Key topics included:
1) Multi-platform remote sensing applications (satellite, UAV, hyperspectral, SAR, thermal infrared) for estimation of soil moisture and crop water requirement.
2) Deep learning and AI-driven crop water stress detection and yield prediction.
3) Precision irrigation systems and automated water management technologies.
4) Multi-sensor data fusion combining space-based, airborne, and ground observations.
5) Real-time monitoring systems and subseasonal-to-seasonal water demand forecasting models.
6) Digital agriculture platforms and decision support tools for sustainable water management.
This session aims to showcase practical solutions that bridge the gap between technological innovation and real-world agricultural applications, emphasizing scalable approaches that support both productivity and environmental sustainability.
Orals: Mon, 4 May, 10:45–12:25 | Room 2.17
Smallholder farmers frequently rely on thumb rules assuming that higher fertilizer inputs guarantee higher yields due to the absence of site-specific edapho-climatological data. This dependence on generalized rules creates a disconnect between site-specific requirements and field management practices, necessitating modeling field dynamics and providing actionable advisories to farmers. To address this disconnect, this study developed ‘CropLizer’ a machine learning and remote sensing based tool (https://mukundn1997-croplizer.hf.space/) to function as an integrated decision support system for rice cultivation. To develop CropLizer, this study synthesized a comprehensive dataset comprising over 45,000 rice field points (60% was reserved for training and the rest for validation) integrated with broadly yields, irrigation, nutrient practices, social status (education and ethnic group), climatic variables (precipitation), soil quality variables (carbon, nitrogen, and bulk density), as well as market accessibility. Subsequently, seven models (linear, support vector, decision tree, random forest, neural network, LSTM, transformer) were trained and hyperparameter tuned to predict yield and fertilizer requirements based on 43 agro-edapho-socio-climatological variables through ‘sklearn’, ‘tensorflow’, and ‘optuna’ libraries in python using IIT Roorkee’s super computer PARAMGANGA. After optimization, a web application was developed to allow users to simulate different scenarios by adjusting specific farming inputs to identify optimal management practices. Consequently, the system generates prescriptions for nitrogen and phosphorus and potassium application rates based on the predicted yields. Moreover, a user could find the potential yield for their field and what adjustments in field practices are required to obtain the potential yield sustainably (without loss of soil carbon). Considering the practical difficulties of gathering meteorological record and soil data, an application programme interface was set up for automatic retrieval of these variables from the field coordinates from open-meteo and soilgrids datasets. Upon validation, the performance of the best performing model (random forest) demonstrated a satisfactory accuracy (65%). Beyond agronomic parameters the tool calculates economic viability by integrating local market prices to estimate potential net profit margins and benefit-cost ratios under current yield and potential yield. This framework bridges the gap between scientific research and field application by providing assured predictions for pre-season planning to mitigate financial risks.
How to cite: Narayanan, M., Sharma, A., and Ilampooranan, I.: CropLizer: An Agro-Socio-Edapho-Climatological Tool for Rice Nutrient Management and Profitability Assessment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1114, https://doi.org/10.5194/egusphere-egu26-1114, 2026.
Winter wheat, maize, rice, and sugarcane are among the most important crops for global food security and bioenergy production. However, consistent high-resolution crop distribution maps across large regions and long time periods remain limited. In this study, we developed crop-specific identification algorithms that integrate spectral and phenological characteristics derived from satellite observations. Using these methods, we generated high-resolution (≤30 m) distribution maps for winter wheat, maize, rice, and sugarcane in China from 2001 to 2024. In addition, we produced sugarcane maps for Brazil (2016–2019), global winter cereal maps (2017–2022), and rice maps across Asia (1990–2023). Validation against independent samples shows that producer’s and user’s accuracies for winter wheat, maize, and rice in China reached 89.3% and 90.6%, 76.2% and 81.6%, and 88.4% and 89.1%, respectively. The global winter cereal maps achieved producer’s and user’s accuracies of 81.1% and 87.9%, while overall accuracies for sugarcane exceeded 91% in both China and Brazil. Estimated crop planting areas exhibit strong agreement with official statistics across regions. The resulting datasets provide consistent, long-term, and high-resolution crop distribution information, offering valuable support for crop monitoring, food security assessment, and climate and land-use change studies.
How to cite: Peng, Q., Shen, R., Fu, Y., Dong, J., Pan, B., Zheng, Y., Chen, X., Li, S., Li, X., and Yuan, W.: High-resolution long-term mapping of major crops using satellite data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4845, https://doi.org/10.5194/egusphere-egu26-4845, 2026.
