TY - GEN A1 - Mengsuwan, Konlavach A1 - Rivera-Palacio, Juan C. A1 - Ryo, Masahiro T1 - ChatGPT and general-purpose AI count fruits in pictures surprisingly well without programming or training T2 - Smart Agricultural Technology N2 - General-purpose artificial intelligence (AI) can facilitate agricultural digitalization as many tools do not require coding. Yet, it remains unclear how well the emerging general-purpose AI technologies can perform object counting, which is a fundamental task in agricultural digitalization, in comparison to the current standard practice. We show that ChatGPT (GPT4 V) demonstrated moderate performance in counting coffee cherries from images, while the T-Rex, foundation model for object counting, performed with high accuracy. Testing with a hundred images, we examined that ChatGPT can count cherries, and the performance improves with human feedback (R2 = 0.36 and 0.46, respectively). The T-Rex foundation model required only a few samples for training but outperformed YOLOv8, the conventional best practice model (R2 = 0.92 and 0.90, respectively). Obtaining the results with these models was 100x shorter than the conventional best practice. These results bring two surprises for deep learning users in applied domains: a foundation model can drastically save effort and achieve higher accuracy than a conventional approach, and ChatGPT can reveal a relatively good performance especially with guidance by providing some examples and feedback. No requirement for coding skills can impact education, outreach, and real-world implementation of generative AI for supporting farmers. KW - Foundation model KW - General purpose ai KW - Chatgpt KW - Large language model KW - Large vision language model KW - Agriculture Y1 - 2024 U6 - https://doi.org/10.1016/j.atech.2024.100688 SN - 2772-3755 VL - 9 PB - Elsevier BV ER - TY - GEN A1 - Ryo, Masahiro A1 - Schiller, Josepha A1 - Stiller, Stefan A1 - Palacio, Juan Camilo Rivera A1 - Mengsuwan, Konlavach A1 - Safonova, Anastasiia A1 - Wei, Yuqi T1 - Deep learning for sustainable agriculture needs ecology and human involvement T2 - Journal of Sustainable Agriculture and Environment N2 - Deep learning is an emerging data analytic tool that can improve predictability, efficiency and sustainability in agriculture. With a bibliometric analysis of 156 articles, we show how deep learning methods have been applied in the context of sustainable agriculture. As a general publication trend, China and India are leading countries for publication, international collaboration is still minor. Deep learning has been popularly applied in the context of smart agriculture across scales for individual plant monitoring, field monitoring, field operation and robotics, predicting soil, water and climate conditions and landscape‐level monitoring of land use and crop types. We identified that the potential of deep learning had been investigated mainly for predicting soil (abiotic), water, climate and vegetation dynamics, but ecological characteristics are critically understudied. We also highlight key themes that can be better addressed with deep learning for fostering sustainable agriculture: (i) including above‐ and belowground ecological dynamics such as ecosystem functioning and ecotone, (ii) evaluating agricultural impacts on other ecosystems and (iii) incorporating the knowledge and opinions of domain experts and stakeholders into artificial intelligence. We propose that deep learning needs to go beyond automatic data analysis by integrating ecological and human knowledge to foster sustainable agriculture. KW - biodiversity KW - deep learning KW - ecology KW - smart agriculture KW - sustainable agriculture Y1 - 2022 U6 - https://doi.org/10.1002/sae2.12036 SN - 2767-035X VL - 2 IS - 1 SP - 40 EP - 44 PB - Wiley ER - TY - GEN A1 - Synodinos, Alexis D. A1 - Karnatak, Rajat A1 - Aguilar‐Trigueros, Carlos A. A1 - Gras, Pierre A1 - Heger, Tina A1 - Ionescu, Danny A1 - Maaß, Stefanie A1 - Musseau, Camille L. A1 - Onandia, Gabriela A1 - Planillo, Aimara A1 - Weiss, Lina A1 - Wollrab, Sabine A1 - Ryo, Masahiro T1 - The rate of environmental change as an important driver across scales in ecology T2 - Oikos N2 - Global change has been predominantly studied from the prism of ‘how much' rather than ‘how fast' change occurs. Associated to this, there has been a focus on environmental drivers crossing a critical value and causing so‐called regime shifts. This presupposes that the rate at which environmental conditions change is slow enough to allow the ecological entity to remain close to a stable attractor (e.g. an equilibrium). However, environmental change is occurring at unprecedented rates. Equivalently to the classical regime shifts, theory shows that a critical threshold in rates of change can exist, which can cause rate‐induced tipping (R‐tipping). However, the potential implications of R‐tipping in ecology remain understudied. We aim to facilitate the application of R‐tipping theory in ecology with the objective of identifying which properties (e.g. level of organisation) increase susceptibility to rates of change. First, we clarify the fundamental difference between tipping caused by the magnitude as opposed to the rate of change crossing a threshold. Then we present examples of R‐tipping from the ecological literature and seek the ecological properties related to higher sensitivity to rates of change. Specifically, we consider the role of the level of ecological organisation, spatial processes, eco‐evolutionary dynamics and pair–wise interactions in mediating or buffering rate‐induced transitions. Finally, we