Monitoring and Optimization of Potato Growth Dynamics under Different Nitrogen Forms and Rates Using UAV RGB Imagery

Abstract

The temporal dynamics of canopy growth are closely related to the accumulation and distribution of plant dry matter. Recently, unmanned aerial vehicles (UAVs) equipped with various sensors have been increasingly adopted in crop growth monitoring. In this study, two potato varieties were used as materials, and treated with different combinations of nitrogen forms (nitrate and ammonium) and application rates (0, 150, and 300 kg ha⁻¹). A canopy development model was then constructed using low-cost time-series RGB imagery acquired by UAV. The objectives of this study were to quantify the variation in canopy development parameters under different nitrogen treatments and to explore the model parameters that represent the dynamics of plant dry matter accumulation, as well as those that contribute significantly to yield. The results showed that, except for the thermal time to canopy senescence (t₂), other parameters of the potato canopy development model exhibited varying degrees of variation under different nitrogen treatments. The model parameters were more sensitive to nitrogen forms, such as ammonium and nitrate, than to application rates. The integral area (A_t) under the canopy development curve had a direct effect on plant dry matter accumulation (path coefficient of 0.78), and the two were significantly positively correlated (Pearson correlation coefficient of 0.93). Integral area at peak flowering (A_tII) was significantly correlated with yield for both single and mixed potato varieties, having the greatest effect on yield (total effect of 1.717). In conclusion, UAV-acquired time-series RGB imagery could effectively quantify the variation of potato canopy development parameters under different nitrogen treatments and monitor the dynamic changes in plant dry matter accumulation. The regulation of canopy development parameters is of great importance and practical value for optimizing nitrogen management strategies and improving yield.

1. Introduction

Potato (Solanum tuberosum L.) is one of the most important non-cereal crops, with high nutritional and economic value, and plays a crucial role in ensuring global food security [1]. There is a competitive and dependent relationship between potato canopy growth and tuber yield formation [2]. Proper nitrogen fertilizer management not only promotes the timely transfer of photosynthetic assimilation products and increases potato tuber yield, but also reduces negative impact on the environment [3,4]. Therefore, it is important to conduct a comprehensive study on the relationship between potato canopy growth dynamics and plant dry matter accumulation and distribution under different nitrogen fertilizer management practices, which is of great importance for optimizing nitrogen management and improving tuber yield.

Nitrogen is an essential nutrient for plant growth and development. Appropriate nitrogen fertilizer dosage can enhance the growth of potato plants and contribute to increased dry matter accumulation and tuber yield [5]. However, more nitrogen is not always better; excessive nitrogen application not only fails to increase tuber yield, but also reduces tuber quality and causes negative environmental impacts [6]. The main sources of nitrogen taken up by potatoes are nitrate and ammonium nitrogen, but research on the effects of different forms of nitrogen on potatoes has yielded conflicting results. While Lorenz et al. [7] found that ammonium nitrogen was the most effective nitrogen source for potato growth, other studies have shown that nitrate nitrogen promotes greater stem and leaf growth compared to ammonium nitrogen and urea, resulting in higher dry matter accumulation and tuber yield [8].

The canopy is the primary site of photosynthesis in crops, and its growth dynamics are closely linked to plant dry matter accumulation and distribution [2,9]. Environmental factors such as temperature, photoperiod, light, water, and nitrogen availability influence canopy growth dynamics in the field, with temperature and nitrogen having the most significant effects [10]. In addition, nitrogen supply is a critical factor influencing potato dry matter accumulation and distribution [11,12,13,14,15]. Studies have shown that adequate nitrogen supply can enhance chlorophyll synthesis, boost photosynthetic capacity, and produce more organic matter, thus facilitating canopy expansion and dry matter accumulation [16]. Variations in canopy growth, duration of maximum green canopy cover, and senescence rate influence the dry matter accumulation rate by affecting light interception and radiation use efficiency [17].

