During the wheat maturity period from 2022 to 2023, sampling was conducted in 11 typical wheat-growing counties/cities (3 in 2022 and 8 in 2023) across eight prefectures in Xinjiang ( Figure 1 ). In each county/city, five representative wheat-growing villages were selected, and two typical wheat fields were chosen per village. Within each sampling plot, wheat was manually cut close to the ground from a 5 m² area using the “five-point sampling method.” Surface dust and other weight-affecting impurities were removed before the samples were packed and transported to the laboratory for harvest index determination.

All seeds were manually separated from the wheat plants and dried in an oven at 105°C for 30 minutes. The oven temperature was then adjusted to 75°C until a constant weight was achieved. The wheat yield and aboveground biomass were weighed and recorded separately. The wheat harvest index (HI) was calculated as follows (Bilandžija et al., 2023; Du et al., 2024):

where “Yield” refers to wheat yield (kg/m²), and “AGB” represents aboveground biomass (kg/m²).

Geographical coordinates (longitude(lon), latitude(lat) and altitude (alt)) of sampling sites were recorded in situ using Huace X90-series GPS receivers during wheat sampling. For soil accumulated temperature (SAT), we calculated the total effective accumulated soil temperature during the wheat growing season (March to July) using daily temperatures ≥ 10°C. All temperature data were obtained from the National Meteorological Science Data Center (http://data.cma.cn/).

Plant height (H): At wheat harvest, more than five plots were selected in each sampling area, with over 10 wheat plants randomly measured per plot using a 1 mm-precision steel tape. The height from the plant base to the apical growing point was recorded, and the average across all plots represented the plant height for that sampling area. Wheat yield (yield): The dry weight of wheat grains from each 5 m² sampling plot was measured to calculate yield per unit area. Aboveground biomass (AGB): The harvested 5 m² wheat samples were oven-dried to constant weight to determine AGB per unit area. Irrigation (irrigation): Based on farmers’ cultivation practices, seasonal water availability, and the arid climate, wheat irrigation regimes were classified as either irrigated or rainfed.

Soil samples (0–20 cm depth) were collected from each plot using a soil auger following the “five-point sampling method”. After homogenization, air-drying under natural conditions, and removal of visible debris, samples were ground for analysis. The measurement protocols for soil nutrients were as follows: Soil Organic Matter (SOM, g/kg): Potassium dichromate oxidation-external heating method; Total Nitrogen (TN, g/kg): Sulfuric acid-perchloric acid digestion (Foss Kjeltec 1035 auto-analyzer); Total Phosphorus (TP, g/kg): Acid digestion-molybdenum antimony colorimetry (Agilent Cary 60 UV-Vis spectrophotometer); Total Potassium (TK, g/kg): Acid digestion-atomic absorption spectrometry (Thermo Scientific S-series AAS); Alkali-hydrolyzable Nitrogen (ASN, mg/kg): Alkali diffusion method; Available Phosphorus (AP, mg/kg): Sodium bicarbonate extraction-molybdenum antimony colorimetry (Agilent Cary 60 UV-Vis spectrophotometer); Available Potassium (AK, mg/kg): Ammonium acetate extraction-atomic absorption spectrometry (Thermo Scientific S-series AAS); pH: Measured using a pH meter (PHS-2F); Soil Electrical Conductivity (EC, mS/cm): Determined with a conductivity meter (DDSJ-308F); Soil Water Content (SWC, g/cm³): Measured by oven-drying core samples (Φ200) at 105°C to constant weight; Bulk Density (BD, g/cm³): Calculated from the dry weight of core samples (Φ200) per unit volume.

Based on Pearson correlation analysis and hierarchical clustering results ( Figure 4 ), the association characteristics between wheat harvest index (HI) and environmental factors in arid regions were revealed: Agronomic management factors: HI showed significant positive correlations with yield, aboveground biomass (AGB), and irrigation (p < 0.01), but a significant negative correlation with plant height (H) (p < 0.01). Soil nutrient factors: HI exhibited significant positive correlations with total phosphorus (TP), available phosphorus (AP), and soil water content (SWC) (p < 0.01). Geographic-climatic factors: Only longitude (lon) demonstrated a significant negative correlation with HI (p < 0.01). The hierarchical clustering dendrogram ( Figure 4 ) further indicated that: AP clustered most closely with HI, suggesting that available phosphorus is the most critical driver of HI. HI, AP, SWC, irrigation, AGB, and yield formed a core group of key driving factors, highlighting their synergistic effects on HI. Geographic-climatic factors formed a distinct cluster, reflecting their indirect influence on HI through interactions with other variables.

