Determination of Water Stress in Okra Plants Using Canopy Temperature Measurement by Infrared Thermometer

Abstract

Water scarcity represents one of the most severe constraints on agricultural productivity in semi-arid and water-stressed cultivation zones. In traditional vegetable cultivation, farmers predominantly rely on visual symptoms of wilting to detect water deficit stress, which typically manifest only after substantial physiological damage, cellular dehydration, and irreversible yield loss have occurred. This study presents a non-destructive, precision agriculture framework for the early detection and classification of water stress in okra (Abelmoschus esculentus) using handheld infrared thermometry coupled with an ensemble Random Forest machine learning classifier. A primary field dataset comprising 100 observations was acquired under variable diurnal meteorological conditions in Jammu, India. The recorded and derived parameters encompassed canopy surface temperature (Tc), wet-bulb reference temperature (Twet), dry-bulb reference temperature (Tdry), ambient air temperature (Ta), relative humidity (RH), air saturation vapor pressure (SVPair), actual vapor pressure (AVP), leaf saturation vapor pressure (SVPleaf), leaf vapor pressure deficit (VPDleaf), canopy-air thermal differential (Tc – Ta), and the Crop Water Stress Index (CWSI). Using an empirical CWSI threshold of 0.30, samples were categorized into non-stressed (CWSI ? 0.30) and stressed (CWSI > 0.30) physiological states. The dataset was partitioned into an 80:20 training and testing split (80 training samples, 20 testing samples). The trained Random Forest classifier achieved an overall classification accuracy of 95.0% (19/20 correct classifications) on the unseen test set. For the non-stressed class, the model demonstrated a precision of 1.00, recall of 0.92, and an F1-score of 0.96 (support = 12). For the water-stressed class, the model yielded a precision of 0.89, recall of 1.00, and an F1-score of 0.94 (support = 8), with zero false negatives (FN = 0), ensuring that no stressed crops were missed. Gini feature importance analysis revealed that dry-bulb reference temperature (Tdry, score = 0.187), leaf saturation vapor pressure (SVPleaf, score = 0.172), wet-bulb reference temperature (Twet, score = 0.160), leaf vapor pressure deficit (VPDleaf, score = 0.112), and Tc – Ta (score = 0.101) were the primary drivers governing classification. The findings confirm that coupling thermal radiometry with psychrometric feature engineering and ensemble learning provides a reliable, non-contact diagnostic mechanism for precision irrigation scheduling in smallholder horticulture.

Introduction

The text presents a precision-irrigation approach for detecting water stress in okra using infrared thermometry, environmental measurements, and Random Forest machine learning.

• Background: Agriculture is a major economic activity in developing countries but faces increasing pressure from freshwater scarcity, climate variability, and population growth. Irrigated agriculture consumes about 70% of global freshwater withdrawals, while conventional flood and furrow irrigation often has low efficiency. Precision irrigation can reduce water use while maintaining or improving crop yields.

• Problem in okra: Okra is particularly sensitive to soil-water shortages during vegetative growth and flowering. Water stress can cause flower drop, poor pod development, tougher pods, and reduced yield. Traditional visual indicators such as wilting detect stress relatively late.

• Proposed approach: The study uses a handheld infrared thermometer to measure canopy temperature, along with air temperature, relative humidity, and artificial wet/dry reference temperatures. Because stressed plants close their stomata and lose less water through transpiration, their canopies become warmer. These measurements were converted into variables such as vapor pressure deficit (VPD), canopy-air temperature difference, and Crop Water Stress Index (CWSI).

• Experimental setup: The experiment was conducted on okra in Jammu, India, under well-watered and progressive deficit-irrigation conditions. Measurements were collected around midday under clear skies. A dataset of 100 observations was divided into 80% training and 20% testing data.

• Stress classification: A CWSI threshold of 0.30 was used to distinguish non-stressed plants (≤0.30) from water-stressed plants (>0.30). The observed CWSI distributions showed a clear separation between the two groups, with median values of 0.17 for non-stressed and 0.47 for stressed plants.

• Machine-learning results: The Random Forest classifier achieved 95% accuracy on the independent 20-sample test set. It correctly identified all 8 water-stressed samples (100% recall) and misclassified only one non-stressed sample as stressed. This indicates that the method was particularly effective at avoiding missed cases of water stress in this experiment.

• Important predictors: The most influential features were dry reference temperature, leaf saturation vapor pressure, wet reference temperature, and leaf VPD. Together, the wet and dry reference temperatures contributed about 35% of the model's feature importance, highlighting the value of normalizing canopy temperature against environmental reference conditions.

• Limitations: The study was limited by changing weather conditions, the need to maintain artificial wet/dry reference targets, measurements from only one okra cultivar and one geographic location, and discrete midday observations rather than continuous monitoring.

• Future scope: The authors propose integrating low-cost infrared sensors with IoT devices, using UAV thermal imaging for field-scale monitoring, applying CNN-based computer vision, and developing smartphone applications that could provide real-time irrigation recommendations

Conclusion

This investigation demonstrated that coupling handheld infrared thermometry with psychrometric environmental features and Random Forest ensemble learning provides a highly accurate (95.0%), non-destructive framework for early water stress diagnosis in okra. The complete elimination of false negatives (recall = 1.00 for stressed plants) validates its practical agronomic utility. Future initiatives will: (1) interface long-wave infrared sensors (e.g., MLX90614) with IoT edge microcontrollers for autonomous canopy monitoring; (2) deploy thermal imaging cameras on UAVs for spatial field mapping [17], [20]; (3) incorporate computer vision CNN architectures; and (4) develop smartphone applications to deliver real-time irrigation advisories directly to smallholder farmers.