discuss how targeted experiments can investigate the mechanisms associated to increasing rates of change. Ultimately, we seek to highlight the need to better understand how rates of environmental change may induce ecological responses and to facilitate the systematic study of rates of environmental change in the context of current global change. KW - climate change KW - ecological communities KW - eco-evo feedbacks KW - transitions KW - global change KW - R-tipping KW - temporal ecology Y1 - 2023 U6 - https://doi.org/10.1111/oik.09616 SN - 0030-1299 VL - 2023 IS - 4 PB - Wiley ER - TY - GEN A1 - Edlinger, Anna A1 - Garland, Gina A1 - Banerjee, Samiran A1 - Degrune, Florine A1 - García‐Palacios, Pablo A1 - Herzog, Chantal A1 - Pescador, David Sánchez A1 - Romdhane, Sana A1 - Ryo, Masahiro A1 - Saghaï, Aurélien A1 - Hallin, Sara A1 - Maestre, Fernando T. A1 - Philippot, Laurent A1 - Rillig, Matthias C. A1 - Heijden, Marcel G. A. van der T1 - The impact of agricultural management on soil aggregation and carbon storage is regulated by climatic thresholds across a 3000 km European gradient T2 - Global Change Biology N2 - Organic carbon and aggregate stability are key features of soil quality and are important to consider when evaluating the potential of agricultural soils as carbon sinks. However, we lack a comprehensive understanding of how soil organic carbon (SOC) and aggregate stability respond to agricultural management across wide environmental gradients. Here, we assessed the impact of climatic factors, soil properties and agricultural management (including land use, crop cover, crop diversity, organic fertilization, and management intensity) on SOC and the mean weight diameter of soil aggregates, commonly used as an indicator for soil aggregate stability, across a 3000 km European gradient. Soil aggregate stability (−56%) and SOC stocks (−35%) in the topsoil (20 cm) were lower in croplands compared with neighboring grassland sites (uncropped sites with perennial vegetation and little or no external inputs). Land use and aridity were strong drivers of soil aggregation explaining 33% and 20% of the variation, respectively. SOC stocks were best explained by calcium content (20% of explained variation) followed by aridity (15%) and mean annual temperature (10%). We also found a threshold‐like pattern for SOC stocks and aggregate stability in response to aridity, with lower values at sites with higher aridity. The impact of crop management on aggregate stability and SOC stocks appeared to be regulated by these thresholds, with more pronounced positive effects of crop diversity and more severe negative effects of crop management intensity in nondryland compared with dryland regions. We link the higher sensitivity of SOC stocks and aggregate stability in nondryland regions to a higher climatic potential for aggregate‐mediated SOC stabilization. The presented findings are relevant for improving predictions of management effects on soil structure and C storage and highlight the need for site‐specific agri‐environmental policies to improve soil quality and C sequestration. KW - aggregate stability KW - agro-ecosystems KW - aridity KW - climatic threshold KW - environmental gradient KW - intensive agriculture KW - soil organic carbon Y1 - 2023 U6 - https://doi.org/10.1111/gcb.16677 SN - 1354-1013 VL - 29 IS - 11 SP - 3177 EP - 3192 PB - Wiley ER - TY - GEN A1 - Riva, Federico A1 - Graco-Roza, Caio A1 - Daskalova, Gergana N. A1 - Hudgins, Emma J. A1 - Lewthwaite, Jayme M. M. A1 - Newman, Erica A. A1 - Ryo, Masahiro A1 - Mammola, Stefano T1 - Toward a cohesive understanding of ecological complexity T2 - Science Advances N2 - Ecological systems are quintessentially complex systems. Understanding and being able to predict phenomena typical of complex systems is, therefore, critical to progress in ecology and conservation amidst escalating global environmental change. However, myriad definitions of complexity and excessive reliance on conventional scientific approaches hamper conceptual advances and synthesis. Ecological complexity may be better understood by following the solid theoretical basis of complex system science (CSS). We review features of ecological systems described within CSS and conduct bibliometric and text mining analyses to characterize articles that refer to ecological complexity. Our analyses demonstrate that the study of complexity in ecology is a highly heterogeneous, global endeavor that is only weakly related to CSS. Current research trends are typically organized around basic theory, scaling, and macroecology. We leverage our review and the generalities identified in our analyses to suggest a more coherent and cohesive way forward in the study of complexity in ecology. Y1 - 2023 U6 - https://doi.org/10.1126/sciadv.abq4207 SN - 2375-2548 VL - 9 IS - 25 PB - American Association for the Advancement of Science (AAAS) ER - TY - GEN A1 - Wei, Yuqi A1 - Wei, Bin A1 - Ryo, Masahiro A1 - Bi, Yixian A1 - Sun, Xiangyun A1 - Zhang, Yingjun A1 - Liu, Nan T1 - Grazing facilitates litter-derived soil organic carbon formation in grasslands by fostering microbial involvement through microenvironment modification T2 - CATENA N2 - Grasslands store 10–30 % of the global soil organic carbon (SOC) and have the potential to mitigate the increase in atmospheric CO2 concentrations. Grazing plays a crucial role in regulating SOC storage in grassland ecosystems. However, the mechanistic understanding of how grazing influences the SOC dynamic still needs to be improved. We investigated how grazing-induced microenvironment changes influence the microbial