For irrigated wheat, the harvest index (HI) showed a significant positive correlation (p < 0.01) only with yield among the agronomic management factors ( Figure 5a ). In contrast, for rainfed wheat, HI exhibited: A significant positive correlation (p < 0.01) with plant height (H) among agronomic factors ( Figure 5b ). Significant positive correlations (p < 0.01) with bulk density (BD), total potassium (TK), total nitrogen (TN), available phosphorus (AP), and available soil nitrogen (ASN) among soil nutrient factors. A significant negative correlation (p < 0.01) with soil water content (SWC). The hierarchical clustering dendrogram further revealed distinct patterns: Irrigated wheat HI did not cluster closely with other factors but instead formed a major group with longitude (lon) and latitude (lat), indicating that geographical location is a key driver of HI in irrigated wheat systems. Rainfed wheat HI clustered most closely with AP, forming a distinct subgroup that further grouped with ASN, H, TN, TK, and BD, collectively representing the dominant influencing factors for rainfed wheat HI. The effects of different factor types on the overall, irrigated, and rainfed wheat Harvest Index (HI) exhibited significant differences in their statistical significance ( Table 2 ).

A Random Forest model was employed to analyze the relative importance and contribution values of various factors affecting wheat Harvest Index (HI) in arid regions, aiming to identify critical influencing factors. Given that HI is directly calculated from yield and aboveground biomass (AGB), these two variables were excluded from the Random Forest analysis to avoid circular reasoning. The results demonstrated that the model explained 47.4% of the variance in overall wheat HI across arid regions ( Figure 6a ). Plant height (H, 13.27%) and total nitrogen (TN, 12.42%) emerged as key drivers (importance >10%), followed by available phosphorus (AP, 9.62%), total phosphorus (TP, 7.33%), soil water content (SWC, 6.83%), latitude (lat, 6.69%), soil available potassium (SAT, 6.52%), altitude (alt, 5.74%), bulk density (BD, 5.62%), and longitude (lon, 5.28%), classified as secondary factors (importance 5-10%).

For irrigated wheat systems ( Figure 6b ), the hierarchy of factor importance was: longitude (lon, 8.37%), plant height (H, 7.96%), latitude (lat, 6.10%), bulk density (BD, 5.13%) (importance >5%). In contrast, rainfed wheat systems ( Figure 6c ) exhibited distinct prioritization: total nitrogen (TN, 12.20%), available phosphorus (AP, 10.48%), total potassium (TK, 5.77%), total phosphorus (TP, 5.61%) (importance >5%). While soil nutrient factors collectively dominated in importance across all systems, the relative contributions of different factor types (geographic, climatic, and management) to HI variation diverged markedly among overall, irrigated, and rainfed wheat systems ( Table 3 ).

Based on the key variables identified through prior analyses, we employed Structural Equation Modeling (SEM) to investigate the interactions among the three factor types (geographic-climatic, soil nutrient, and agricultural management) and quantify their direct and indirect contributions to the Harvest Index (HI). This approach enables systematic visualization of causal pathways, effectively disentangling the complex interdependencies between these factors and HI through standardized path coefficients and model fit indices. We selected: All factors from Geographic-climatic factors (GCF). Plant height (H) as the representative of wheat plant characteristics (plant). The top five most important factors from soil nutrient factors (soil). We then constructed structural equation models (SEM) for: The overall wheat samples, Irrigated wheat systems and Rainfed wheat systems ( Figure 7 ).

Key findings from path analysis: Overall wheat HI ( Figure 7a ): Soil nutrients showed the strongest positive effect. GCF exhibited significant negative effects. Plant characteristics had weak negative effects. Irrigated wheat ( Figure 7b ): All three factor categories negatively affected HI. Effect strength: Plant > GCF > Soil. Rainfed wheat ( Figure 7c ) (most stable model, GOF=0.633): All three factor categories positively influenced HI, Effect strength: Soil > GCF > Plant.

Based on the results of correlation analysis, Random Forest and Structural Equation Modeling(SEM), a linear fit of the three types of significant influences on the impact factors to wheat HI systematically quantifies how changes in these key drivers modulate wheat HI trends in arid regions. The similarities and differences between irrigated and rainfed wheat HI trends can also be visualized. It was found that Geographic-climatic factors had relatively less effect on wheat HI. Overall, wheat HI increased with increasing longitude (lon) and latitude (lat), and in irrigated wheat, HI increased with increasing latitude and longitude; whereas in rainfed wheat, HI had the opposite trend, decreasing with increasing latitude and longitude ( Figures 8d, e ). Meanwhile, wheat HI showed a decreasing trend with increasing altitude (alt) and soil accumulated temperature (SAT) ( Figures 8f, g ).