The degree of canopy development can be quantified by the percentage of soil surface area covered by green leaves and has been recognized as one of the effective indicators for assessing the factors affecting crop growth characteristics [16,17,18]. The rate of canopy growth, the duration of canopy maintenance, and the rate of canopy senescence together determine the crop canopy cover throughout the growth cycle [19]. It has been shown that nitrogen promotes crops to reach maximum canopy cover earlier, prolongs canopy cover time, and slows canopy senescence, thereby increasing the canopy’s ability to intercept photosynthetically active radiation [10,20]. Khan [18] was the first to construct a physiological-ecological model of potato canopy cover dynamics with heat accumulation throughout the growth cycle, and quantifiable model parameters were analyzed as new traits. Some studies have shown that model parameters can be affected by multiple factors such as genetic variation, nitrogen fertilizer dosage, environmental conditions, and their interactions [10,16,17,18]. However, there is still a relative lack of studies on the temporal dynamics of canopy growth, plant dry matter accumulation and partitioning, and their interrelationships in potato under different combinations of nitrogen forms and rates throughout the life span.

Traditional crop monitoring methods are often labor-intensive and time-consuming and have difficulties comprehensively and accurately reflecting actual conditions in large agricultural fields [21]. In recent years, high-throughput phenotyping technologies, exemplified by unmanned aerial vehicle (UAV) remote sensing, have been widely applied for non-destructive, efficient, accurate, and continuous crop monitoring [22,23]. UAVs equipped with visible light sensors, such as consumer-grade digital cameras, offer the advantages of low cost, ease of operation, and rich data acquisition, making them widely used for crop growth monitoring, disease detection, and yield estimation [24,25]. In this study, we used a UAV as a remote sensing platform to periodically and continuously acquire high-resolution RGB images of potato throughout the growing season, and subsequently we fitted potato canopy development curves by extracting canopy cover. The objectives of this study were to (1) quantify the variations in canopy development parameters of potato under different combinations of nitrogen application and nitrogen form, (2) explore the canopy development parameter that represents the dynamics of total dry matter accumulation in potato plants, and (3) identify stable canopy development parameters that significantly contribute to potato yield and are not influenced by potato variety.

2. Materials and Methods

2.1. Experimental Sites and Design

The field experiment was conducted in the Chabei Administrative District of Zhangjiakou City, Hebei Province, located at 115.06° E longitude, 41.48° N latitude, and an altitude of 1390 m. The average temperature during the potato growing season in this region is about 16.49 °C, with average solar radiation of 3.64 × 10⁴ W m⁻², typical of a temperate continental monsoon climate. The experimental field soil is a calcareous chernozem with 16.27 g kg⁻¹ organic matter, 1.08 g kg⁻¹ total nitrogen, 0.07 g kg⁻¹ hydrolyzable nitrogen, 8.81 mg kg⁻¹ available phosphorus, 73.00 mg kg⁻¹ available potassium, and a pH of 8.4.

The experimental design of “2 potato varieties × 2 nitrogen forms × 3 nitrogen application rates” was adopted, using a split-plot design with four replications, resulting in a total of 40 experimental plots. The main plots were assigned to the nitrogen treatments (nitrogen forms and application rates), while the sub-plots were assigned to the two potato varieties. Each experimental plot was planted with 6 rows, each 10 m long and 5.4 m wide, with 20 cm plant spacing and 90 cm row spacing; the field layout is shown in Figure 1. Two potato varieties, ‘Shapody’ (V1) and ‘Zhongshu 18’ (V2), were selected for the experiment. The forms of nitrogen used were nitrate nitrogen (potassium nitrate, KNO₃) and ammonium nitrogen (ammonium sulfate, (NH₄)₂SO₄, with 1% nitrification inhibitor). Nitrogen application rates were set at three levels: 0, 150, and 300 kg ha⁻¹ of nitrogen. Nitrogen treatments were randomly assigned to the main plots, while the two potato varieties were randomly distributed within each main plot as sub-plots.

The nitrogen was applied as follows: 40% of the total nitrogen fertilizer was applied as a basal dose at planting, while the remaining nitrogen was administered through fertigation in split applications, with all nitrogen applied before the peak flowering stage. Specifically, 20% was applied before flowering, 20% during the early flowering stage, 10% just before peak flowering, and the final 10% during full flowering. Phosphorus was applied at 240 kg ha⁻¹ as superphosphate (Ca(H₂PO₄)₂), and potassium was applied at 450 kg ha⁻¹ as potassium sulfate (K₂SO₄). Irrigation was provided via drip irrigation to meet crop water requirements throughout the growing season. Pesticides and fungicides were applied as necessary based on local agronomic recommendations, and weed control was performed manually as needed. The nitrogen treatments were coded as N0, N1, N2, N3, and N4, representing the following: control (no nitrogen fertilizer), 150 kg ha⁻¹ KNO₃, 300 kg ha⁻¹ KNO₃, 150 kg ha⁻¹ (NH₄)₂SO₄, and 300 kg ha⁻¹ (NH₄)₂SO₄, respectively. All other field management practices were consistently applied across all treatments and replications to ensure uniform experimental conditions.

