基于UAV与地面高光谱的杜兰小麦氮肥分施策略评价

1. Introduction

Wheat (Triticum durum Desf.) is one of the world’s most important cereal crops and plays a central role in global food security by supplying a substantial proportion of dietary calories and protein [1,2]. Achieving sustainable wheat production under increasing food demand and climate variability requires improved nitrogen (N) management. Nitrogen is the nutrient most closely associated with wheat growth, photosynthetic capacity, grain yield, and grain protein concentration [3]. However, excessive N fertilization reduces the effectiveness of fertilizer utilization and increases environmental risks through nitrate leaching, ammonia volatilization, and nitrous oxide emissions [4]. Consequently, modern wheat production aims to optimize nitrogen management by ensuring an adequate nitrogen supply for crop growth while minimizing environmental impacts [5,6]. Improving the synchronization between crop nitrogen demand and fertilizer application has therefore become a major objective of sustainable and precision agriculture [7].

Among the agronomic practices used to improve nitrogen management, split nitrogen application is widely recognized as an effective strategy because it synchronizes fertilizer application with crop nitrogen demand during successive phenological stages. Compared with a single basal application, split nitrogen application generally enhances nitrogen availability throughout the growing season, sustains canopy photosynthetic activity, improves grain yield and quality, and reduces nitrogen losses to the environment. Nevertheless, the effectiveness of split nitrogen applications varies with climatic conditions, soil properties, and crop developmental stages. These factors are particularly important in Mediterranean environments, where irregular rainfall and high evaporative demand strongly influence soil nitrogen dynamics. Therefore, identifying the optimum timing and frequency of split nitrogen applications remains a key challenge for improving wheat productivity while promoting more effective nitrogen management under region-specific conditions [7,8].

Accurate assessment of crop nitrogen status is essential for optimizing split nitrogen fertilization and supporting effective nitrogen management throughout the growing season. Conventional methods, including plant tissue analysis and destructive biomass sampling, provide reliable measurements but are labor-intensive, time-consuming, and unsuitable for repeated field-scale assessments. Consequently, remote sensing has emerged as an effective alternative for rapid, non-destructive, and spatially explicit monitoring of crop physiological responses, enabling timely evaluation of crop nitrogen status and supporting informed nitrogen management decisions [9].

Among available approaches, UAV-based multispectral imaging has been widely adopted because it provides high spatial resolution and flexible data acquisition under field conditions. Vegetation indices derived from multispectral imagery, such as NDVI and SR, have demonstrated strong relationships with canopy vigor, chlorophyll content, biomass accumulation, and nitrogen status. However, their performance is influenced by crop phenology, canopy structure, and environmental conditions, indicating that a single spectral indicator may not adequately characterize crop responses throughout the entire growing season [10,11].

Ground-based hyperspectral sensing complements UAV observations by providing detailed spectral information that is highly sensitive to changes in canopy biochemistry, particularly chlorophyll and nitrogen-related traits. These measurements are frequently used to calibrate and validate UAV-derived observations, improving the interpretation of crop physiological responses under different nitrogen management strategies [12].

In addition to spectral information, UAV photogrammetry enables the extraction of structural crop characteristics, including plant height and canopy architecture. Structural traits provide complementary information on crop growth and biomass that cannot be obtained from spectral data alone. Therefore, integrating spectral and structural information offers a more comprehensive assessment of crop development and nitrogen status than either approach independently [13].

Although the combined use of UAV multispectral imagery, ground-based hyperspectral sensing, and structural canopy traits has shown considerable potential, relatively few studies have applied these complementary techniques to evaluate different split nitrogen application strategies throughout successive wheat growth stages, particularly under Mediterranean environmental conditions. This limitation highlights the need for integrated sensing approaches capable of supporting more efficient, data-driven nitrogen management.

Despite substantial advances in nitrogen management and remote sensing technologies, several important knowledge gaps remain. Most previous studies have evaluated nitrogen fertilization rates or remote sensing techniques independently, with limited attention to the combined effects of different split nitrogen application schedules and multi-source sensing approaches. Moreover, many investigations have relied on a single sensing platform, providing either spectral or structural information, which may not fully capture the temporal physiological responses of wheat throughout the growing season. These limitations are particularly evident under Mediterranean environments, where variable rainfall and temperature strongly influence nitrogen availability and crop uptake.

To address these gaps, this study integrates UAV-based multispectral imagery, ground-based hyperspectral measurements, vegetation indices, and UAV-derived canopy structural traits within a unified experimental framework. Unlike previous studies, which generally focused on individual sensing techniques or nitrogen application rates, the proposed approach simultaneously evaluates biochemical and structural crop responses to six split nitrogen application strategies across successive Zadoks growth stages. This integrated framework provides a more comprehensive understanding of crop nitrogen dynamics and supports the development of more efficient and environmentally sustainable nitrogen management practices.