Agronomic management factors significantly influenced wheat HI, with plant biomass (AGB) and yield (yield) having the most direct effect on wheat HI, which significantly increased as they increased. In addition, AGB, yield and HI were significantly higher in irrigated wheat than in rainfed wheat ( Figures 8a, b ). Overall, there was a negative correlation between HI and water stress (H), with irrigated wheat showing a similar trend, whereas rainfed wheat showed an increase in HI when water stress increased instead ( Figure 8c ).

Soil nutrient factors are particularly important in influencing HI of wheat in arid regions, which is the basic security for wheat growth. The present study showed that the total nitrogen (TN) content of irrigated wheat was significantly higher than that of rainfed wheat, and HI increased with increasing TN, with rainfed wheat showing a more pronounced increase in this respect ( Figure 8h ). Overall, wheat HI was elevated with the increase in effective phosphorus (AP); in irrigated wheat, although HI showed a slight decrease with the increase in AP, the AP content was significantly higher in irrigated wheat than in rainfed wheat. The effect of AP on HI was more sensitive in the rainfed case, where HI was significantly elevated ( Figure 8i ). In addition, soil water content (SWC) was significantly higher in irrigated wheat than in rainfed wheat, and HI generally increased with increasing SWC, where HI was positively correlated with SWC in irrigated wheat, whereas HI decreased sharply with increasing SWC in rainfed wheat ( Figure 8j ). Finally, HI also increased with increasing soil density (BD), but rainfed wheat responded more sensitively to this ( Figure 8k ). It is worth noting that in arid regions, where the pH of wheat soils is alkaline, HI decreased overall with increasing pH ( Figure 8l ).

Geographic-climatic factors exhibit relatively limited but discernible spatial influence on wheat HI in arid regions. The study reveals that HI increases with longitude but decreases with accumulated effective soil temperature. In Xinjiang, a typical arid region, the distinct winter-cold and summer-hot climate renders water availability the primary limiting factor for wheat cultivation. Water supply patterns play a decisive role in determining HI (van den Boogaard et al., 1996). Compared to irrigated conditions, wheat grown under natural precipitation shows reduced yield, biomass, and heading days, along with significantly lower HI. Notably, drought-resistant cultivars typically demonstrate higher water-use efficiency than irrigated varieties (van den Boogaard et al., 1996), indicating differential adaptation strategies to water stress among wheat ecotypes. Globally, water deficit remains a key ecological constraint limiting crop productivity (Zhang et al., 2006). However, arid agricultural systems face dual challenges of water scarcity and inefficient utilization, severely constraining sustainable crop production (Ogola et al., 2002). Consequently, optimized water management remains crucial for improving wheat HI in arid regions.

The harvest index (HI) has been demonstrated to be a modifiable factor in crop production. The extent to which agricultural systems are influenced by climatic conditions varies depending on crop species, region, and management practices (Schlindwein et al., 2015). Our findings similarly indicate that agronomic management factors significantly affect wheat HI in arid regions, where wheat HI increases with higher grain yield and aboveground biomass (AGB), but decreases with plant height. Given the inherent upper limit of crop HI, recent introductions of dwarfing genes have reduced wheat plant height, altering dry matter partitioning while improving lodging resistance and yield potential, thereby facilitating HI enhancement (Brancourt-Hulmel et al., 2003; Zhang et al., 2012; Ukozehasi et al., 2022; Xi et al., 2024). Variations in crop HI are primarily attributable to differences in crop management. Research shows that wheat HI exhibits a negative correlation with the stem-to-AGB ratio but a positive correlation with the spike-to-AGB ratio. Foulkes et al. (2011) estimated that a 10% increase in AGB could theoretically raise HI to 0.64. Under drought stress conditions, the positive effect of HI on yield becomes more directly pronounced (Yang et al., 2000; Peltonen-Sainio et al., 2008; Zhang et al., 2012).