2.2. UAV Image Acquisition and Preprocessing

A small quadcopter drone, the Inspire 2 (DJI Technology Co., Ltd., Shenzhen, China), was used as the remote sensing platform. It was equipped with a 20.8 megapixel Zenmuse X5S image sensor (red, green, and blue channels) to acquire digital images of the potato throughout its growth cycle, with UAV surveys conducted approximately once a week at different growth stages. The flight altitude was set to 20 m (corresponding to a spatial resolution of 0.44 cm/pixel) to balance sufficient coverage and resolution for accurate analysis, and the front and side overlaps were 80% and 70%, respectively. To avoid the influence of light on the results, clear and cloudless weather was chosen for the flight operation, and the flight time was at 11:00~13:00. Before the mission, five ground control points were evenly placed in the field area for post-image geometric correction. Image preprocessing was performed using Agisoft Metashape Professional software (Agisoft LLC, St. Petersburg, Russia), which consists of the steps of importing raw images, aligning, geometric correction, generating point clouds, constructing grids, and finally generating a digital orthomosaic (DOM).

2.3. Canopy Cover Extraction and Canopy Development Curve Fitting

The vegetation canopy information in the DOM generated from UAV RGB images was accurately extracted using the green leaf index (GLI), a vegetation index based on the RGB bands (Equation (1)). A threshold of 0.05 was applied to differentiate vegetation from non-vegetation areas in this study. The DOMs were processed to classify each pixel as either vegetation or non-vegetation based on the GLI threshold. The ratio of the number of canopy pixels to the total number of pixels in the central area (4 m × 5 m) of each experimental plot was then used to determine the canopy coverage of each plot.

$$GLI={\displaystyle \frac{2G-R-B}{2G+R+B}}$$

(1)

The dynamics of potato canopy coverage over thermal days was modeled according to the method described by Khan [18], which was used to evaluate the dynamics of canopy development under different nitrogen treatments (Figure 2). The formula for converting days after emergence to thermal days is as follows:

$$g\left(T\right)=\left({\displaystyle \frac{T_c-T}{T_c-T_o}}\right){\left({\displaystyle \frac{T-T_b}{T_o-T_b}}\right)}^{{\displaystyle \frac{T_{\mathrm{o}}-T_{\mathrm{b}}}{T_{\mathrm{c}}-T_{\mathrm{o}}}}},$$

(2)

where $g(T)$ represents the growth or development rate of the plant as a function of temperature, with $T_b$, $T_o$, and $T_c$ being the base, optimum, and ceiling temperatures, respectively [26]. In this study, these values were set to 5.5 °C, 23.4 °C, and 34.6 °C [17].

The dynamics of potato canopy cover can be divided into three stages: the canopy expansion phase, the maximum canopy coverage duration, and the canopy decline phase [17,18]. These stages can be described mathematically as follows:

$$v={\nu }_{\mathrm{m}\mathrm{a}\mathrm{x}}(1+{\displaystyle \frac{t_1-t}{t_1-t_{\mathrm{m}1}}}){({\displaystyle \frac{t}{t_1}})}^{{\displaystyle \frac{t_1}{t_1-t_{\mathrm{m}1}}}},\mathrm{with}0\le t\le t_1,$$

(3)

$$\nu ={\nu }_{\mathrm{m}\mathrm{a}\mathrm{x}},\mathrm{with}{t}_1\le t\le t_2,$$

(4)

$$\nu ={\nu }_{max}\left({\displaystyle \frac{t_e-t}{t_e-t_2}}\right){\left({\displaystyle \frac{t+t_1-t_2}{t_1}}\right)}^{{\displaystyle \frac{t_1}{t_e-t_2}}},\mathrm{with}{t}_2\le t\le t_{\mathrm{e}},$$