Global climate change has led to more frequent and severe droughts in the middle and lower reaches of the Yangtze River, intensifying the spatiotemporal variability of crop yields in this region. Winter rapeseed, a major oilseed crop in China, is particularly vulnerable to these drought conditions, which now pose greater risks to local food security. Accurate and timely regional yield predictions are increasingly important for effective agricultural management and disaster response. However, predicting rapeseed yield at the city level is challenging due to complex climate patterns and the strengthened impact of drought. Addressing these challenges requires the integration of multi-source data, including both remote sensing and weather data, to capture the full range of environmental influences on crop growth. Traditional statistical and machine learning methods have often proven inadequate for robust, transferable yield prediction across different regions and years.
This study presents a deep learning–based yield prediction framework that integrates multi-temporal remote sensing indicators and meteorological variables to estimate winter rapeseed yield under both normal and drought conditions. Using data from 2014 to 2023 for the middle and lower reaches of the Yangtze River, an Attention–Long Short-Term Memory (Attention-LSTM) model was developed by jointly incorporating time-series remote sensing indices, meteorological factors, and statistical yield records. Key phenological periods for yield estimation were identified through multi-temporal and multi-variable combinations, and input configurations were systematically optimized. The proposed framework outperformed LSTM, Random Forest, and Support Vector Regression models, achieving an R2 of 0.81 and RMSE of 306.73 kg/ha on the validation dataset. Spatiotemporal yield dynamics and regional applicability were further analyzed, and the model’s robustness and adaptability were assessed under drought conditions. Under drought scenarios, the model maintained high accuracy, with an R2 of 0.76 and RMSE of 358.32 kg/ha. These results indicate the framework’s potential for drought-resilient yield prediction and its value for agricultural management and drought assessment under future climate change.
How to cite: Liu, S., Dong, S., and Guan, Q.: Iintegrating deep learning and multi-source datasets for drought-resilient winter rapeseed yield prediction in the Yangtze River Basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2352, https://doi.org/10.5194/egusphere-egu26-2352, 2026.
Hydrosat leverages land surface temperature measured by thermal infrared (TIR) satellite technology to help growers save water and increase yields. Key agronomical parameters, e.g., soil moisture, crop development, and crop water demand, are monitored daily over arbitrarily large areas by solving the surface energy balance. Coupled to a soil water balance model based on meteorological data, remote sensing algorithms also estimate the amount of irrigation water applied by farmers and generate irrigation recommendations optimizing water productivity, i.e., maximizing crop yield while minimizing irrigation water consumption.
IrriWatch is Hydrosat’s irrigation management decision support system, which allows growers to track water demand and growth progress of their crops down to individual 10x10 m² pixels, daily, in near-real-time. With governments becoming more conscious about conserving their water reserves, applying high-resolution remote sensing algorithms over large irrigation districts - and potentially even whole nations - is becoming increasingly relevant. Compared to small proof-of-concept models, this requires careful balancing of complex steps, including automated field delineation and crop identification at scale early in the growing season, data fusion and sharpening to obtain high-fidelity daily TIR data at a spatial resolution compatible with detecting in-field variations, as well as energy and water balance modelling capable to handle diverse environmental conditions and soil or crop types without any local data or farm management information available. To effectively help governments preserve water while increasing farmers’ crop yields, the immense amount of data generated by our models must be condensed to clear actionable indicators that are intuitive to an audience not necessarily familiar with remote sensing concepts.
We will present an overview of an operational processing pipeline to support both field-level precision agriculture applications and large-scale water productivity monitoring and optimization. Leveraging daily high-resolution land surface temperature, both from Hydrosat’s own satellite constellation and from a novel thermal sharpening algorithm, allows to track water productivity over tens of thousands of square kilometers. We find that high spatio-temporal resolution is critical to accurately monitor crop development even at regional or seasonal scale, as insufficient resolution introduces substantial errors in actual evapotranspiration estimates. In addition, correcting for geomorphological factors, e.g., microclimate or effect of elevation or slope on surface temperature, becomes increasingly important over large spatial scales.
Statistical analysis of field-scale results over large areas reveals spatial patterns of conditions responsible for yield losses or excessive water consumption. We will demonstrate how such insights support automatic identification of root causes for low water productivity, forming the basis for efficiently implementing data-driven mitigation actions.
How to cite: Werner, F., Ziliani, M., Chartrand, R., Hopkins, L., Pohankova, T., Stefan, V., Sleimi, R., Vinholi, J., Abello, A., and Bastiaanssen, W.: More crop per drop: precision irrigation and water productivity from field-scale to global scale, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20878, https://doi.org/10.5194/egusphere-egu26-20878, 2026.