assimilation of plant litter C and SOC formation from decomposed litter C in a multi-year field experiment, where grazing was simulated with mowing, dung and urine return, and trampling. We incubated 13C labeled litter in PVC collars to trace the microbial assimilation of litter C and the fate of litter C in the SOC after decomposition. While the grazing treatments changed soil properties marginally, mowing decreased above-ground plant biomass, litter mass, plant height, and plant cover (−12 % to −79 %). Accordingly, mowing treatment increased the exposure of litter to UV radiation (+38 %) and therefore facilitated the microbial assimilation of litter C (+20 %) and the SOC formation (+15 %). Trampling treatment promoted the transformation of litter C to SOC pools by mixing litter and soil (+34 %). Dung and urea return treatment did not affect SOC formation due to a marginal change in available nitrogen. Collectively, our results suggest that grazing facilitates litter-derived SOC formation by regulating microbial involvement through changes in the microenvironment. Our study indicates that grazing promotes SOC formation from plant litter, which maintains SOC storage in grasslands. Accurate quantification of the contribution of plant C input to SOC pools in different grasslands under various utilization is the next step to better predict SOC dynamics. KW - 13C-stable isotope KW - Grazing KW - Microenvironment KW - PLFA-SIP KW - Soil microbial assimilation KW - Soil organic carbon (SOC) formation Y1 - 2023 U6 - https://doi.org/10.1016/j.catena.2023.107389 SN - 0341-8162 VL - 232 PB - Elsevier BV ER - TY - GEN A1 - Safonova, Anastasiia A1 - Ghazaryan, Gohar A1 - Stiller, Stefan A1 - Main-Knorn, Magdalena A1 - Nendel, Claas A1 - Ryo, Masahiro T1 - Ten deep learning techniques to address small data problems with remote sensing T2 - International Journal of Applied Earth Observation and Geoinformation N2 - Researchers and engineers have increasingly used Deep Learning (DL) for a variety of Remote Sensing (RS) tasks. However, data from local observations or via ground truth is often quite limited for training DL models, especially when these models represent key socio-environmental problems, such as the monitoring of extreme, destructive climate events, biodiversity, and sudden changes in ecosystem states. Such cases, also known as small data problems, pose significant methodological challenges. This review summarises these challenges in the RS domain and the possibility of using emerging DL techniques to overcome them. We show that the small data problem is a common challenge across disciplines and scales that results in poor model generalisability and transferability. We then introduce an overview of ten promising DL techniques: transfer learning, self-supervised learning, semi-supervised learning, few-shot learning, zero-shot learning, active learning, weakly supervised learning, multitask learning, process-aware learning, and ensemble learning; we also include a validation technique known as spatial k-fold cross validation. Our particular contribution was to develop a flowchart that helps DL users select which technique to use given by answering a few questions. We hope that our review article facilitate DL applications to tackle societally important environmental problems with limited reference data. KW - Small data problems KW - Remote sensing KW - Deep learning KW - Transfer learning KW - Few-shot learning KW - Zero-shot learning KW - Self-supervised learning Y1 - 2023 U6 - https://doi.org/10.1016/j.jag.2023.103569 SN - 1569-8432 VL - 125 PB - Elsevier BV ER - TY - GEN A1 - Stiller, Stefan A1 - Dueñas, Juan F A1 - Hempel, Stefan A1 - Rillig, Matthias C. A1 - Ryo, Masahiro T1 - Deep learning image analysis for filamentous fungi taxonomic classification: dealing with small datasets with class imbalance and hierarchical grouping T2 - Biology Methods and Protocols N2 - Deep learning applications in taxonomic classification for animals and plants from images have become popular, while those for microorganisms are still lagging behind. Our study investigated the potential of deep learning for the taxonomic classification of hundreds of filamentous fungi from colony images, which is typically a task that requires specialized knowledge. We isolated soil fungi, annotated their taxonomy using standard molecular barcode techniques, and took images of the fungal colonies grown in petri dishes (n = 606). We applied a convolutional neural network with multiple training approaches and model architectures to deal with some common issues in ecological datasets: small amounts of data, class imbalance, and hierarchically structured grouping. Model performance was overall low, mainly due to the relatively small dataset, class imbalance, and the high morphological plasticity exhibited by fungal colonies. However, our approach indicates that morphological features like color, patchiness, and colony extension rate could be used for the recognition of fungal colonies at higher taxonomic ranks (i.e. phylum, class, and order). Model explanation implies that image recognition characters appear at different positions within the colony (e.g. outer or inner hyphae) depending on the taxonomic resolution. Our study suggests the potential of deep learning applications for a better understanding of the taxonomy and ecology of filamentous fungi amenable to axenic culturing. Meanwhile, our study also highlights some technical challenges in deep learning image analysis in ecology, highlighting that the domain of applicability of these methods needs to be carefully