Treffer: مدل‌سازی تبخیر – تعرق خیار گلخانه‌ای با استفاده از یادگیری ماشین: رویکرد جنگل تصادفی در مقابل ضرایب گیاهی سنتی و غیرخطی.

Introduction Accurate estimation of crop evapotranspiration (ET<subscript>c</subscript>) is fundamental for the development of efficient irrigation strategies in greenhouse systems, where environmental conditions differ significantly from open-field farming. In Iran, greenhouse agriculture, particularly for cucumbers (Cucumis sativus L.), has expanded considerably, making irrigation optimization a critical priority. The specific microclimate within greenhouses, including controlled humidity, temperature, and solar radiation levels, affects plant water needs, requiring tailored approaches to predict ET<subscript>c</subscript>. Traditional models, like the FAO 56 crop coefficient (K<subscript>c</subscript>) method, provide a standardized way to estimate ET<subscript>c</subscript> but are generally suited to field crops under variable outdoor conditions. The limitations of fixed K<subscript>c</subscript> values in capturing the complexity of greenhouse environments have prompted the exploration of alternative models. In recent years, machine learning (ML) techniques, especially ensemble methods like the Random Forest (RF) algorithm, have emerged as promising tools for ET<subscript>c</subscript> modeling due to their capacity to manage non-linear interactions among meteorological variables and enhance model flexibility. This study evaluates the performance of three ET<subscript>c</subscript> estimation approaches for greenhouse-grown cucumber: the conventional FAO 56 K<subscript>c</subscript> method, a nonlinear K<subscript>c</subscript> model using a third-degree polynomial, and direct ET<subscript>c</subscript> prediction through the RF algorithm. These methods are assessed across two growth cycles, autumn-winter (A-W) and spring-summer (S-S), to capture seasonal differences in crop water requirements. Materials and Methods The study was conducted in a research greenhouse located at the College of Agriculture and Natural Recourses, University of Tehran, focusing on daily ET<subscript>c</subscript> of cucumber over two distinct growth periods. Environmental parameters were measured both inside and outside the greenhouse, including maximum, minimum, and average temperatures, relative humidity, and solar radiation. Reference evapotranspiration inside the greenhouse (ET<subscript>o</subscript>G) was derived using a micro-lysimeter installed with a turfgrass surface, while daily ET<subscript>c</subscript> was measured using a soil water balance method, where soil moisture content was monitored daily across three experimental plots to ensure precision. ET<subscript>c</subscript> calculations were performed through three modeling approaches. In the first approach, the FAO 56 K<subscript>c</subscript> model estimated ET<subscript>c</subscript> by applying fixed crop coefficients and multiplying them by ET<subscript>o</subscript>G. Although this method has been widely applied in field conditions, its applicability to greenhouses is limited due to fixed K<subscript>c</subscript> assumptions. In the second approach, a non-linear K<subscript>c</subscript> model was developed using third-degree polynomial regression on K<subscript>c</subscript> values calculated as the ratio of ET<subscript>c</subscript> to ET<subscript>o</subscript>G, capturing growth-stage-specific variations. In the final approach, the RF model directly predicted ET<subscript>c</subscript> based on a broad range of meteorological inputs. To optimize the RF model, hyperparameters were tuned using Python’s GridSearchCV tool, and data were split into training (70%) and testing (30%) sets to validate model performance. After initial RF modeling, a feature selection process using Permutation Feature Importance (PFI) was applied to identify the most influential variables, refining the RF model to the top four parameters. Results and Discussion The results highlighted seasonal variability in cumulative ET<subscript>c</subscript>, with the S-S period exhibiting nearly double the ET<subscript>c</subscript> of the A-W period due to higher ambient temperature and increased solar radiation. These findings underscore the necessity of dynamic ET<subscript>c</subscript> models that can accommodate seasonal and environmental variations. The FAO 56 K<subscript>c</subscript> method produced a mean RMSE of 0.915 mm/day across both growth cycles, demonstrating limitations in fixed K<subscript>c</subscript> approaches under greenhouse conditions. The non-linear K<subscript>c</subscript> model, with an average RMSE of 0.64 mm/day, provided improved accuracy by adjusting K<subscript>c</subscript> values across different growth stages, especially during mid-growth when water demand peaks. This improvement aligns with the premise that non-linear models can better capture the ET<subscript>c</subscript> variability within controlled environments. The RF algorithm demonstrated superior accuracy and flexibility, outperforming both K<subscript>c</subscript>-based models with R² values of 0.96 for the A-W period and 0.94 for the S-S period in training datasets, and with respective RMSE values of 0.365 mm/day and 0.57 mm/day in testing datasets. These results illustrate the RF model’s capacity to accurately model ET<subscript>c</subscript> by capturing complex, non-linear interactions among variables such as maxTG (maximum temperature inside the greenhouse), meanRHG (average relative humidity inside the greenhouse), and RadiationG (solar radiation inside the greenhouse) during the A-W period, with RadiationG and ET<subscript>o</subscript>G emerging as key variables during the S-S period. By emphasizing critical seasonal drivers of ET<subscript>c</subscript>, the RF model offers a robust alternative that adjusts to environmental changes without relying on static K<subscript>c</subscript> values. This adaptability supports RF’s potential as a powerful tool for ET<subscript>c</subscript> estimation, accurately reflecting seasonal influences on greenhouse crop water needs. Conclusion The findings from this study demonstrate that the RF algorithm, when applied to ET<subscript>c</subscript> modeling in greenhouse conditions, provides a flexible, high-accuracy alternative to traditional K<subscript>c</subscript> methods. Unlike the FAO 56 K<subscript>c</subscript> and non-linear K<subscript>c</subscript> models, which rely on predefined or growth-stage specific coefficients, the RF approach enables direct ET<subscript>c</subscript> prediction using real-time meteorological data. By optimizing input variables through feature selection, RF efficiently reduced the model complexity, focusing on the top four influential parameters while retaining high predictive accuracy. This reduction not only streamlines data collection requirements but also enhances the model's applicability in practical greenhouse operations. The results indicate that RF's capacity to model complex relationships among variables makes it especially suited for greenhouse environments, where precision irrigation is crucial for sustainable water management. Ultimately, this research underscores the importance of integrating machine learning techniques in ET<subscript>c</subscript> estimation, providing greenhouse operators with adaptive, resource-efficient tools for managing water use in controlled agricultural settings. [ABSTRACT FROM AUTHOR]