Increasing climate variability and water scarcity are intensifying the need for adaptive, data-driven water management solutions in African agriculture. While digital and precision technologies are increasingly available, their effectiveness depends on how diverse data streams are integrated into usable decision support tools that respond to local conditions. This research focuses on the development of an open-source digital agriculture platform designed to support sustainable water management.
The developed platform brings together multiple sources of real-time and near-real-time data, including IoT-based soil moisture and climate sensors, weather forecasts, crop information, and remote sensing products, and systematically compares these observations with inputted crop water and irrigation models. By continuously comparing field-level data with input models, the decision support system aims to enable more accurate irrigation scheduling, early detection of water stress, and adaptive responses to climate variability, supporting more informed, timely, and context-specific decision-making for farmers. Artificial intelligence and machine learning components can be integrated to further enhance the platform by identifying patterns, improving forecasts, and refining model performance over time.
The presentation highlights the design and functionality of farmer-oriented decision support systems, outlining how open-source digital platforms can be tested, adapted, and refined in climate-vulnerable settings. By emphasizing interoperability, transparency, and community-driven innovation, the approach demonstrates how digital agriculture platforms can move beyond standalone technologies toward integrated decision support ecosystems for sustainable water management.
How to cite: Johnson, F., Cheo, A. E., Santillano Sanchez, F. A., Tambo, E., Djibo, R. S., Rabani, A., and Boubacar, I.: Leveraging Digital and Innovative Technologies for Agricultural Water Management , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-22159, https://doi.org/10.5194/egusphere-egu26-22159, 2026.
Reliable, field-scale indicators of crop water status and plant condition are needed to support plant breeding, precision irrigation and climate adaptation, yet UAV-based monitoring must balance predictive accuracy with deployability. We present an explainable Deep Learning workflow (TriNet) for scalable UAV phenotyping from multispectral time series, aligned with agronomic and breeding practice through high-granularity in situ scoring in accordance with established standards (such as those of the AGES - Österreichische Agentur für Gesundheit und Ernährungssicherheit). TriNet disentangles spatial, temporal, and spectral information and incorporates attention-based interpretability to identify influential inputs and guide efficient acquisition strategies. The framework supports handling multispectral data acquired from comparatively high altitudes with respect to the state of the art (e.g., 60 m with 2.5 cm Ground Sampling Distance (GSD)), and allows the exploration of the trade-off between model performance and GSD. This supports a reduction of flight times and data volumes (e.g., 1.74 GB at 60 m vs. 5.96 GB at 20 m in our reference setup) while maintaining controlled predictive accuracy.
We study the case of winter wheat breeding, and extend this approach with new results for the traits drought stress and plant health and a comprehensive analysis of flight height as an operational design variable, systematically simulating and evaluating acquisitions from 20 to 120 m. Results indicate that predictive accuracy is largely insensitive to flight height across this range, supporting higher-altitude, high-coverage monitoring until the release of larger datasets provide a clear justification for lower-altitude, higher-resolution acquisitions. Finally, we translate these findings into practitioner-oriented operational insights for drone-based High-Throughput Phenotyping.
How to cite: Salzinger, J., Beltrame, L., Schawerda, L.-T., and Fanta-Jende, P.: Addressing deployability concerns for AI-supported UAV-based High Throuput Phenotyping, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16806, https://doi.org/10.5194/egusphere-egu26-16806, 2026.
In a context marked by water scarcity as is the case in Morocco - particularly in semi-arid regions most exposed to water challenges such as the Souss Massa region - cultivation under cover exerts enormous pressure on water resources, making increasingly precise irrigation management essential. This work proposes a hybrid approach to improve the accuracy, generalization and stability of reference evapotranspiration predictions, integrating physical laws into the neural network architecture, which makes it possible to create a model that respects both the observed data and the physical knowledge governing reference evapotranspiration. The proposed methodology is based firstly on the evaluation of three deep learning architectures with advanced attention mechanism (Attention-based LSTM, Attention-based bidirectional-LSTM, Attention-based CNN-LSTM), secondly the evaluation of the best architecture before and after the integration of the physical component (Physics-Informed Neural Networks) using a convex combination integrating the Priestley-Taylor physical model. The results show the superiority of the hybrid architectures outperforming the others, the Attention-based CNN-LSTM architecture already obtaining interesting performances (R2 = 0.934).
However, the PINNs architecture with a balance coefficient set at λ = 0.1 outperforms all other architectures with less error and better data explanation (R² = 0.945). This combination allows a reduction of the average absolute error of 7.5% compared to the Attention-based CNN-LSTM model also ensuring better stability of predictions against extreme values. The validation is carried out in a prototype connected greenhouse equipped with IoT sensors and a monitoring dashboard.
This hybrid physico-learned approach offers a scalable and interpretable solution for intelligent irrigation management in semi-arid conditions.