considered. KW - Metagenomics and Environmental Biology KW - Computational Methods KW - Imaging technologies Y1 - 2024 U6 - https://doi.org/10.1093/biomethods/bpae063 SN - 2396-8923 VL - 9 IS - 1 PB - Oxford University Press (OUP) ER - TY - GEN A1 - Ryo, Masahiro T1 - Ecology with artificial intelligence and machine learning in Asia: a historical perspective and emerging trends T2 - Ecological Research N2 - The use of artificial intelligence (AI) and machine learning (ML) has significantly enhanced ecological research in Asia by improving data processing, analysis, and pattern extraction. Analyzing 1550 articles, I show an overview of the use of AI and ML for Asian ecological research. Following the last 20 year trend, I found that the topics in Asian ecological research have transitioned from technical perspectives to more applied issues, focusing on biodiversity conservation, climate change, land use change, and societal impacts. Non‐Asian countries, on the other hand, have focused more on theoretical understanding and ecological processes. The difference between Asian and non‐Asian regions may have emerged due to the ecological challenges faced by Asian countries, such as rapid economic growth, land development, and climate change impacts. In both regions, deep learning related technology has been emerging (e.g., big data collection including image and movement). Within Asia, China has been the Asia‐leading country for AI/ML applications followed by Korea, Japan, India, and Iran. The number of computer science education programs in China has been increasing 3.5× times faster than that in the United States, indicating that a nationwide strategy for computer science development is key for ecological science with AI. Overall, the adoption of AI and ML technologies in ecological studies in Asia has propelled the field forward and opened new avenues for innovative research and conservation practices. KW - artificial intelligence (AI) KW - biodiversity conservation KW - China KW - deep learning KW - machine learning (ML) Y1 - 2023 U6 - https://doi.org/10.1111/1440-1703.12425 SN - 0912-3814 VL - 39 IS - 1 SP - 5 EP - 14 PB - Wiley ER - TY - GEN A1 - Stiller, Stefan A1 - Grahmann, Kathrin A1 - Ghazaryan, Gohar A1 - Ryo, Masahiro T1 - Improving spatial transferability of deep learning models for small-field crop yield prediction T2 - ISPRS Open Journal of Photogrammetry and Remote Sensing N2 - Predicting crop yield using deep learning (DL) and remote sensing is a promising technique in agriculture. In smallholder agriculture (<2 ha), where 84% of the farms operate globally, it is crucial to build a model that can be useful across several fields (high spatial transferability). However, enhancing spatial model transferability in a small-scale setting faces significant challenges, including spatial autocorrelation, heterogeneity and scale dependence of spatial dynamics, as well as the need to address limited data points. This study aimed to test the hypothesis that spatial cross validation (SCV) is a more suitable model validation practice than random cross validation (RCV) to enhance model transferability for spatial prediction in a small-scale farming setting. We compared the performances of DL models that predict crop yield for several settings including three crop types and two DL architectures based on RCV with and without overlapping samples and SCV. Notably, we conducted model performance tests on external, equally sized fields instead of the field used for training. We used high resolution RGB imagery taken with a drone as input. Our results show that the models using SCV outperformed those using RCV when the models were tested on external fields (on average r = 0.37 for SCV, r = 0.18 for RCV with overlap and r = 0.07 without), even though the models using SCV showed a substantially lower performance for cross validation (CV) than those using RCV (r with SCV and RCV w/o overlap = 0.73 and 0.98/0.73, respectively). The results suggest that RCV leads to over-optimism by overfitting the spatial structure and remembering image-specific information (so called memorization). Our study offers the first empirical evidence in agriculture that SCV is preferable to RCV in small field settings for making DL models more transferable. KW - Smallholder farming KW - UAV KW - Crop yield KW - Convolutional neural network KW - Spatial cross validation KW - External validation Y1 - 2024 U6 - https://doi.org/10.1016/j.ophoto.2024.100064 SN - 2667-3932 VL - 12 PB - Elsevier BV ER - TY - GEN A1 - Palacio, Juan Camilo Rivera A1 - Bunn, Christian A1 - Rahn, Eric A1 - Little-Savage, Daisy A1 - Schmidt, Paul Günter A1 - Ryo, Masahiro T1 - Geographic-scale coffee cherry counting with smartphones and deep learning T2 - Plant Phenomics N2 - Deep learning and computer vision, using remote sensing and drones, are 2 promising nondestructive methods for plant monitoring and phenotyping. However, their applications are infeasible for many crop systems under tree canopies, such as coffee crops, making it challenging to perform plant monitoring and phenotyping at a large spatial scale at a low cost. This study aims to develop a geographic-scale monitoring method for coffee cherry counting, supported by an artificial intelligence (AI)-powered citizen science approach. The approach uses basic smartphones to take a few pictures of coffee trees; 2,968 trees were investigated with 8,904 pictures in Junín and Piura (Peru), Cauca, and Quindío (Colombia) in 2022, with the help of nearly 1,000 smallholder coffee farmers. Then, we trained and validated YOLO (You Only Look Once) v8 for detecting