(5)

In the formula above, $\nu$ represents the canopy cover as a decimal between 0 and 1, where $\nu =0$ means no cover and $\nu =1$ means full cover. $t$ represents thermal time. The model includes five key parameters: $t_{\mathrm{m}1}$, $t_1$, $t_2$, $t_{\mathrm{e}}$, and ${\nu }_{\mathrm{m}\mathrm{a}\mathrm{x}}$. $t_{\mathrm{m}1}$, $t_1$, $t_2$, and $t_{\mathrm{e}}$ represent the thermal times corresponding to key points in the canopy development (the inflection point described in Equation (3), and the onset, maximum, and end of canopy cover, respectively). ${\nu }_{\mathrm{m}\mathrm{a}\mathrm{x}}$ represents the maximum canopy cover.

The main secondary parameters derived from the model were: the thermal days of maximum canopy cover duration (t₂−t₁); the thermal days of canopy senescence duration (t_e−t₂); the maximum canopy growth rate (C_m1, Equation (6)); the integral area of the three phases under the canopy development curve (A₁, A₂, and A₃); and the total integral area (A_sum). When the thermal days reach t_e, the integral area (A_t) under the dynamic development curve of canopy cover versus thermal days equals A_sum. Furthermore, A_tI, A_tII, A_tIII, and A_tIV represent the cumulative areas under the aforementioned curve for potato plants during the budding stage, full flowering stage, end of flowering stage, and leaf senescence stage, respectively. These cumulative areas correspond to the plant sampling time points described in Section 2.5.

$$C_{\mathrm{m}1}={\displaystyle \frac{2t_1-t_{\mathrm{m}1}}{t_1(t_1-t_{\mathrm{m}1})}}({\displaystyle \frac{t_{\mathrm{m}1}}{t_1}}{)}^{{\displaystyle \frac{t_{\mathrm{m}1}}{t_1-t_{\mathrm{m}1}}}}{v}_{\mathrm{m}\mathrm{a}\mathrm{x}}$$

(6)

2.4. Determination of Net Photosynthetic Rate at the Leaf Level

We randomly selected and marked three healthy potato plants with representative growth in each experimental plot. Photosynthetic gas exchange parameters of the newest functional leaves on the marked plants were measured using a CIRAS-3 photosynthesis system (PP Systems, Amesbury, MA, USA) on sunny days during key growth stages from 8:30 A.M. to 11:30 A.M. The parameters were set as follows: light intensity of 1200 µmol m⁻² s⁻¹, leaf chamber temperature of 25 °C, and CO₂ concentration of 400 µmol mol⁻¹.

2.5. Determination of Total Plant Dry Weight and Tuber Yield

Plant samples were collected at four key growth stages, the budding stage, full flowering stage, end of flowering stage, and leaf senescence stage, which correspond to the BBCH (Biologische Bundesanstalt, Bundessortenamt and CHemical industry) scale stages 50, 65, 69, and 91, respectively. The BBCH scale is an internationally standardized system for describing the phenological development stages of plants. Eight plants per plot were sampled at the bud and full flowering stages, while four plants were sampled at the late flowering and senescence stages due to the larger size of the plants during these stages. To ensure representativeness, plants were selected from different parts of the plot. The sampled plants were immediately placed in an oven at 105 °C for 30 min for rapid drying, followed by drying at a constant 70 °C until completely dried. After drying, their dry weight was measured. Finally, the total dry weight per plant for each sampling period was calculated based on the number of destructively sampled plants. At harvest, the fresh tuber weight of each experimental plot was measured. To calculate tuber yield per plant, the total fresh tuber weight was divided by the number of plants remaining in each plot after accounting for the number of plants destructively sampled at planting and at specific intervals throughout the growing season. The yield was then standardized per plant by subtracting the sampled plants from the total planted population in each plot.

2.6. Data Analysis

The canopy development curves were fitted using Python software (version 3.10.9), and the resulting data were subjected to one-way analysis of variance and significance tests in IBM SPSS Statistics (version 27.0.1). Linear regression analysis and graphing were performed with Microsoft Excel (version 2016). R software (version 4.2.3) was used for path analysis, correlation analysis (Pearson correlation coefficient), and additional plotting. In the path model, the paths between variables represent causal relationships, with the direction of the arrows indicating the order of causality. The path coefficients represent the magnitude of influence one variable has on another. We calculated both direct and indirect effects using standardized path coefficients to further evaluate the relationships between the variables. The significance of each path was tested, and paths with a p-value less than 0.01 were considered statistically significant.