How to cite: Zlaiga, J., Rghioui, A., Elhachemy, S., Elyaqouti, M., and Belaqziz, S.: A hybrid physics-artificial intelligence approach for accurate prediction of reference evapotranspiration, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13372, https://doi.org/10.5194/egusphere-egu26-13372, 2026.
The sustainability of global agricultural systems is increasingly dependent on the precision and efficiency of water distribution networks. In regions facing water scarcity, the design of irrigation subunits is a critical factor; however, the complexity of irregular field geometries often leads to designs based on manual approximations that ignore the full potential of hydraulic and economic optimization. This research introduces a sophisticated computational approach that integrates spatial network generation with advanced diameter optimization within a unified geographic information environment.
The core of this methodology lies in its ability to simultaneously address two fundamental aspects of irrigation engineering: the automated spatial layout of the pipe network and the discrete optimization of pipe diameters. By leveraging a high-precision hydraulic simulation engine, a genetic algorithm evaluates multiple potential configurations to identify the most cost-effective solution that satisfies pressure uniformity and flow requirements. This dual-integrated approach replaces traditional fragmented workflows, where layout design and hydraulic dimensioning are often performed in separate, disconnected steps.
The framework’s performance was validated through a practical application. This case study demonstrates how the system processes complex topographical data and irregular field boundaries to generate a complete infrastructure plan. The results indicate that the automated selection of commercial diameters, combined with an optimized spatial distribution of laterals and manifolds, leads to a significant reduction in total investment costs compared to conventional engineering methods.
By streamlining the transition from raw geospatial data to a fully optimized hydraulic network, this work provides a robust decision-support tool for precision agriculture. It offers a scalable and adaptable solution that enhances the efficiency of irrigation projects, supporting long-term water conservation goals and improving the economic viability of modern farming practices in the face of a changing climate.
How to cite: Campo-Bescós, M. Á., Barberena, I., and Casalí, J.: A GIS-Based Framework for the Spatial Design and Discrete Optimization of Drip Irrigation Subunits, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12062, https://doi.org/10.5194/egusphere-egu26-12062, 2026.
Mounting climate volatility, characterized by increasingly frequent and severe events, poses a critical threat to global food security. Traditional irrigation methods, which react only to visible drought symptoms, often fail to prevent irreversible physiological damage to crops. This underscores the need for precise, early detection of sub-lethal plant stress—a core challenge for precision agriculture. Effective early warning would enable proactive, smart irrigation, optimizing water use while protecting crop yields in a changing climate. Current drought assessment methods face significant trade-offs. Direct physiological measurements, though accurate, are destructive and impractical for field-scale use. Hyperspectral imaging (HSI) offers a non-destructive alternative by capturing detailed reflectance spectra. While it has identified signatures of advanced drought stress, a critical gap still remains, reliably predicting the initial metabolic perturbations that precede visible decline, particularly the early drop in net photosynthetic assimilation (An) which is a sensitive indicator of plant metabolic function and stress tolerance.
Our research directly addresses this need. Through controlled drought experiments of model plant Arabidopsis thaliana, we simultaneously collected high-resolution HSI data, transcriptome profiles, and ground-truth An measurements. A partial least squares regression model trained on spectral features accurately predicted An values two days in advance. Feature analysis identified wavelengths near 700 nm within the red-edge and near-infrared transition, as optimal early predictors. Strikingly, transcriptome data revealed a concurrent increase in gene activity linked to red and far-red light response pathways in drought-stressed plants. This convergence of spectral and molecular evidence indicates that early drought-induced photosynthetic alterations, predictive of An decline, manifest in canopy reflectance at ~700 nm and are underpinned by specific light-responsive molecular changes. By integrating hyperspectral phenotyping with mechanistic transcriptomics, we bridge prediction and biological causality, transforming HSI from a correlative tool into a mechanistically grounded early-warning system. This approach enables proactive, physiologically informed water management, paving the way for more climate-resilient agriculture.
How to cite: Cheng, K., Chen, C., Fan, K., Lam, H.-M., and Wu, J.: Drought-induced early alterations in photosynthetic efficiency revealed by convergence of spectral and molecular evidence, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6918, https://doi.org/10.5194/egusphere-egu26-6918, 2026.