Significant differences in wheat harvest index (HI) were observed among different irrigation methods. Irrigated wheat demonstrated significantly higher grain yield, aboveground biomass (AGB), and HI compared to rainfed wheat (p < 0.01). These findings are consistent with a meta-analysis by Yang et al. (2022), which reported that irrigation can increase wheat yield by an average of 50%. Irrigation directly affects grain yield and HI (Ma et al., 2025) by regulating soil moisture conditions during grain filling (Uddling et al., 2008). In rainfed systems, natural precipitation serves as the sole water source, determining soil water availability (Noor et al., 2023). Under limited irrigation conditions, yield reduction caused by water constraints depends on both the severity and duration of soil water deficit. The impact of water deficit on crop yield varies according to specific phenological stages, with the most sensitive stages showing regional variations (Mishra et al., 1991). Since these variations are associated with regional differences in environmental conditions and agronomic practices, location-specific information is required to develop and optimize limited irrigation strategies. Optimal irrigation levels, nitrogen application rates, and cultivar selection can significantly improve winter wheat cultivation area, AGB, and yield, while nitrogen uptake increases with both irrigation and nitrogen inputs (Bingjun et al., 2018; Li et al., 2021). A key question regarding wheat (Triticum aestivum L.) HI in arid regions is whether HI manipulation in crop production can simultaneously achieve the dual objectives of increasing grain production and improving water conservation (Yang et al., 2000; Yang and Zhang, 2010).

Soil nutrient factors exert significant influences on wheat harvest index (HI) in arid regions, including soil nitrogen (N), phosphorus (P), water content, and pH. In long-term experiments, these factors induce changes in soil physical and chemical properties through differential inputs of mineral nutrients and organic matter, thereby directly affecting crop yield (Girma et al., 2007; Wang et al., 2017). Phosphorus represents one of the most critical limiting factors for plant growth (Albayrak et al., 2009). In arid soils, phosphorus deficiency inhibits wheat root development, consequently affecting growth rates (Zhao et al., 2024). Our results demonstrate significant positive correlations between soil total phosphorus (TP), available phosphorus (AP) and wheat HI in arid regions (p < 0.01). Soil available phosphorus not only serves as a key driver in soil carbon, nitrogen, phosphorus and sulfur cycles (Zhao et al., 2024), but also constitutes an essential nutritional source determining wheat harvest index.

Furthermore, both soil total nitrogen and available nitrogen demonstrated significant effects on wheat harvest index (HI) in arid regions (p < 0.05). Hay (1995) reported that nitrogen fertilizer had minimal impact on HI when applied at optimal rates. However, studies have indicated that in water-limited environments, excessive nitrogen application can lead to reduced yield and HI due to over-stimulated vegetative growth, which is associated with decreased post-anthesis carbon assimilation caused by soil water deficit (Herwaarden et al., 1998; Porker et al., 2020). Divergent perspectives exist regarding the effects of nitrogen application strategies on HI, likely attributable to the complex interactions between environmental factors and plant growth/yield formation. The prevailing view suggests that excessive nitrogen application may lead to decreased HI (Porker et al., 2020). Research has also revealed a positive correlation between wheat grain protein content and soil nitrogen availability, but an inverse relationship with actual grain yield (Prystupa et al., 2018). The ratio of soil available nitrogen to grain yield shows close association with grain protein content and has been proposed as a useful indicator for estimating the required amount of available nitrogen to achieve target yields (Prystupa et al., 2018). Given the generally positive correlation between soil available nitrogen and crop nitrogen uptake, it can be inferred that a similar relationship exists between grain nitrogen concentration and nitrogen uptake per unit grain yield in wheat (Gómez et al., 2021). Some studies have observed greater nitrogen uptake under irrigated conditions compared to rainfed systems (Zhang et al., 2017; Yan et al., 2022). Nevertheless, the underlying mechanisms through which soil nitrogen influences wheat HI in arid regions require further investigation.

A significant positive correlation was observed between soil water content and wheat harvest index (HI) in arid regions (p < 0.01), with HI progressively increasing as soil moisture levels rose. Given that wheat primarily grows under rainfed conditions in these areas, its growth and yield are highly dependent on both water availability and fertilizer inputs. Consequently, investigating the relationships among soil moisture, fertilizer levels, and crop growth is crucial for developing more effective dryland management practices (Halitligil et al., 2000; Huang et al., 2003). The synergistic interaction between soil water and fertilizers plays a determining role in wheat yield. Insufficient soil moisture may diminish the positive effects of nitrogen fertilization on wheat production, while heavy rainfall or excessive irrigation could lead to nitrogen leaching, thereby adversely affecting yield (Basso et al., 2012). Soil pH was found to be a limiting factor for wheat HI in arid regions (p < 0.05). Wheat typically thrives in neutral soils, whereas arid regions often have alkaline soils. This inherent alkalinity in arid-area soils consequently constrains potential improvements in wheat harvest index.