References

[1] D. E. Evans, E. J. Sadler, C. R. Camp, and J. A. Millen, \"Spatial canopy temperature measurements using centre pivot mounted IRTs,\" in Proc. 5th Int. Conf. Precision Agric., Bloomington, MN, USA, 2000, pp. 1–14. [2] H. G. Jones, M. Stoll, T. Santos, C. de Sousa, M. M. Chaves, and O. M. Grant, \"Use of infrared thermography for monitoring stomatal closure in the field: application to grapevine,\" J. Exp. Bot., vol. 53, no. 378, pp. 2249–2260, 2002, doi: 10.1093/jxb/erf083. [3] A. R. Aston and C. H. M. van Bavel, \"Soil surface water depletion and leaf temperature,\" Agron. J., vol. 64, no. 3, pp. 368–373, 1972, doi: 10.2134/agronj1972.00021962006400030034x. [4] T. B. Raper, D. M. Oosterhuis, and E. M. Barnes, \"In-season cotton drought-stress quantification: Previous approaches and future directions,\" J. Cotton Sci., vol. 20, no. 3, pp. 179–192, 2016. [5] J. L. Van Zyl, \"Canopy temperature as a water stress indicator in vines,\" S. Afr. J. Enol. Vitic., vol. 7, no. 2, pp. 64–70, 1986, doi: 10.21548/7-2-2326. [6] J. T. Scott and J. N. Gallagher, \"An assessment of infra-red thermometry for scheduling irrigation of bean crops,\" Agron. N. Z., vol. 15, pp. 27–34, 1985. [7] J. C. O\'Toole and J. L. Hatfield, \"Effect of wind on the crop water stress index derived by infra-red thermometry,\" Agron. J., vol. 75, no. 5, pp. 811–817, 1983, doi: 10.2134/agronj1983.00021962007500050019x. [8] M. Fuchs, \"Infrared measurements of canopy temperature and detection of plant water stress,\" Theor. Appl. Climatol., vol. 42, no. 4, pp. 253–261, 1990, doi: 10.1007/BF00865986. [9] H. G. Jones, \"Use of infrared thermometry for estimation of stomatal conductance as a possible aid to irrigation scheduling,\" Agric. For. Meteorol., vol. 95, no. 3, pp. 139–149, 1999, doi: 10.1016/S0168-1923(99)00030-1. [10] J. P. A. Gardner, \"Infrared thermometry and the crop water stress index,\" in Proc. Irrig. Assoc. Int. Expo. Tech. Conf., New Orleans, LA, USA, 1992, pp. 267–275. [11] C. K. Parry, \"Biophysically based measurement of plant water status using canopy temperature,\" Ph.D. dissertation, Dept. Plants, Soils, and Climate, Utah State Univ., Logan, UT, USA, 2014, doi: 10.26076/a6aa-e1ed. [12] M. S. Akhtar, S. Ahmad, and M. A. Randhawa, \"Infrared thermometry as a rapid and non-destructive technique for screening drought tolerance in wheat genotypes,\" Plant Physiol. Biochem., vol. 127, pp. 411–418, 2018, doi: 10.1016/j.plaphy.2018.03.005. [13] I. Alves and L. S. Pereira, \"Non-water-stressed baselines for irrigation scheduling with infrared thermometers,\" in Proc. 5th Int. Conf. Precision Agric., Bloomington, MN, USA, 2000, pp. 1–12. [14] S. B. Idso, R. D. Jackson, P. J. Pinter, R. J. Reginato, and J. L. Hatfield, \"Normalizing the stress-degree-day parameter for environmental variability,\" Agric. Meteorol., vol. 24, pp. 45–55, 1981, doi: 10.1016/0002-1571(81)90032-7. [15] R. D. Jackson, S. B. Idso, R. J. Reginato, and P. J. Pinter, \"Canopy temperature as a crop water stress indicator,\" Water Resour. Res., vol. 17, no. 4, pp. 1133–1138, 1981, doi: 10.1029/WR017i004p01133. [16] L. Breiman, \"Random Forests,\" Mach. Learn., vol. 45, no. 1, pp. 5–32, 2001, doi: 10.1023/A:1010933404324. [17] A. Del Pozo, R. Calderón, J. Peña, V. González-Dugo, and P. J. Zarco-Tejada, \"Evaluation of crop water stress index for irrigation scheduling in walnut orchards,\" Plant Methods, vol. 21, no. 1, Art. no. 12, 2025, doi: 10.1186/s13007-025-01334-x. [18] S. Taghvaeian, S. A. O’Shaughnessy, S. R. Evett, M. A. Andrade, P. D. Colaizzi, and J. A. Tolk, \"Challenges and opportunities in precision irrigation decision-support systems for center pivots,\" Environ. Res. Lett., vol. 16, no. 5, Art. no. 053003, 2021, doi: 10.1088/1748-9326/abf129. [19] D. R. Cutler, T. C. Edwards, K. H. Beard, A. Cutler, K. T. Hess, J. Gibson, and J. J. Lawler, \"Random forests for classification in ecology,\" Ecology, vol. 88, no. 11, pp. 2783–2792, 2007, doi: 10.1890/07-0539.1. [20] M. E. Karar, F. Alotaibi, A. Al Rasheed, and O. Reyad, \"A pilot study of smart agricultural irrigation using unmanned aerial vehicles and IoT-based cloud system,\" Inf. Process. Agric., vol. 8, no. 3, pp. 415–427, 2021, doi: 10.1016/j.inpa.2020.07.003.

Copyright

Copyright © 2026 Muneeb Ajmer, Sayam Prajapati, Somil Narang, Rudraksh Sharma, Saksham Khajuria. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.