We hypothesized that: (i) synchronizing nitrogen applications with crop developmental stages would enhance crop growth and generate distinct spectral and structural responses in durum wheat; and (ii) integrating UAV multispectral imagery with ground-based hyperspectral sensing and canopy structural measurements would provide a more comprehensive and reliable assessment of crop responses to split nitrogen application strategies than individual sensing techniques.

Accordingly, the objectives of this study were to: (i) evaluate the physiological responses of durum wheat to six split nitrogen application strategies under Mediterranean conditions; (ii) assess the performance of integrated UAV multispectral and ground-based hyperspectral sensing for monitoring crop development across different growth stages; (iii) quantify the relationships among spectral, structural, and agronomic traits under different nitrogen management strategies; and (iv) identify the optimum split nitrogen application schedule for enhancing crop growth, maintaining plant nutritional status, and supporting sustainable wheat production.

2. Materials and Methods

2.1. Study Area and Experimental Design

The field experiment was conducted at the Agricultural Research Farm of the Faculty of Agriculture, Akdeniz University, located on the Akdeniz University campus in Antalya, Türkiye (36°53′54″ N, 30°38′13″ E), during the 2019–2020 growing season (November 2019 to May 2020) (Figure 1). The experimental field was located on a clay loam soil classified as a Typic Xerochrept, with a soil pH ranging from 7.94 to 8.02. The durum wheat (Triticum durum Desf.) cultivar Svevo was used in this study. Svevo is a widely cultivated, high-yielding Italian spring durum wheat cultivar developed by Società Produttori Sementi (SPS). It is characterized by its high grain quality, good adaptation to Mediterranean agroecological conditions, excellent pasta-making quality, and relatively high yield potential. Owing to its agronomic performance and broad adaptation, Svevo has been extensively used in breeding programs and scientific studies as a reference durum wheat genotype.

Figure 1. Study area located in Antalya, Turkey.

The experiment was established in a randomized complete block design (RCBD) comprising seven nitrogen treatments with three replications, resulting in a total of 21 experimental plots. Each plot measured 5.0 m × 2.5 m (12.5 m 2) (Figure 2).

Figure 2. Experimental plot layout and pre-establishment coordinate measurements: (a) randomized complete block design (RCBD) showing the treatments (N0–N6) and three replicates (1–3); (b) coordinate measurements conducted prior to experimental establishment.

Prior to the establishment of the experiment, the field was surveyed using a Differential Global Positioning System (DGPS). The acquired sub-centimeter-accuracy coordinate data were used to generate vector boundaries for each experimental plot, ensuring the precise positioning and layout of all treatments within the experimental field. This high-precision georeferencing also provided an accurate spatial framework for subsequent UAV image acquisition, photogrammetric processing, and geospatial analyses.

Before the first nitrogen application, soil samples were collected from the experimental field to determine its physical and chemical properties. Samples were obtained from two soil depths (0–20 and 20–40 cm), air-dried, gently crushed, and passed through a 2 mm sieve prior to laboratory analysis. Soil pH was determined in a 1:2.5 soil-to-water suspension using a digital pH meter following the method of Jackson [14], while electrical conductivity (EC) was measured in the same suspension according to Page et al. [15]. Total carbonate content was determined using a Scheibler calcimeter [16]. Soil texture, organic matter, and total nitrogen contents were determined using the hydrometer method [17], the modified Walkley–Black method [18], and the modified Kjeldahl method [19], respectively. Plant-available phosphorus was extracted with 0.5 M NaHCO 3 and quantified using the molybdate blue colorimetric method with a UV–Vis spectrophotometer Shimadzu UV-1800 (Shimadzu Corporation, Tokyo, Japan) [20]. Exchangeable potassium (K), calcium (Ca), and magnesium (Mg) were extracted with 1 M ammonium acetate (NH 4 OAc) and determined by inductively coupled plasma optical emission spectrometry (ICP–OES) [20]. Available iron (Fe), zinc (Zn), manganese (Mn), and copper (Cu) were extracted using diethylenetriaminepentaacetic acid (DTPA) and quantified by ICP–OES.

The physical and chemical properties of the soil samples collected from the 0–20 and 20–40 cm depths are summarized in Table 1. The analytical results indicated that the experimental soil had a clay loam texture with generally adequate nutrient availability for durum wheat production. (Table 1).

Table 1. Physical and chemical properties of the study area soils.

Based on soil analysis results (Table 1) and a target grain yield of 7500 kg ha−1, 140 kg ha−1 of phosphorus and potassium were applied as base fertilization during sowing. Phosphorus was applied as diammonium phosphate (DAP) and potassium as potassium sulfate (K 2 SO 4), taking into account commonly applied fertilizers. For nitrogen fertilization, after deducting the amount of nitrogen from DAP fertilizer, a total of 175 kg N ha−1 of urea fertilizer was applied as top dressing. In the control application, no nitrogen fertilization was performed other than base fertilization. To evaluate the effects of split nitrogen application, the total nitrogen rate was divided into different application schedules according to the principal phenological stages of wheat defined by the Zadoks growth scale [21]. Split nitrogen applications were carried out at sowing, tillering, stem elongation, booting, heading, and flowering, depending on the treatment.