cherries in the dataset in Peru. An average number of cherries per picture was multiplied by the number of branches to estimate the total number of cherries per tree. The model's performance in Peru showed an R2 of 0.59. When the model was tested in Colombia, where different varieties are grown in different biogeoclimatic conditions, the model showed an R2 of 0.71. The overall performance in both countries reached an R2 of 0.72. The results suggest that the method can be applied to much broader scales and is transferable to other varieties, countries, and regions. To our knowledge, this is the first AI-powered method for counting coffee cherries and has the potential for a geographic-scale, multiyear, photo-based phenotypic monitoring for coffee crops in low-income countries worldwide. Y1 - 2024 U6 - https://doi.org/10.34133/plantphenomics.0165 SN - 2643-6515 VL - 6 PB - American Association for the Advancement of Science (AAAS) ER - TY - GEN A1 - Schiller, Josepha A1 - Jänicke, Clemens A1 - Reckling, Moritz A1 - Ryo, Masahiro T1 - Higher crop rotational diversity in more simplified agricultural landscapes in Northeastern Germany T2 - Landscape Ecology N2 - Context Both crop rotational diversity and landscape diversity are important for ensuring resilient agricultural production and supporting biodiversity and ecosystem services in agricultural landscapes. However, the relationship between crop rotational diversity and landscape diversity is largely understudied. Objectives We aim to assess how crop rotational diversity is spatially organised in relation to soil, climate, and landscape diversity at a regional scale in Brandenburg, Germany. Methods We used crop rotational richness, Shannon’s diversity and evenness indices per field per decade (i.e., crop rotational diversity) as a proxy for agricultural diversity and land use and land cover types and habitat types as proxies for landscape diversity. Soil and climate characteristics and geographical positions were used to identify potential drivers of the diversity facets. All spatial information was aggregated at 10 × 10 km resolution, and statistical associations were explored with interpretable machine learning methods. Results Crop rotational diversity was associated negatively with landscape diversity metrics and positively with soil quality and the proportion of agricultural land use area, even after accounting for the other variables. Conclusion Our study indicates a spatial trade-off between crop and landscape diversity (competition for space), and crop rotations are more diverse in more simplified landscapes that are used for agriculture with good quality of soil conditions. The respective strategies and targets should be tailored to the corresponding local and regional conditions for maintaining or enhancing both crop and landscape diversity jointly to gain their synergistic positive impacts on agricultural production and ecosystem management. KW - Cropping systems KW - Explainable artificial intelligence KW - Land use KW - Landscape heterogeneity KW - Multiple scales KW - Soil quality Y1 - 2024 U6 - https://doi.org/10.1007/s10980-024-01889-x SN - 1572-9761 VL - 39 IS - 4 PB - Springer Science and Business Media LLC ER - TY - GEN A1 - Degrune, Florine A1 - Dumack, Kenneth A1 - Ryo, Masahiro A1 - Garland, Gina A1 - Romdhane, Sana A1 - Saghaï, Aurélien A1 - Banerjee, Samiran A1 - Edlinger, Anna A1 - Herzog, Chantal A1 - Pescador, David Sánchez A1 - García‐Palacios, Pablo A1 - Fiore-Donno, Anna Maria A1 - Bonkowski, Michael A1 - Hallin, Sara A1 - Heijden, Marcel G. A. van der A1 - Maestre, Fernando T. A1 - Philippot, Laurent A1 - Glemnitz, Michael A1 - Sieling, Klaus A1 - Rillig, Matthias C. T1 - The impact of fungi on soil protist communities in European cereal croplands T2 - Environmental Microbiology N2 - Protists, a crucial part of the soil food web, are increasingly acknowledged as significant influencers of nutrient cycling and plant performance in farmlands. While topographical and climatic factors are often considered to drive microbial communities on a continental scale, higher trophic levels like heterotrophic protists also rely on their food sources. In this context, bacterivores have received more attention than fungivores. Our study explored the connection between the community composition of protists (specifically Rhizaria and Cercozoa) and fungi across 156 cereal fields in Europe, spanning a latitudinal gradient of 3000 km. We employed a machine‐learning approach to measure the significance of fungal communities in comparison to bacterial communities, soil abiotic factors, and climate as determinants of the Cercozoa community composition. Our findings indicate that climatic variables and fungal communities are the primary drivers of cercozoan communities, accounting for 70% of their community composition. Structural equation modelling (SEM) unveiled indirect climatic effects on the cercozoan communities through a change in the composition of the fungal communities. Our data also imply that fungivory might be more prevalent among protists than generally believed. This study uncovers a hidden facet of the soil food web, suggesting that the benefits of microbial diversity could be more effectively integrated into sustainable agriculture practices. Y1 - 2024 U6 - https://doi.org/10.1111/1462-2920.16673 SN - 1462-2912 VL - 26 IS - 7 PB - Wiley ER - TY - GEN A1 - Guo, Tongtian A1 - Guo, Meiqi A1 - Pang, Yue A1 - Sun, Xiangyun A1 - Ryo, Masahiro A1 - Liu, Nan A1 - Zhang, Yingjun T1 - Ungulate herbivores promote beta diversity and drive stochastic plant community assembly by selective defoliation and trampling: From a four‐year simulation experiment T2 - Journal of Ecology N2 - 1. Ungulate herbivores shape grassland plant communities at multiple scales, ultimately affecting ecosystem function. However, ungulates have complex effects on grasslands, including defoliation, trampling, excreta return and their interactions. Moreover, the effects of ungulate density on grasslands are regulated by these three mechanisms. Nevertheless, how these three mechanisms affect biodiversity at multiple scales and community assembly remains poorly understood. 