Vila et al. (2025) Potential of thermal imaging for yield and soil water content prediction in leafy vegetables

Identification

• Journal: Smart Agricultural Technology
• Year: 2025
• Date: 2025-10-31
• Authors: Vinícius Villa e Vila, Silas Alves Souza, Fernando Campos Mendonça, Tamara Maria Gomes, Peterson Ricardo Fiorio, Patrícia Angélica Alves Marques
• DOI: 10.1016/j.atech.2025.101587

Research Groups

• University of São Paulo/USP, Luiz de Queiroz College of Agriculture/ESALQ, Biosystems Engineering Department, Piracicaba, SP, Brazil
• University of São Paulo/USP, Faculty of Animal Science and Food Engineering/FZEA, Biosystems Engineering Department, Pirassununga, SP, Brazil

Short Summary

This study developed predictive models for yield and soil water content in lettuce and arugula by integrating thermal images. The models, based on Crop Water Stress Index (CWSI) and normalized temperature difference (ΔT), demonstrated good performance for yield (R² up to 0.82) and soil water content (R² up to 0.92), providing critical thresholds for efficient irrigation management.

Objective

• To develop predictive models of yield and soil water content in leafy vegetables (lettuce and arugula) by integrating thermal images obtained using infrared radiation cameras.

Study Configuration

• Spatial Scale: Two experiments conducted in a protected environment (greenhouse) at the University of São Paulo, Piracicaba, São Paulo, Brazil. Plants were grown in beds of 0.5 square meters effective area, with specific plant spacing for lettuce (8 plants per bed) and arugula (100 plants per bed). Thermal images were acquired at a standardized distance of 1.5 meters from the canopy.
• Temporal Scale: Two growing cycles for each crop (lettuce and arugula), each lasting 40 days (First cycle: May 30 to July 9, 2022; Second cycle: July 19 to August 28, 2022). Meteorological data were monitored every 15 minutes, and tensiometer readings were taken daily. Thermal images were acquired between 12:00 and 13:00 local time, at the final stage of development (35 days after transplanting).

Methodology and Data

• Models used:
  • FAO 56 Penman-Monteith equation for reference evapotranspiration (ETo).
  • Crop evapotranspiration (ETc) calculation (ETc = ETo * Kc).
  • Van Genuchten model for fitting the soil water retention curve.
  • Crop Water Stress Index (CWSI) calculation.
  • Normalized temperature difference (ΔT) calculation.
  • Regression and correlation analysis for developing predictive models.
  • Statistical indicators for model evaluation: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Nash–Sutcliffe Efficiency (NSE), Willmott’s index of agreement (d), Pearson correlation coefficient (r), and confidence index (c).
• Data sources:
  • Thermal images: Acquired using a FLIR T640 Duo Pro R thermal camera (spectral band 7.5–13.5 µm, resolution 640 × 512 pixels, thermal sensitivity <50 mK).
  • Soil water content: Measured using tensiometers installed at 0–0.20 m soil depth, with readings converted to volumetric soil water content (m³ m⁻³) using a soil water retention curve.
  • Crop yield: Assessed as fresh shoot biomass (leaves and stems) at harvest, extrapolated to tonnes per hectare (t ha⁻¹).
  • Meteorological data: Collected from an on-site meteorological station (air temperature, relative humidity, solar radiation, atmospheric pressure) with sensors (HMP45C, LI200X, CS106) connected to a CR23X datalogger.
  • Irrigation levels: Three treatments corresponding to 100%, 80%, and 60% of crop evapotranspiration (ETc) replacement.
  • Soil properties: Physical, hydro-physical, and chemical analyses of the 0–0.20 m soil layer.

Main Results

• Yield prediction models for lettuce showed R² values of 0.71 (CWSI) and 0.75 (ΔT), with a mean RMSE of 4.87 t ha⁻¹. For arugula, R² values were 0.82 (CWSI) and 0.79 (ΔT), with the same mean RMSE.
• Soil water content prediction models exhibited R² values of 0.92 (CWSI) and 0.73 (ΔT), with a mean RMSE of 0.00428 m³ m⁻³. The CWSI-based model demonstrated superior performance for soil water content prediction.
• Model performance, evaluated by the confidence index (c), ranged from "very good" to "great" for yield prediction (0.764 to 0.861) and from "very good" to "great" for soil water content prediction (0.787 to 0.939).
• Critical thresholds to avoid water deficit stress in lettuce and arugula were identified as CWSI values above 0.35 and ΔT values above -0.96 °C.
• Canopy temperature increased with decreasing irrigation levels (e.g., lettuce average 25.94 °C at 100% ETc vs. 29.39 °C at 60% ETc), confirming water deficit impact on plant temperature regulation.
• Yield reductions were significant under water deficit: lettuce yield decreased by 17.98% (80% ETc) and 37.67% (60% ETc) compared to 100% ETc; arugula yield decreased by 31.96% (80% ETc) and 48.36% (60% ETc).
• Soil water potential values confirmed water stress in 80% and 60% ETc treatments, with only the 100% ETc treatment maintaining values above the -20 kPa threshold for leafy vegetables.

Contributions

• Developed and validated robust predictive models for crop yield and soil water content in leafy vegetables (lettuce and arugula) using thermal imaging, addressing a gap in quantitative irrigation management for these crops.
• Quantified the strong inverse relationships between thermal indices (CWSI and ΔT) and both crop yield and soil water content, providing practical tools for real-time assessment.
• Established specific critical thresholds for CWSI (0.35) and ΔT (-0.96 °C) that can be directly applied in irrigation scheduling to prevent water deficit stress in leafy vegetables.
• Demonstrated the potential of thermal imaging as a non-destructive, rapid, and efficient technology for early detection of water stress and prediction of yield gaps in short-cycle, water-sensitive crops.

Funding

• Coordination for the Improvement of Higher Education Personnel (CAPES), finance code 001.
• National Council for Scientific and Technological Development (CNPq – Case number 156142/2021–0).
• Luiz de Queiroz Agricultural Studies Foundation (FEALQ).

Citation

@article{Vila2025Potential,
  author = {Vila, Vinícius Villa e and Souza, Silas Alves and Mendonça, Fernando Campos and Gomes, Tamara Maria and Fiorio, Peterson Ricardo and Marques, Patrícia Angélica Alves},
  title = {Potential of thermal imaging for yield and soil water content prediction in leafy vegetables},
  journal = {Smart Agricultural Technology},
  year = {2025},
  doi = {10.1016/j.atech.2025.101587},
  url = {https://doi.org/10.1016/j.atech.2025.101587}
}

Generated by BiblioAssistant using gemini-2.5-flash (Google API)