Canopy cover (CC) reflects canopy density, leaf area development, and early stress conditions acts as a significant indicator for crop health. Accurate CC estimation helps in mapping spatial variability in crops and facilitates early detection of disease or stress due to nutrient or water deficiency. For estimation of canopy cover, UAV multispectral data was acquired at different crop growth stages. This study estimated wheat canopy cover percentage from tillering to dough stage using Random forest classifier and MSAVI index thresholding for more accurate and robust assessment of canopy dynamics. Supervised classification approach was used based on given training samples for three different classes, i.e., soil, canopy and shadow and classification was performed through Random forest (RF) algorithm. The extracted canopy pixels were then used for finding canopy cover percentage. Additionally, a simplified approach was used based on MSAVI index thresholds to identify crop pixels, enabling reliable CC estimation through vegetation index segmentation method. The experiment was conducted on wheat crop using three ETc (Crop evapotranspiration) based irrigation treatments i.e., 100%, 80%, and 60% ETc and each treatment had three replications. In addition to ETc based treatments, the Farmer’s and rainfed treatments were also considered. The rainfed treatments with two replications, received a single life-saving irrigation and farmer treatments, with three replications were irrigated based on local farmer’s practice.
Canopy cover percentage observed across different growth stages (40 to 114 DAS) showed distinct variation in crop development among varying irrigation treatments. In the treatments with 100%, 80%, and 60% ETc irrigation, RF based CC ranged from 35.3–98.5%, 36.1–97.9%, and 29.2–95.2%, while MSAVI-based CC ranged from 33.8–96.5%, 34.2–95.9%, and 28.1–94.5%, respectively. In comparison to ETc treatments, farmers treatment exhibited lower canopy cover, with ranges of 28.6–95.6% (RF) and 28.9–92.8% (MSAVI). Rainfed treatment recorded the lowest CC values across the growing season, varying between 23.1–72.4% using RF and 25.3–69.7% using MSAVI. Canopy cover estimates from the Random Forest algorithm and the MSAVI index showed consistent seasonal patterns, with RF generally producing slightly higher CC values. The NDVI patterns were also observed for all stages to validate these findings and the values ranged from 0.29–0.89, 0.26–0.88, and 0.24–0.85 in 100%, 80%, and 60% ETc treatments, respectively. Rainfed (0.22–0.74) and Farmer’s treatments (0.26–0.81) had lower NDVI values, supported the CC trends observed with RF and MSAVI methods. The highest CC and NDVI values were obtained around flowering stage i.e., (85-95) DAS and the lowest at tillering stages for all treatments, followed by a gradual decline after the flowering stage as the crop progressed toward maturity. Canopy cover trends were comparable in the 100% and 80% ETc treatments, whereas CC in 60% ETc treatment remained lower at all stages, indicating the impact of water deficit on canopy growth.
The study highlights that MSAVI based vegetation-index methods can provide a reliable and highly efficient pathway for estimating canopy cover, reducing the need for extensive training datasets and complex classification models.
How to cite: Yadav, A., Upreti, H., and Singhal, G. D.: Assessment of UAV based Canopy Cover for Varying Irrigation Treatments using Random Forest Classifier and MSAVI Index Thresholding, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-773, https://doi.org/10.5194/egusphere-egu26-773, 2026.
Posters on site: Mon, 4 May, 08:30–10:15 | Hall A
Quantification of crop evapotranspiration (ET) and yield is essential for precision agricultural water management and food security, particularly over long temporal and large regional scales. In this study, we combined a water-carbon coupled model with a GPP-driven crop growth simulation method utilizing remote sensing datasets, to simultaneously estimate ET and yield over the past two decades in the North China Plain. The developed model was tested for two major crops (winter wheat and summer maize) using approximately 20 site-years of observations. For wheat, the root mean square error (RMSE) values of ET and gross primary production (GPP) were 0.57 mm d-1 and 1.65 gC m-2 d-1, and for maize were 0.80 mm d-1 and 2.92 gC m-2 d-1, respectively. Besides, the crop growth simulation agreed well with measurements that R2 values were mostly larger than 0.66, and the RMSE of yield was 554.7 for wheat and 1346.6 kg hm-2 for maize, respectively. The results revealed an increasing trend in the crop water productivity (WP = yield/ET) of wheat, while maize maintained an overall higher WP than wheat during 2001-2018. In addition, the impacts of climate change and human management on the spatiotemporal dynamics of ET-GPP fluxes over the agroecosystems were evaluated. The significantly increased GPP rather than ET dominated the significant increase in water use efficiency (WUE=GPP/ET) in the NCP, accounting for 38.6% of its cropland area. The temporal dynamic of regional mean WUE indicated a significantly increased rate of 0.026 gC kg-1H2O per year during 2001-2018. The experimental simulations demonstrated that agricultural management dominated the interannual trend of WUE, with a relative contribution of 79.5%, which was obviously larger than that of atmospheric CO2 concentration (40.2%) and changes in climate variables (-19.7%). The effects of agricultural management on WUE were further disaggregated across the classified six cropping systems, and 82.4% could be attributed to the management of winter wheat-summer maize rotation system. The remote sensing-based model developed in this study effectively quantifies regional ET and yield for two typical crops, providing critical information for smart agricultural water management. The analysis of agroecosystem WUE under changing environments underscores the dominant role of agricultural management and offers insights for climate adaption in agriculture.