2. Here, we conducted a 4‐year novel field experiment to disentangle the effects of defoliation, trampling, and excreta return by ungulates on plant community assembly in a temperate grassland in Inner Mongolia, China. This experiment set two different scenarios: moderate ungulate density (Moderate, characterised by selective defoliation and moderate trampling) and high ungulate density (Intense, characterised by non‐selective defoliation and heavy trampling), including different combinations of defoliation, trampling and excreta return in each scenario. 3. We found that defoliation and trampling increased stochasticity in community assembly and promoted alpha and beta diversity under both scenarios. Specifically, defoliation promoted the coexistence of species with multiple resource acquisition strategies (higher functional trait diversity) by reducing interspecific competition; trampling tended to facilitate random species colonisation. Conversely, excreta return favoured grasses, promoting deterministic assembly and impacting species coexistence. Notably, selective defoliation in the Moderate scenario led to a dominance of stochastic processes during community assembly, whereas non‐selective defoliation still did not change the dominance of deterministic processes. Further, communities subject to selective defoliation were insensitive to changes in soil properties caused by trampling and excreta return, maintaining a high‐level beta diversity and the stochastic of community assembly. 4. Synthesis: Our study provides important insights into the mechanisms by which ungulate herbivores influence plant community assembly, suggesting that defoliation and trampling have the potential to drive stochastic processes, while excreta return plays the opposite role. Our study also suggests that selective foraging by ungulates acts as stronger stochastic forces during community assembly compared to non‐selective defoliation. These results imply that considering ungulate feeding preferences and foraging behaviour in grassland management will help prevent biodiversity loss and biotic homogenisation. KW - ungulate herbivores KW - community assembly KW - selective defoliation KW - trampling KW - excreta return KW - grassland KW - stochastic and deterministic processes KW - plant–herbivore interactions Y1 - 2024 U6 - https://doi.org/10.1111/1365-2745.14370 SN - 0022-0477 VL - 112 IS - 9 SP - 1992 EP - 2006 PB - Wiley ER - TY - GEN A1 - Safonova, Anastasiia A1 - Ryo, Masahiro T1 - Deep learning improves point density in PS-InSAR data toward finer-scale land surface displacement detection T2 - IEEE Access N2 - The permanent scatterer interferometric aperture radar (PS-InSAR) technique is used to measure and monitor displacements of the Earth’s surface over time. While the approach is promising for large-scale deformation, the density of the received PS points is insufficient for localized deformation analysis. In this first work, we aim to improve the technique by increasing the point density of high-precision deformation monitoring in PS-InSAR data by developing a convolutional long short-term memory (ConvLSTM) model that predicts PS points on different land covers, such as forest, urban, natural, water, and combinations among them. The proposed architecture, PS-ConvLSTM, was trained on a temporary dataset with interferograms to classify stable and unstable PS pixels from over 200,000 site images obtained from the city of Barcelona, Spain. The result showed that the trained PS-ConvLSTM model is highly compatible with the method currently used, which requires a large manual effort by an expert (accuracy: 99%). In addition, the proposed approach increased the point density by 15%, indicating that ConvLSTM is a promising approach for increasing the point density in PS-InSAR data and thus improving localized deformation analysis. KW - Deep learning KW - recurrent neural network KW - long short-term memory KW - point density KW - persistent scatter KW - InSAR process Y1 - 2024 U6 - https://doi.org/10.1109/ACCESS.2024.3459099 SN - 2169-3536 VL - 12 SP - 132754 EP - 132762 PB - Institute of Electrical and Electronics Engineers (IEEE) ER - TY - GEN A1 - Bi, Mohan A1 - Li, Huiying A1 - Meidl, Peter A1 - Zhu, Yanjie A1 - Ryo, Masahiro A1 - Rillig, Matthias C. T1 - Number and dissimilarity of global change factors influences soil properties and functions T2 - Nature Communications N2 - Soil biota and functions are impacted by various anthropogenic stressors, including climate change, chemical pollution or microplastics. These stressors do not occur in isolation, and soil properties and functions appear to be directionally driven by the number of global change factors acting simultaneously. Building on this insight, we here hypothesize that co-acting factors with more diverse effect mechanisms, or higher dissimilarity, have greater impacts on soil properties and functions. We created a factor pool of 12 factors and calculated dissimilarity indices of randomly-chosen co-acting factors based on the