تخمین مناسب تبخیر – تعرق گیاه (ET<subscript>c</subscript>) برای برنامه‌ریزی آبیاری کارآمد بسیار حیاتی است. این مطالعه با هدف ارزیابی و مدل‌سازی ET<subscript>c</subscript> و ضریب گیاهی (Kc) خیار گلخانه‌ای در دو دوره رشد پاییز – زمستان 1401 (A-W) و بهار – تابستان 1402 (S-S) در گلخانه تحقیقاتی پردیس کشاورزی و منابع طبیعی دانشگاه تهران انجام شد. تبخیر – تعرق مرجع داخل گلخانه (ET<subscript>o</subscript>G) با استفاده از میکرولایسیمتر چمن و ET<subscript>c</subscript> با استفاده از روش بیلان آب خاک و اندازه‌گیری روزانه رطوبت خاک در سه کرت آزمایشی انجام شد. Kc روزانه خیار گلخانه‌ای از نسبت ET<subscript>c</subscript> به ET<subscript>o</subscript>G محاسبه شد. ابتدا مدل FAO56 برای برآورد Kc (Kc <subscript>FAO56</subscript>) و ET<subscript>c</subscript> حاصل از حاصل‌ضرب آن در ET<subscript>o</subscript>G (ET<subscript>c</subscript> = Kc FAO56 × ET<subscript>o</subscript>G) مورد ارزیابی قرار گرفت که میانگین NRMSE و ME مربوط به ET<subscript>c</subscript> در دو دوره رشد موردبررسی به ترتیب برابر 31.2 درصد و 0.62 به‌دست آمد. در مرحله دوم از یک معادله درجه سه غیرخطی به‌منظور برآورد Kc و ET<subscript>c</subscript> حاصل از حاصل‌ضرب آن در ET<subscript>o</subscript>G (ET<subscript>c</subscript> = Kc <subscript>NonLinear </subscript>× ET<subscript>o</subscript>G) استفاده شد که میانگین NRMSE و ME در این روش نسبت به روش قبلی بهبودیافته و به ترتیب برابر 21.58 درصد و 0.82 به دست آمد. در مرحله سوم از الگوریتم جنگل تصادفی (RF) به‌منظور مدل‌سازی مستقیم ET<subscript>c</subscript> خیار گلخانه‌ای با استفاده از ET<subscript>o</subscript>G و پارامترهای هواشناسی اندازه‌گیری‌شده در داخل و بیرون گلخانه استفاده شد. نتایج ET<subscript>c</subscript> حاصل از الگوریتم RF نشان داد که این الگوریتم با داشتن انعطاف‌پذیری و دقت بالا توانست پیش‌بینی‌های قابل قبول تری از ET<subscript>c</subscript> ارائه دهد. میانگین NRMSE و ME مربوط به الگوریتم RF در دو دوره رشد به ترتیب برابر با 6.8 درصد و 0.95 در مرحله آموزش و 11.3 درصد و 0.86 در مرحله آزمایش بدست آمد. در مرحله سوم با بررسی اهمیت ویژگی‌ها به روش Permutation Feature Importance (PFI) و انتخاب تنها چهار پارامتر اندازه‌گیری‌شده در گلخانه به‌عنوان ورودی RF می‌توان بدون نیاز به Kc خطی یا غیرخطی، ET<subscript>c</subscript> را با دقت بالایی برآورد کرد. این تحقیق اهمیت استفاده از روش‌های یادگیری ماشین در مدل‌سازی ET<subscript>c</subscript> گلخانه‌ای را نشان داده و می‌تواند به بهبود مدیریت آب در گلخانه‌ها کمک کند. [ABSTRACT FROM AUTHOR]

Copyright of Water & Soil Management & Modeling / Mudil Sazī va Mudīriyyat-i āb va Khāk is the property of University of Mohaghegh Ardabili, Water Management Research Center and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)