How to cite: Wang, X., Lei, H., and Huo, Z.: Coupled estimation of crop evapotranspiration-yield and assessment of water use efficiency in the North China Plain through a remote sensing-based model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3713, https://doi.org/10.5194/egusphere-egu26-3713, 2026.
Climate change will heavily impact agriculture through alterations of precipitation dynamics which leads to more frequent agroecological droughts and intense precipitation events. Strategies to adapt to these changes are necessary to maintain food safety and sustain livelihoods in the agricultural sector. One method for farmers to mitigate the impacts from climate change is the Keylines design which can be described as open ditches parallel to the elevation line. These are designed to retain runoff, reduce erosion and increase infiltration which should lead to a higher amount of water available to plants during dry periods (e.g., Ponce-Rodríguez et al. 2021). However, scientific research and corresponding data regarding Keyline systems and their influence on field hydrological dynamics is sparse.
To quantify the hydrological impact of Keyline systems, a comprehensive field experiment has been set up combining Keyline systems with agroforest on two agricultural fields, one of which is located in the eastern Jura-range and one outside of Zurich. The goal of this study is to assess the optimal integration of tools to investigate how the soil moisture patterns are altered by Keyline systems and quantify the timing and amount of water retained. The work presented here shows the first results of a comparison between different soil moisture analysis methods applied to agricultural fields, including (a) soil hydrological modelling, (b) electric resistivity tomography, (c) UAV-based L-band radiometry, (d) in-situ soil matrix potential and volumetric water content measurements and (e) destructive gravimetric water content measurements. Several of these methods currently undergo rapid developments due to the technological advancements made in recent years, leading to an increased accessibility for a broader range of users (e.g. Du 2020; Zhou et al. 2025). This highlights the need to assess the tools regularly to showcase possible applications and directions for further development. The results presented here demonstrate the capabilities as well as the limitations of the individual methods and shows how the different systems can be used complementary with each other to obtain a complete assessment of the soil hydrological dynamics. This will help researchers investigating soil moisture dynamics to make informed choices regarding their research tools for the assessment of nature-based solutions to adapt to climate change impacts within but also beyond agriculture.
Literature:
Du, C. (2020). Comparison of the performance of 22 models describing soil water retention curves from saturation to oven dryness. Vadose Zone Journal, 19(1), e20072. https://doi.org/10.1002/vzj2.20072.
Ponce-Rodríguez, M. D. C., Carrete-Carreón, F. O., Núñez-Fernández, G. A., Muñoz-Ramos, J. de J., & Pérez-López, M. E. (2021): Keyline in bean crop (Phaseolus vulgaris l.) for soil and water conservation. Sustainability, 13(17), 9982. https://doi.org/10.3390/su13179982.
Zhou, Y., Schwank, M., Boutin, J., Richaume, P., Mialon, A., Holmberg, M., Kalescke, L. Zeiger, P., Leduc-Leballeur, M., ... , Kerr, Y. (2025, in review): Setellite Microwave Radiometry at L-band for Monitoring Earth’s Essential Climate Variables. IEEE Geoscience and Remote Sensing Magazine.
How to cite: Villiger, M. N., Leuteriz, A., Carminati, A., and Stähli, M.: Towards an Optimal Method for Assessing the Spatial and Temporal Hydrological Dynamics of a Keyline System, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2540, https://doi.org/10.5194/egusphere-egu26-2540, 2026.
Root-zone soil moisture (RZSM) reflects the water accessible to plants and is therefore central to precision irrigation support and agricultural drought monitoring, yet direct RZSM observations are limited. Satellite missions provide surface soil moisture (SSM), but they do not directly observe deeper layers, and in-situ measurements remain too sparse for broad coverage. We present a machine learning approach to estimate daily RZSM in vineyards in the Terra Alta region of Catalonia in northeastern Spain, using daily 2020 to 2024 in-situ observations from eight stations as reference data. This model provides a baseline for later experiments using satellite SSM to extend applicability beyond the instrumented network.
We train a multilayer perceptron (MLP) to predict soil moisture at 25 cm, taken as RZSM, using in-situ SSM at 5 cm, daily precipitation, mean, minimum and maximum temperature, a cyclic encoding of day of year, and static soil descriptors from SoilGrids. Robustness is assessed with year-block cross-validation to evaluate temporal generalisation and leave-station-out experiments to evaluate transferability across vineyards. Performance is quantified using non-parametric Kling–Gupta efficiency (KGE) and RMSE.