3. Results

3.1. Parameters of Potato Canopy Development under Different Nitrogen Treatments

Figure 3 shows the fitted curves of potato canopy development under different nitrogen treatments, and it can be seen that there are obvious differences in the canopy development curves of ‘Shapody’ and ‘Zhongshu 18’ under different nitrogen treatments, especially under different nitrogen forms.

3.1.1. Primary Parameters

The analysis of the main parameters for the fitted canopy development curves of potato populations across all field experiment plots are presented in

Table 1. Results of the fitting analysis of the primary parameters of the potato canopy development model.

Varieties	Treatments	Vmax	tm1	t1	t2	te
‘Shapody’	N0	0.74 ± 0.06 de	23.76 ± 0.92 bc	50.00 ± 0.00 a	58.93 ± 1.21 abc	70.56 ± 0.95 e
	N1	0.86 ± 0.04 abc	25.05 ± 1.09 b	50.00 ± 0.00 a	59.66 ± 1.89 abc	69.30 ± 1.92 e
	N2	0.91 ± 0.04 ab	25.91 ± 0.81 b	47.12 ± 1.82 ab	56.06 ± 0.84 c	70.89 ± 0.31 e
	N3	0.98 ± 0.00 a	25.65 ± 0.73 b	37.12 ± 1.14 de	57.90 ± 1.84 bc	76.26 ± 0.90 d
	N4	0.97 ± 0.01 a	28.35 ± 0.21 a	41.13 ± 0.48 cd	59.80 ± 1.15 abc	76.61 ± 1.60 d
‘Zhongshu 18’	N0	0.67 ± 0.06 e	15.59 ± 1.21 f	45.25 ± 2.97 abc	62.34 ± 1.11 ab	84.27 ± 1.35 c
	N1	0.75 ± 0.06 cde	18.82 ± 1.05 e	44.00 ± 4.06 bc	63.08 ± 1.45 a	86.86 ± 0.79 bc
	N2	0.84 ± 0.06 bcd	20.09 ± 0.82 de	48.08 ± 1.31 ab	61.74 ± 1.92 ab	89.53 ± 4.27 b
	N3	0.97 ± 0.01 a	22.33 ± 0.34 cd	35.53 ± 2.06 e	61.59 ± 0.50 ab	88.59 ± 2.44 bc
	N4	0.95 ± 0.00 ab	24.13 ± 0.89 bc	36.47 ± 1.87 de	62.91 ± 1.24 a	96.81 ± 1.94 a

Bolded data are maximum values under all treatments of the variety and italicized data are minimum values. Same lowercase letters indicate no significant difference at the p < 0.05 level. V_max: maximum canopy cover; t_m1: inflection point during canopy expansion; t₁: duration of canopy expansion; t₂: onset of canopy senescence; t_e: the time point when canopy cover is zero.

. It can be seen that different forms of nitrogen and application rates have significant effects on most parameters (p < 0.05). Both potato varieties, ‘Shapody’ and ‘Zhongshu 18’, showed the highest mean maximum canopy cover (V_max) under the 150 kg ha⁻¹ ammonium nitrogen (N3) treatment, with increases of 31.08% and 44.78%, respectively, compared to the control (N0).

Nitrogen fertilization delayed the thermal time (tm1) to reach the maximum canopy growth rate (C_m1) to varying degrees and prolonged the above-ground life cycle (t_e), with the most significant effects observed under the 300 kg ha⁻¹ ammonium nitrogen (N4) treatment. Overall, the t_e value of the mid-to-late maturing variety ‘Shapody’ was lower than that of the late-maturing variety ‘Zhongshu 18’. Canopy decay thermal time (t₂) did not differ significantly among nitrogen treatments, indicating that t₂ is a relatively conservative parameter for a given cultivar. The thermal days (t₁) required to reach V_max for both varieties were significantly shortened (p < 0.05) under ammonium nitrogen treatment. The t₁ value for the N3 treatment was the smallest, reduced by 12.88 td and 9.72 td, respectively, compared to the N0 treatment. This indicates that ammonium nitrogen can significantly promote potato populations to reach maximum canopy coverage earlier and delay the senescence of the above-ground parts of the plants.