measured responses of soil properties and functions to the single factors. Results show that not only was the number of factors important, but factor dissimilarity was also key for predicting factor joint effects. By analyzing deviations of soil properties and functions from three null model predictions, we demonstrate that higher factor dissimilarity and a larger number of factors could drive larger deviations from null models and trigger more frequent occurrence of synergistic factor net interactions on soil functions (decomposition rate, cellulase, and β-glucosidase activity), which provides mechanistic insights for understanding high-dimensional effects of factors. Our work highlights the importance of considering factor similarity in future research on interacting factors. Y1 - 2024 U6 - https://doi.org/10.1038/s41467-024-52511-2 SN - 2041-1723 VL - 15 IS - 1 PB - Springer Science and Business Media LLC ER - TY - GEN A1 - Breitenbach, Romy A1 - Gerrits, Ruben A1 - Dementyeva, Polina A1 - Knabe, Nicole A1 - Schumacher, Julia A1 - Feldmann, Ines A1 - Radnik, Jörg A1 - Ryo, Masahiro A1 - Gorbushina, Anna A. T1 - The role of extracellular polymeric substances of fungal biofilms in mineral attachment and weathering T2 - npj Materials Degradation N2 - The roles extracellular polymeric substances (EPS) play in mineral attachment and weathering were studied using genetically modified biofilms of the rock-inhabiting fungus Knufia petricola strain A95. Mutants deficient in melanin and/or carotenoid synthesis were grown as air-exposed biofilms. Extracted EPS were quantified and characterised using a combination of analytical techniques. The absence of melanin affected the quantity and composition of the produced EPS: mutants no longer able to form melanin synthesised more EPS containing fewer pullulan-related glycosidic linkages. Moreover, the melanin-producing strains attached more strongly to the mineral olivine and dissolved it at a higher rate. We hypothesise that the pullulan-related linkages, with their known adhesion functionality, enable fungal attachment and weathering. The released phenolic intermediates of melanin synthesis in the Δ sdh1 mutant might play a role similar to Fe-chelating siderophores, driving olivine dissolution even further. These data demonstrate the need for careful compositional and quantitative analyses of biofilm-created microenvironments. KW - Biofilms KW - Cytogenetics KW - Geochemistry Y1 - 2022 U6 - https://doi.org/10.1038/s41529-022-00253-1 SN - 2397-2106 VL - 6 IS - 1 PB - Nature Publishing Group UK ER - TY - GEN A1 - Kakhani, Nafiseh A1 - Taghizadeh‐Mehrjardi, Ruhollah A1 - Omarzadeh, Davoud A1 - Ryo, Masahiro A1 - Heiden, Uta A1 - Scholten, Thomas T1 - Towards explainable AI : interpreting soil organic carbon prediction models using a learning‐based explanation method T2 - European Journal of Soil Science N2 - An understanding of the key factors and processes influencing the variability of soil organic carbon (SOC) is essential for the development of effective policies aimed at enhancing carbon storage in soils to mitigate climate change. In recent years, complex computational approaches from the field of machine learning (ML) have been developed for modelling and mapping SOC in various ecosystems and over large areas. However, in order to understand the processes that account for SOC variability from ML models and to serve as a basis for new scientific discoveries, the predictions made by these data‐driven models must be accurately explained and interpreted. In this research, we introduce a novel explanation approach applicable to any ML model and investigate the significance of environmental features to explain SOC variability across Germany. The methodology employed in this study involves training multiple ML models using SOC content measurements from the LUCAS dataset and incorporating environmental features derived from Google Earth Engine (GEE) as explanatory variables. Thereafter, an explanation model is applied to elucidate what the ML models have learned about the relationship between environmental features and SOC content in a supervised manner. In our approach, a post hoc model is trained to estimate the contribution of specific inputs to the outputs of the trained ML models. The results of this study indicate that different classes of ML models rely on interpretable but distinct environmental features to explain SOC variability. Decision tree‐based models, such as random forest (RF) and gradient boosting, highlight the importance of topographic features. Conversely, soil chemical information, particularly pH, is crucial for the performance of neural networks and linear regression models. Therefore, interpreting data‐driven studies requires a carefully structured approach, guided by expert knowledge and a deep understanding of the models being analysed. KW - Explainable AI KW - Germany KW - Google Earth Engine KW - HLS product KW - Remote sensing KW - Soil organic carbon Y1 - 2025 U6 - https://doi.org/10.1111/ejss.70071 SN - 1351-0754 VL - 76 IS - 2 SP - 1 EP - 18 PB - Wiley ER - TY - GEN A1 - Rivera-Palacio, Juan C. A1 - Bunn, Christian A1 - Ryo, Masahiro T1 - Factors affecting deep learning model performance in citizen science–based image data collection for agriculture : a case study on coffee crops T2 - Computers and Electronics in Agriculture N2 - Citizen science is an effective approach for collecting extensive data scalable for deep learning, although data quality is debatable. However, few studies have determined the factors associated with data collection that affect model performance