The experimental treatments consisted of seven split nitrogen application strategies: N0 (no nitrogen application), N1 (100% of the total nitrogen applied at sowing), N2 (50% at sowing + 50% at tillering), N3 (33% at sowing + 33% at tillering + 33% at stem elongation), N4 (25% at sowing + 25% at tillering + 25% at stem elongation + 25% at booting), N5 (20% at sowing + 20% at tillering + 20% at stem elongation + 20% at booting + 20% at heading), and N6 (16.7% at sowing + 16.7% at tillering + 16.7% at stem elongation + 16.7% at booting + 16.7% at heading + 16.7% at flowering (Figure 3).

Figure 3. Experimental scheme of split nitrogen application treatments across durum wheat phenological stages.

In the agricultural ecological conditions of Antalya province, durum wheat (Triticum durum Desf.) is typically sown in mid-November. However, due to prevailing weather conditions, wheat sowing was carried out on 16 December 2019. Harvesting took place on 15 June 2020.

2.2. Leaf Sampling and Remote Sensing Studies

Plant sampling was initiated at the tillering stage and continued at five subsequent phenological stages throughout the growing season. During each sampling event, representative leaf samples were collected from every experimental plot for laboratory analyses following the procedures described by Kacar and Inal [22]. For each plot, the youngest fully expanded leaves were collected from 20 randomly selected plants. Rather than sampling all leaves on the shoot, only the youngest fully expanded leaves were used, and the collected leaves were pooled to obtain a representative composite sample for each experimental plot (Figure S1). Total nitrogen (N) concentration was determined using the modified Kjeldahl method, while phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), iron (Fe), zinc (Zn), manganese (Mn), and copper (Cu) concentrations were measured in the filtrate obtained by wet digestion using an inductively coupled plasma optical emission spectrometer (ICP-OES; PerkinElmer Avio 2000, PerkinElmer, Shelton, CT, USA).

Ground-based hyperspectral measurements and UAV-based multispectral image acquisition were conducted concurrently with each plant sampling period to ensure consistency between field observations and remote sensing data (Figure 4 and Figure S2). Hyperspectral reflectance measurements were acquired using an ASD FieldSpec® 3 spectroradiometer (Analytical Spectral Devices Inc., Boulder, CO, USA), covering the 325–1075 nm spectral range. Measurements were collected under clear-sky conditions by positioning the sensor at a height of 1.35 m above the crop canopy, producing a ground field of view (GFOV) with a diameter of approximately 60 cm (Figure 3). The GFOV was calculated according to the method described by Mac Arthur et al. [23], as expressed in Equation (1).

$$
G F O V = 2 \textrm{ } \left(h \times tan \frac{\theta}{2}\right)
$$

(1)

where GFOV is the ground field of view, h is the height above the target, and θ is the field of view (FOV).

Figure 4. Hyperspectral measurement setup. (a) Field spectroradiometer; (b) ground field of view (FOV) during spectral measurements.

UAV-based multispectral image acquisition was performed at flight altitudes of 10 and 45 m during each sampling period. Multispectral imagery was collected using a MAPIR Survey3W (RGN) camera mounted on the UAV platform. The camera captures reflectance in the green, red, and near-infrared (NIR) spectral bands and is equipped with a 12-megapixel sensor weighing 76 g. The MAPIR Survey3W RGN (Red, Green, and Near-Infrared) camera is equipped with three spectral bands centered at 550 nm (green), 660 nm (red), and 850 nm (near-infrared). To ensure the accuracy and temporal consistency of reflectance measurements, radiometric calibration was performed for all multispectral images using a MAPIR Reflectance Calibration Panel before image processing. The calibration procedure converted raw digital numbers (DNs) into surface reflectance values based on reference reflectance targets, thereby minimizing the effects of variations in illumination conditions among different flight dates and enabling reliable temporal comparisons of multispectral imagery [24].

Following radiometric calibration, the images were processed to generate georeferenced multispectral datasets. Orthorectification, resampling, and band stacking were performed to produce four-band multispectral image products for subsequent vegetation index calculations and spatial analyses.

Ground-based hyperspectral reflectance measurements were acquired at each phenological stage for every nitrogen treatment and replication, resulting in a total of 751 spectral observations. Spectral measurements were collected using an ASD FieldSpec handheld spectroradiometer (Analytical Spectral Devices Inc., Boulder, CO, USA), operating over the 325–1075 nm wavelength range. All measurements were conducted under clear-sky conditions between 11:00 and 13:00 local time to minimize the effects of variations in solar illumination. The instrument was configured with a 25° field of view, and the distance between the sensor and the crop canopy was maintained at 1.35 m throughout the measurement campaign.

To facilitate spectral interpretation and reduce data complexity, the measured reflect