3.1.2. Secondary Parameters

Compared to nitrogen application rates, different forms of nitrogen have a more significant effect on secondary parameters (

Table 2. Results of the fitting analysis of the secondary parameters of the potato canopy development model.

Varieties	Treatments	Cm1	t2−t1	te−t2	A1	A2	A3	Asum
‘Shapody’	N0	0.02 ± 0.00 cd	8.93 ± 1.21 e	11.63 ± 1.80 ef	19.04 ± 1.62 abc	6.75 ± 1.23 e	5.90 ± 0.69 f	31.68 ± 2.31 e
	N1	0.03 ± 0.00 cd	9.66 ± 1.89 e	9.64 ± 3.81 f	21.48 ± 0.95 a	8.16 ± 1.48 e	6.03 ± 2.59 ef	35.67 ± 1.98 de
	N2	0.03 ± 0.00 c	8.95 ± 1.14 e	14.82 ± 0.84 def	20.14 ± 0.44 ab	8.25 ± 1.33 e	9.42 ± 0.78 def	37.82 ± 1.76 cd
	N3	0.05 ± 0.00 a	20.78 ± 2.48 abc	18.36 ± 2.02 cde	13.83 ± 0.64 c	20.27 ± 2.39 ab	12.33 ± 1.36 cd	46.44 ± 1.27 b
	N4	0.04 ± 0.00 b	18.67 ± 1.36 cd	16.82 ± 2.10 cdef	15.21 ± 0.54 bc	18.01 ± 1.20 bc	11.26 ± 1.45 cd	44.47 ± 1.30 b
‘Zhongshu 18’	N0	0.02 ± 0.00 d	17.10 ± 2.93 cd	21.93 ± 2.24 bcd	17.46 ± 2.89 abc	11.00 ± 1.35 de	10.00 ± 1.23 de	38.46 ± 3.36 cd
	N1	0.03 ± 0.00 cd	19.09 ± 4.67 bcd	23.77 ± 1.00 bc	17.72 ± 3.25 abc	13.78 ± 2.76 cd	12.18 ± 1.32 cd	43.67 ± 3.51 bc
	N2	0.03 ± 0.00 cd	13.67 ± 2.50 de	27.79 ± 5.42 ab	21.65 ± 2.04 a	11.51 ± 2.25 de	15.14 ± 1.82 bc	48.29 ± 1.70 b
	N3	0.05 ± 0.00 ab	26.06 ± 1.87 ab	27.00 ± 2.91 ab	14.58 ± 1.74 bc	25.18 ± 1.68 a	17.60 ± 1.74 ab	57.36 ± 1.45 a
	N4	0.05 ± 0.00 ab	26.44 ± 2.82 a	33.90 ± 1.07 a	13.97 ± 0.99 c	25.13 ± 2.73 a	21.54 ± 0.65 a	60.64 ± 2.19 a

Bolded data are maximum values under all treatments of the variety and italicized data are minimum values. Same lowercase letters indicate no significant difference at the p < 0.05 level. C_m1: maximum growth rate during canopy expansion; t₂−t₁: duration of maximum canopy cover; t_e−t₂: duration of canopy senescence; A₁, A_2, A₃: areas under individual curve segments; A_sum: integrated ground cover.

). C_m1, t₂−t₁, t_e−t₂, A₂, A₃, and A_sum increased significantly under ammonium nitrogen treatment (p < 0.05), except for A₁ which was greatly reduced. The A_sum values of ‘Shapody’ and ‘Zhongshu 18’ reached their maximum under N3 and N4 treatments, respectively, increasing by 46.59% and 57.67% compared to the control group. This may be related to the duration of the fertility periods of these varieties.

3.2. Relationship between Integral Area under Canopy Development Curve and Plant Dry Matter Accumulation

The relationship between total dry weight (TDW) of plants and area of integration (A_t) under the dynamic development curve of canopy cover with thermal time is shown in Figure 4a. The results indicate that A_t had a strong linear relationship with TDW, with a high coefficient of determination (R²) of 0.87. Compared to the single indicator At, the correlation between the composite indicator A_t × P_n, constructed using the net photosynthetic rate (P_n) at the leaf level, and TDW showed only a slight improvement (R² = 0.88) (Figure 4b).

