and promoting natural vegetation restoration (Zhang et al., 2010). The increasing trend in soil salinity in western Xinjiang may be closely related to agricultural expansion, improper water resource management, and secondary salinization caused by irrigation in the region (Zaitunah et al., 2018). More than 60% of the sample points in Central Asia and Xinjiang show an increasing trend in soil salinity, further confirming that the expansion of saline land has become a widespread phenomenon in these regions. Soil salinization poses a serious threat to agricultural production and may trigger a range of ecological and environmental issues, such as vegetation degradation, soil erosion, and the deterioration of soil structure (Tarolli et al., 2024). These issues are particularly pronounced in arid and semi-arid regions, as soil salinization often exacerbates water scarcity, further impacting regional ecological balance and sustainable development (Doan et al., 2023). The spatiotemporal variation of soil salinity is influenced by multiple factors, including climate change, human activities, and natural geographic features (Haj-Amor et al., 2023). The increase in salinity in southern Kazakhstan and around Lake Balkhash may be related to reduced precipitation and increased evaporation caused by global warming, and it could also be the result of agricultural expansion and improper irrigation (Tarolli et al., 2024). On the other hand, the decline in salinity in northern Kazakhstan may be related to improvements in agricultural management strategies in recent years, such as introducing more scientific irrigation techniques and natural vegetation restoration measures (Rakhmanov et al., 2024). The increase in salinity in western Xinjiang also warrants attention, as this phenomenon may be due to improper water resource management and secondary soil salinization caused by irrigation during the region’s agricultural development (Bai et al., 2023). Therefore, improving irrigation techniques, increasing soil organic matter, promoting salt-tolerant crops, and encouraging vegetation restoration are effective methods to slow the process of salinization (Liu et al., 2022). At the same time, using advanced remote sensing technologies like SIF for long-term monitoring allows for precise tracking of the dynamic changes in soil salinity and provides vital reference data for formulating more scientific and practical soil management policies. This will help ensure the sustainability of regional ecosystems and support the long-term stable development of agricultural production (Aburas et al., 2015).

Using Solar-Induced Chlorophyll Fluorescence (SIF) data from 2000 to 2020, combined with machine learning methods and the Standardized Solar-Induced Chlorophyll Fluorescence Index (SIFI), a systematic analysis of the spatiotemporal changes in soil salinity across the five Central Asian countries and the Xinjiang region of China was conducted. It was found that the increase in soil salinity poses a severe threat to agricultural production, with significant differences in the degree and trends of soil salinization across the countries. The increase in soil salinity is typically negatively correlated with crop growth inhibition (Shi et al., 2023). Optical and thermal remote sensing technologies effectively monitor this process, especially by using data that reflect vegetation biochemical and physical characteristics, such as leaf water content, leaf area, and chlorophyll content, which can indirectly infer soil salinity levels (Jiang and Shu, 2019). Chlorophyll fluorescence technology is particularly effective, as salinity stress directly inhibits plant photosynthesis, affecting the chlorophyll fluorescence signal (Zaghdoudi et al., 2011). Additionally, by measuring the content of ions such as sodium, potassium, and calcium in the leaves, as well as leaf conductivity, the effects of soil salinity on plants can be further reflected and assessed (Munns and Tester, 2008). These physiological response parameters provide direct evidence of plant health and serve as effective indicators for monitoring soil salinization. Between 2000 and 2020, the area of salt-affected land (soil electrical conductivity, EC ≥ 2–4 dS·m⁻¹) in Tajikistan fluctuated significantly, ranging from 20% to 60%, as shown in Figure 10 . This fluctuation may be closely related to Tajikistan’s climate conditions, land-use patterns, and changes in agricultural management practices (Orlovsky and Orlovsky, 2002). In contrast, the area of salt-affected soil in Xinjiang, China, has remained relatively stable, with fluctuations around 15%. This relative stability in Xinjiang may be attributed to the region’s stricter water resource management and ongoing land improvement measures (Yang et al., 2019). However, the area changes of saline land (EC ≥ 4~8 dSm⁻¹) in various countries have been more drastic than in salt-affected land, highlighting the severity of soil salinization in Central Asia and Xinjiang. For example, the fluctuation range of saline land in Tajikistan is between 50% and 150%, far exceeding that of other countries. This may be related to the country’s extreme climatic conditions, limited water resources, and improper irrigation practices (Turdaliev et al., 2023). In contrast, the relatively stable changes in saline land in Kazakhstan may reflect the progress the country has made in recent years in agricultural management and improvements in irrigation techniques (Duan et al., 2022a). Kyrgyzstan and Tajikistan exhibited significant interannual fluctuations, with sharp alternations in the extent of both salt-affected and saline lands, reflecting high sensitivity to climate variability and instability in land use and irrigation management. In particular, Tajikistan showed annual variation rates in saline land exceeding 100% in multiple years, highlighting its high risk under extreme climatic conditions and improper irrigation practices. Uzbekistan experienced considerable fluctuations in the early years of the study period, but gradually showed a trend of stabilization after 2010. In contrast, Turkmenistan continued to exhibit intense year-to-year variability, with changes in saline land reaching up to 150%, likely associated with periodic accumulation and leaching cycles of salt. The Xinjiang region displayed clear cyclical patterns, and since 2010, saline land has steadily decreased, indicating the positive effects of long-term governance and mitigation efforts. These interannual variation patterns are valuable for distinguishing between short-term, climate-driven transient salinization and structural salinization caused by poor water resource management, thus providing a basis for region-specific differentiated control strategies. Overall, from 2000 to 2020, the dynamics of soil salinization in the five Central Asian countries and the Xinjiang region exhibited significant spatial heterogeneity and temporal variability. Countries such as Kazakhstan and Xinjiang have shown a certain degree of effectiveness in governance and trend stability, whereas Tajikistan and Turkmenistan still face frequent fluctuations and high-intensity salinization pressure. Such disparities are closely related not only to differences in water resource regulation, agricultural management, and irrigation efficiency, but also to the long-term impacts of climatic conditions and land use structures. Therefore, when formulating regional soil salinization control strategies, it is essential to fully consider each area’s climatic vulnerability, governance capacity, and the underlying mechanisms of salt accumulation. Tailored, locally adapted management approaches should be adopted to ensure the sustainable use of land resources and the resilient development of agricultural ecosystems.

The modeling contribution of standardized solar-induced chlorophyll fluorescence indices (SIFI1–SIFI12), derived from annual SIF time series, was evaluated across eight representative land cover types. This approach produces phenologically interpretable indicators that characterize the statistical structure of SIF variations and emphasize their differential responsiveness across diverse land cover types, enabling robust comparisons among ecosystems with distinct vegetation characteristics. Figure 11 reveals that SIFI1 consistently exhibits high contribution across all land cover types, underscoring its robustness as a general sensitivity indicator. Its wide-ranging applicability may stem from its ability to capture early-season physiological activity or reflect persistent structural traits of vegetation that influence SIF dynamics throughout the year. In addition, SIFI4, SIFI5, SIFI7, and SIFI8 show particularly strong contributions in rainfed cropland, herbaceous cover, irrigated cropland, and shrubland. These indices correspond to the mid-growing season, typically associated with peak photosynthetic activity from April to August. During this period, vegetation in these land types is especially responsive to environmental stressors such as salinity, drought, or nutrient limitations, which enhances the explanatory power of SIF-based indicators. In contrast, SIFI10, SIFI11, and SIFI12 exhibit higher contributions in forest and sparse vegetation, indicating their sensitivity to ecological dynamics in late autumn and early winter. This seasonal window is critical for capturing declines in photosynthetic activity, changes in carbon assimilation efficiency, and other long-term traits of perennial or woody vegetation systems.