Abstract
A major limitation in perennial orchard management is the lack of integrative frameworks that connect canopy nitrogen (N) dynamics with canopy structure, environmental drivers, yield formation, and postharvest quality at the individual-tree scale. This study presents a multi-sensor, tree-level framework combining unmanned aerial vehicle (UAV) multispectral imagery, LiDAR-derived canopy structure (e.g., height, canopy area, plant area index), and environmental variables (e.g., temperature, vapor pressure deficit, and evapotranspiration) within an explainable machine-learning approach to quantify leaf and canopy nitrogen content, yield, and fruit quality in avocado. The framework was evaluated across three contrasting systems: a controlled lysimetric gradient (Gilat; N10–N120 mg N L⁻¹), a semi-arid commercial orchard (Kfar Menachem), and a humid commercial orchard (Kabri), using multi-temporal UAV acquisitions and ground measurements collected during 2022–2024. Random Forest models accurately predicted leaf nitrogen concentration (LNC; R² = 0.92/0.78/0.78 for calibration/validation/prediction) and canopy nitrogen content (CNC; R² = 0.90/0.76/0.73 for calibration/validation/prediction), indicating reliable scaling from leaf to canopy level. Yield prediction showed robust performance (R² = 0.90/0.75/0.71 for calibration/validation/prediction), with consistent results across sites. Postharvest quality models showed high accuracy (R² = 0.83-0.92), with shapley additive explanations (SHAP) analysis indicating that late-season nitrogen status, canopy structure, and meteorological conditions contributed more strongly to decay risk than fertilizer input alone. Nitrogen-use efficiency peaked at an intermediate nitrogen input (∼N20), indicating a constrained optimal range in the controlled lysimetric experiment. The framework links nitrogen status, canopy structure, and environmental conditions with yield and fruit quality outcomes, supporting tree-scale diagnostics for improved nitrogen management, while highlighting the need for further validation across sites, seasons, and cultivars.
Abstract from the paper’s OpenAlex record; open copy licensed cc-by-nc-nd. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 6
Claim · model-written summary
Random Forest predicted avocado leaf nitrogen concentration with R² = 0.92/0.78/0.78 performance.
- Taxon
- avocado
Persea americanat0 - Outcome
- leaf nitrogen concentratione0
- Value
- R² = 0.92/0.78/0.78e0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“Random Forest models accurately predicted leaf nitrogen concentration (LNC; R² = 0.92/0.78/0.78 for calibration/validation/prediction) and canopy nitrogen content (CNC; R² = 0.90/0.76/0.73 for calibration/validation/prediction), indicating reliable scaling from leaf to canopy level.”
t0“Avocado”
Claim · model-written summary
Random Forest predicted avocado canopy nitrogen content with R² = 0.90/0.76/0.73 performance.
- Taxon
- avocadoPersea americanat0
- Outcome
- canopy nitrogen contente0
- Value
- R² = 0.90/0.76/0.73e0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“Random Forest models accurately predicted leaf nitrogen concentration (LNC; R² = 0.92/0.78/0.78 for calibration/validation/prediction) and canopy nitrogen content (CNC; R² = 0.90/0.76/0.73 for calibration/validation/prediction), indicating reliable scaling from leaf to canopy level.”
t0“Avocado”
Claim · model-written summary
Random Forest predicted avocado Yield prediction with R² = 0.90/0.75/0.71 performance.
- Taxon
- avocadoPersea americanat0
- Outcome
- Yield predictione0
- Value
- R² = 0.90/0.75/0.71e0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“Yield prediction showed robust performance (R² = 0.90/0.75/0.71 for calibration/validation/prediction), with consistent results across sites.”
t0“Avocado”
Claim · model-written summary
Random Forest predicted avocado Postharvest quality models with R² = 0.83-0.92 performance.
- Taxon
- avocadoPersea americanat0
- Outcome
- Postharvest quality modelse0
- Value
- R² = 0.83-0.92e0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“Postharvest quality models showed high accuracy (R² = 0.83-0.92), with shapley additive explanations (SHAP) analysis indicating that late-season nitrogen status, canopy structure, and meteorological conditions contributed more strongly to decay risk than fertilizer input alone.”
t0“Avocado”
Claim · model-written summary
Late-season nitrogen status, canopy structure and meteorological conditions contributed more to decay risk than fertilizer input alone.
- Taxon
- avocadoPersea americanat0
- Compound
- nitrogene0
- Outcome
- decay riske0
- Direction
- more stronglye0
- Comparator
- fertilizer input alonee0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“shapley additive explanations (SHAP) analysis indicating that late-season nitrogen status, canopy structure, and meteorological conditions contributed more strongly to decay risk than fertilizer input alone.”
t0“Avocado”
Claim · model-written summary
Nitrogen-use efficiency peaked at intermediate nitrogen input.
- Taxon
- avocadoPersea americanat0
- Treatment
- intermediate nitrogen input (∼N20)e0
- Outcome
- Nitrogen-use efficiencye0
- Direction
- peakede0
- Setting
- controlled lysimetric experimente0
Not reportedComparator
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“Nitrogen-use efficiency peaked at an intermediate nitrogen input (∼N20), indicating a constrained optimal range in the controlled lysimetric experiment.”
t0“Avocado”