Correlation analysis revealed that P_n, A_t, and TDW all have a significant positive linear correlation (p < 0.01). To further clarify the direct and indirect effects of P_n and A_t on TDW, a path analysis was performed (Figure 5). The results showed that P_n had a significant positive effect on A_t (path coefficient = 0.51) and A_t on TDW (path coefficient = 0.78), while P_n had no significant direct effect on TDW (p < 0.01). This indicates that P_n mainly has an indirect effect on TDW by influencing A_t and then TDW.

3.3. Relationship between Tuber Yield and Model Parameters

The correlation matrix of tuber yield with model parameters for all experimental plots is presented in Figure 6. It can be seen that t_m1 and t₁ had significant correlations with tuber yield, V_max, C_m1, and A₁ had no significant correlation with tuber yield, while all other model parameters had a significant positive correlation with tuber yield (p < 0.01). Among these, A_sum, t_e, and the integral area of canopy coverage at full flowering stage with thermal time (A_tII) had the strongest linear correlation with tuber yield, with Pearson correlation coefficients of 0.81, 0.79, and 0.78, respectively.

When analyzed separately by variety, the correlation between tuber yield and model parameters varied significantly (Figure 7). The results showed that V_max, A_sum, A_tII, A_tIII, and A_tIV were significantly correlated with tuber yield for individual varieties (p < 0.01). Among these, the correlation coefficients of A_tII, A_tIII, and A_tIV with tuber yield were all above 0.80. t₁ and A₁ were found to be significantly negatively correlated with tuber yield in variety ‘Shapody’, while t_e, C_m1, t₂−t₁, t_e−t₂, A₂, and A₃ were significantly positively correlated with tuber yield in ‘Shapody’. However, these model parameters did not show significant correlations with tuber yield in ‘Zhongshu 18’. These indicate that A_sum, A_tII, A_tIII, and A_tIV are effective parameters for predicting potato yield performance for both single and mixed varieties.

To further analyze the extent to which the model parameters affected yield, we quantified the contribution of V_max, A_tII, A_tIII, A_tIV, and A_sum to yield using path analysis (

Table 3. Analysis of the total effect of model parameters on yield.

Parameters	Direct Effect	Indirect Effect	Total Effect
Vmax	−0.002	−0.706	−0.708
AtII	0.573	1.144	1.717
AtIII	0.131	−2.206	−2.075
AtIV	−0.768	1.885	1.117
Asum	0.732	0.000	0.732

V_max: maximum canopy cover; A_tII, A_tIII, and A_tIV: area under the canopy curve for the peak flowering, final flowering, stem and leaf senescence stages, respectively. A_sum: integrated ground cover.

). It can be seen that A_sum has the largest positive direct effect on yield (standardized coefficient of 0.73). V_max, A_tIII, and A_tIV have negative direct and/or indirect effects on yield, while A_tII has significant positive direct and indirect effects on yield, with the largest total effect (1.717). The above indicates that adjusting the variables V_max, A_tIII, and A_tIV carries a higher risk and may have a negative impact on yield; increasing the integral area under the canopy development curve at peak flowering (A_tII) is the most effective way to improve yield.

4. Discussion

4.1. Factors Affecting Canopy Development Parameters

Factors such as crop genotype, nitrogen supply, temperature, and water conditions significantly influence canopy development parameters [10,16]. Previous studies have shown that increasing nitrogen inputs can increase maximum canopy cover (V_max) [6,24,25,26,27,28]. Our results further showed that the effect of different nitrogen forms on V_max was more significant compared to the dose of nitrogen applied, where V_max was higher under the 150 kg ha⁻¹ ammonium nitrogen treatment than other treatments. Ospina et al. [27] found shorter t₁ and longer t₂ at high nitrogen fertilization levels, which is not completely consistent with the results of this study and may be due to different physiological response mechanisms of different genotypes to different forms of nitrogen.