The model achieves strong skill when evaluated on independent years at training stations, with median KGE around 0.9. Transfer to unseen vineyards is more heterogeneous, with some stations retaining good performance around 0.85 and others showing biases and reduced efficiency, suggesting that additional information may be needed for consistent transfer across vineyards. Ongoing work aims to improve generalisation by incorporating antecedent moisture and precipitation information and by testing additional predictors such as vegetation, supported by feature importance analysis across the full set of inputs. To enable use beyond the instrumented network, we will transition the model towards configurations driven by or trained with satellite-derived SSM. Taken together, these steps are intended to move towards a transferable tool to support drought monitoring and irrigation-related decisions in agricultural regions.
How to cite: Cid-Giménez, J., Escorihuela, M. J., Barella-Ortiz, A., and Quintana-Seguí, P.: Estimating root-zone soil moisture in Mediterranean vineyards using machine learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21620, https://doi.org/10.5194/egusphere-egu26-21620, 2026.
Spectral sensors have become an integral part of modern precision agriculture. It enables fast, non-destructive and map-based crop monitoring of key crop physiological parameters. The spectral resolution of a sensor plays an important role in determining its ability to detect subtle changes in the crop, nutrient status and canopy development. Sensor’s comparison based on spectral resolution remains limited particularly in the context of field-level agronomic monitoring. This study aims to address this gap by using three sensors UAV-based multispectral (MS), UAV-based hyperspectral (HS) and handheld Greenseeker (GS) NDVI measurements. Hyperspectral sensors provide continuous high-resolution data from visible to near infrared; MS use fewer broad bands; GS limits to two bands for quick NDVI field checks. The experimental study was conducted in the arid region of Uttar Pradesh, India. The experimental setup consisted of plots with same irrigation (100% ETc) and varying nitrogen dosage i.e. 150,120 and 90 kg/ha (Plot 1, Plot 2 and Plot 3, respectively) with three replications. Plots 4 and 5, representing farmer-field conditions with 120 kg ha⁻¹ nitrogen and no nitrogen respectively, followed regional irrigation practices, whereas Plot 6 (rainfed) was irrigated only once initially. A series of UAV-flights were conducted across critical phenological stages, and the reflectance was used to generate Normalized Difference Vegetation Index (NDVI) representing canopy density.
The results showed that NDVI rapidly increased during early vegetative stage (61-75) DAS, saturated around (75-85) DAS, followed by a decline during (101-117) DAS. NDVI peaked around flowering stage for all the sensors. GS-NDVI varied between (0.46-0.78), MS-NDVI displayed (0.52-0.86), whereas HS- NDVI varied between (0.55-0.90). The mean NDVI values were (0.570 ± 0.085) for GS, (0.608 ± 0.075) for MS, and (0.664 ± 0.087) for the HS, with HS exceeding others by 16.5% (vs. GS) and 9.3% (vs. MS). Pearson correlation coefficients confirm strong inter-sensor agreement: Greenseeker-Hyperspectral r = 0.96, Multispectral-Hyperspectral r = 0.91, Greenseeker-Multispectral r = 0.87 (all p<0.001), indicating consistent vegetation health trends despite spectral resolution variances. Across days 61-117, fully irrigated plots with varying nitrogen dosage (Plot 1-3) maintained higher vegetation indices (0.60-0.90) than stressed plots. For Plot 4 (0.57-0.84), Plot 5(0.49-0.79) and Plot 6(0.46-0.76), the decline accelerated under water and nitrogen deficit. Water-stressed and nitrogen deficit plots show greater NDVI drops, indicating higher stress levels leading to early senescence, thus affecting the grain yield.
Overall, the three sensors show strong agreement in NDVI trends. For precision agriculture, HS optimized subtle changes, followed by MS; statistical trends aligned with established NDVI comparison protocols using correlation and regression. Hyperspectral sensor offered the highest diagnostic capability, multispectral provided spatial characteristics and greenseeker served as an efficient tool for rapid monitoring of field. These combined observations emphasize the importance of selecting sensors based on the required level of detail, operational constraints, and monitoring objectives in precision agriculture. Integrating data from multiple sensor types can further enhance crop assessment accuracy and support more informed decision-making in precision agriculture.
How to cite: Adwait, A., Upreti, H., and Singhal, G. D.: Evaluating Spectral Resolution Effects on Crop Monitoring: A Comparison of UAV-based Multispectral, Hyperspectral and handheld Greenseeker sensor, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-743, https://doi.org/10.5194/egusphere-egu26-743, 2026.