Our study showed that high nitrogen levels, especially high doses of ammonium nitrogen, significantly prolonged t_e. The late-maturing variety ‘Zhongshu 18’ had significantly higher t_e and A_sum than the mid-to-late maturing variety ‘Shapody’, which is similar to the findings of Ospina [16] and Tessema [17]. The above indicates that in addition to the amount of nitrogen applied, the form of the nitrogen is also an important factor influencing the model parameters.

4.2. Factors Affecting Plant Dry Matter Accumulation

Canopy light interception and radiation use efficiency are essential for improving dry matter accumulation and yield in field crops [29]. Previous studies have shown that changes in canopy cover can reflect the light interception capacity of plants during the growing season [30,31,32]. Our results further demonstrated that the integral area (A_t) of canopy cover over thermal days had a significant direct effect on plant dry matter accumulation, showing a strong linear relationship between the two.

Radiation use efficiency (RUE) is an important measure of the dry matter produced by a plant per unit of light energy [16]. RUE is strongly correlated with net photosynthetic rate under different environmental conditions [33,34]. Previous studies have found that although net photosynthetic rate (P_n) at the leaf level varies significantly under some conditions, its direct effect on total dry weight of the whole plant is not significant, while total net CO₂ assimilation of the plant and dry weight yield are closely related, and temperature and leaf area index are the most important factors affecting RUE [34,35,36]. Our results showed that the correlation with total plant dry weight was only slightly improved by combining P_n with A_t data. This may be due to the fact that leaf-level P_n does not fully represent the net photosynthetic capacity of the population canopy, and P_n varies significantly at different growth stages of the crop and at different times of the day.

4.3. Relationship between Model Parameters and Tuber Yield

V_max is closely related to tuber yield, and a high V_max usually means that the plant is able to intercept more light, thereby increasing the accumulation of photosynthetic products and yield [10,18]. Our results showed that V_max was significantly and positively correlated with tuber yield for single varieties, whereas there was no significant correlation between V_max and tuber yield for mixed varieties. Khan [18] found a significant positive correlation between C_m1 and tuber dry weight. The canopy development parameters DP2 (i.e., t₂−t₁) and t_e under drought conditions significantly affected yield, with longer DP2 and later t_e contributing to higher yield under drought conditions [11]. However, our results showed that the strength of the correlation between C_m1, t₂−t₁, and t_e with yield was closely related to varieties, suggesting that while canopy development dynamics are closely related to light interception, the correlation of most canopy development parameters with tuber yield can be strongly influenced by genetic characteristics of different varieties.

In our study, A_sum was highly positively correlated with tuber yield, which is consistent with the findings of Khan et al. [18,37] and Ospina et al. [27]. Furthermore, our results showed that although A_sum had the greatest positive direct effect on yield, A_tII had the largest total effect on yield. This suggests that increasing the integrated area of canopy cover over time during the peak flowering period of potatoes is the optimal strategy to improve yield. However, it is important to note that while our findings demonstrated the effects of nitrogen treatments on canopy development and tuber yield, other environmental factors—such as temperature [38], soil moisture [39], irrigation practices, and soil properties [40]—may also influence canopy growth and yield. These factors could lead to variations in results despite consistent nitrogen treatments, highlighting the limitations of our study.

5. Conclusions

In this study, we developed a potato canopy development model based on canopy cover data extracted from UAV RGB imagery, from which we can draw the following conclusions: (1) The model accurately characterizes the potato canopy growth patterns under different combinations of nitrogen application rates and forms and quantifies the differences in canopy development parameters. Most model parameters were more sensitive to nitrogen application forms than to application rates. Specifically, ammonium nitrogen significantly increased the values of V_max, t_e, C_m1, t₂−t₁, t_e−t₂, A₂, A₃, and A_sum, while the values of t₁ and A₁ decreased significantly; (2) the parameter A_t showed a significant positive correlation with plant dry matter accumulation, indicating that the integral area of canopy cover over thermal time is a reliable predictor of plant dry matter status; (3) the integral area of canopy cover during the peak flowering stage of potato had a significant total effect on yield. Therefore, we speculate that increasing the integral area during this critical period, through optimized nitrogen management and other agronomic practices, could effectively improve potato yield.

In the future, we can further explore the optimal combination of nitrogen application forms and doses for potatoes at each growth stage and improve the applicability and predictive accuracy of the model through refined optimization, integrated management practices, and long-term experimental validation.