Posters virtual: Thu, 7 May, 14:00–18:00 | vPoster spot A
EGU26-2155 | ECS | Posters virtual | VPS10
A combined approach of UAV data and machine learning algorithms in weeds detectionThu, 07 May, 14:45–14:48 (CEST) vPoster spot A
The Souss-Massa region is known as the most important agricultural area in Morocco, and one of the most affected regions by climate change and over-exploitation. This situation has required the intervention of new tools to improve water resource management. In this context, the Unmanned Aerial Vehicles (UAVs) images data were used for weeds detection in a Citrus orchard farm. Two sites were considered, the first one planted with 12-years-old and 1.5 years-old clementine trees. After a panoply of image processing from the data collection, following by the georeferencing, the creation of the digital elevation model, the digital surface model, and the elaboration of the orthomosaic image, the machine learning algorithms (MLA) such as Maximum Likelihood Classification, Minimum Distance Classification, Support Vector Machine, were applied for weeds detection and mapping. For both sites, all MLA showed a Cohen’s kappa coefficient higher than 0.6 and an overall accuracy higher than 60%. This study demonstrates how this emerging technology offers farmers opportunities to enhance production while optimizing water usage.
How to cite: El Hafyani, M., Chaaou, A., Sadik, A., Labbaci, A., Hssaisoune, M., Tairi, A., Abdelfadel, F., Taia, S., Ait-Ichou, H., Elhaid, I., and Bouchaou, L.: A combined approach of UAV data and machine learning algorithms in weeds detection , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2155, https://doi.org/10.5194/egusphere-egu26-2155, 2026.
EGU26-2054 | ECS | Posters virtual | VPS10
Real-Time UAV-Deep Learning System for Citrus Orchard Structure and Yield AssessmentThu, 07 May, 14:57–15:00 (CEST) vPoster spot A
Arid and semi-arid regions are facing more frequent and severe droughts, with annual rainfall often below 200 mm. Large-scale, intensive irrigation further strains these limited water resources. Under these conditions, growers need practical tools to estimate yield and monitor tree health at high spatial detail so they can better manage irrigation and inputs. This work develops and tests an automated, data-driven pipeline for estimating citrus yield at the individual-tree level using UAV imagery and Deep Learning. The pipeline comprises three main components. First, individual trees and orchard rows are segmented using a lightweight Tiny U‑Net model. Second, a CNN-based model predicts tree-level yield from vegetation indices and field measurements. Third, these predictions are validated through detailed fruit sampling.
The study was conducted in a commercial citrus orchard in a semi-arid region under climate and water stress. High‑resolution UAV imagery was processed into orthomosaics and vegetation index maps, and the Tiny U‑Net was optimized for fast, near real‑time semantic segmentation, enabling precise tree crown delineation and accurate tree and row counts. For yield prediction, the CNN model exploited spatial features from vegetation indices combined with in‑situ data. The validation relied on direct comparison between UAV‑based yield estimates and yields obtained from field sampling and laboratory weighing. Both mean and median yields per tree were computed to capture tree‑level variability. The final dataset, consisting of 34 trees and approximately 340 fruit samples, provided a robust basis for assessing model performance. The Tiny U‑Net segmentation model reached high accuracy, with precision and recall of 94.74% and 94.88%, and an inference time of 12.55 ms per image tile. This shows the model is suitable for real‑time or on‑board use and can reliably map orchard structure at large scale. Tree and row counts derived from the segmentation achieved an R² greater than 0.99, confirming the robustness of the approach. For yield estimation, the CNN model outperformed other machine learning methods, achieving an R² of 0.88 at tree level. Field validation confirmed the practical usefulness of the pipeline, UAV‑predicted yields closely matched ground‑truth values, with both indicating an average yield of roughly 50 kg per tree. Most trees fell between 40 and 70 kg, and the model’s output histogram mean 50.9 kg, and median 51.4 kg aligned well with these field observations.
This robust agreement between model outputs and independent field validation data underscores the system's reliability and operational readiness for accurate, tree-level yield mapping. By integrating precise tree segmentation, high-resolution vegetation indices, and rigorously collected ground truth measurements, this study demonstrates that automated yield maps can be produced with sufficient accuracy to support operational decisions in orchards. This offers a cost-effective and scalable tool for precision agriculture, enabling optimized resource allocation, improved harvest planning, and adaptive management under increasing climate stress.
How to cite: Bakas, K., Saddik, A., Dliou, A., Hssaisoune, M., El Hachemy, S., Ait Ichou, H., Hmache, F., El Hafyani, M., Labbaci, A., and Bouchaou, L.: Real-Time UAV-Deep Learning System for Citrus Orchard Structure and Yield Assessment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2054, https://doi.org/10.5194/egusphere-egu26-2054, 2026.