and potential sampling bias. This study aims to identify the factors that significantly influence the performance of a deep learning object detection model in agricultural prediction tasks. To do so, we analyzed errors in a You Only Look Once (YOLO v8) model trained for counting the number of coffee cherries in mobile pictures. The model was trained with 436 images taken in Colombia and Peru collected by local farmers as a citizen science approach. We analyzed the prediction errors of the model using 637 additional pictures. We then applied a linear mixed model (LMM) and a decision tree machine learning model to regress the model’s error against predictor variables related to the following categories: photographer influence, geographic location, mobile phone characteristics, picture characteristics, and coffee varieties. Our results show the strong influence of photographer identity and adherence (whether the image collection protocol was followed or not) on model prediction error. Following the protocol can increase model performance from an R² of 0.48 to 0.73. Additionally, model performance varied significantly depending on photographer identity, with R² ranging from 0.45 to 0.93. In contrast, factors such as mobile phone characteristics (e.g., frontal camera resolution, flash type, and screen size), using the screen behind the branch to obscure other cherries, coffee varieties, and geographic location did not significantly affect prediction error. These findings demonstrate that data quality in citizen science–based data collection for enhancing model prediction can be achieved through straightforward and comprehensive protocols, customized volunteer training, and regular feedback from experts. Such measures collectively support the robust application of deep learning models in agriculture. Furthermore, this study demonstrated that any mobile device with a camera can contribute to citizen science initiatives, underscoring the potential and scalability of this approach in agricultural research. KW - Deep learning KW - Bias KW - Citizen science KW - Smartphone imaging KW - Coffee KW - Prediction Y1 - 2025 U6 - https://doi.org/10.1016/j.compag.2025.110096 SN - 0168-1699 VL - 232 SP - 1 EP - 13 PB - Elsevier BV CY - Amsterdam ER - TY - GEN A1 - Jeltsch, Florian A1 - Roeleke, Manuel A1 - Abdelfattah, Ahmed A1 - Arlinghaus, Robert A1 - Berg, Gabriele A1 - Blaum, Niels A1 - De Meester, Luc A1 - Dittmann, Elke A1 - Eccard, Jana Anja A1 - Fournier, Bertrand A1 - Gaedke, Ursula A1 - Gallagher, Cara A1 - Govaert, Lynn A1 - Hauber, Mark A1 - Jeschke, Jonathan M. A1 - Kramer-Schadt, Stephanie A1 - Linstädter, Anja A1 - Lucke, Ulrike A1 - Mazza, Valeria A1 - Metzler, Ralf A1 - Nendel, Claas A1 - Radchuk, Viktoriia A1 - Rillig, Matthias C. A1 - Ryo, Masahiro A1 - Scheiter, Katharina A1 - Tiedemann, Ralph A1 - Tietjen, Britta A1 - Voigt, Christian C. A1 - Weithoff, Guntram A1 - Wolinska, Justyna A1 - Zurell, Damaris T1 - The need for an individual-based global change ecology T2 - Individual-based ecology N2 - Biodiversity loss and widespread ecosystem degradation are among the most pressing challenges of our time, requiring urgent action. Yet our understanding of their causes remains limited because prevailing ecological concepts and approaches often overlook the underlying complex interactions of individuals of the same or different species, interacting with each other and with their environment. We propose a paradigm shift in ecological science, moving from simplifying frameworks that use species, population or community averages to an integrative approach that recognizes individual organisms as fundamental agents of ecological change. The urgency of the biodiversity crisis requires such a paradigm shift to advance ecology towards a predictive science by elucidating the causal mechanisms linking individual variation and adaptive behaviour to emergent properties of populations, communities, ecosystems, and ecological interactions with human interventions. Recent advances in computational technologies, sensors, and analytical tools now offer unprecedented opportunities to overcome past challenges and lay the foundation for a truly integrated Individual-Based Global Change Ecology (IBGCE). Unravelling the potential role of individual variability in global change impact analyses will require a systematic combination of empirical, experimental and modelling studies across systems, while taking into account multiple drivers of global change and their interactions. Key priorities include refining theoretical frameworks, developing benchmark models and standardized toolsets, and systematically incorporating individual variation and adaptive behaviour into empirical field work, experiments and predictive models. The emerging synergies between individual-based modelling, big data approaches, and machine learning hold great promise for addressing the inherent complexity of ecosystems. Each step in the development of IBGCE must systematically balance the complexity of the individual perspective with parsimony, computational efficiency, and experimental feasibility. IBGCE aims to unravel and predict the dynamics of biodiversity in the Anthropocene through a comprehensive study of individual organisms, their variability and their interactions. It will provide a critical foundation for considering individual variation and behaviour for future conservation and sustainability management, taking into account individual-to-ecosystem pathways and feedbacks. KW - Agent-based KW - Biodiversity crisis KW - Climate change KW - Ecological theory KW - Individual trait variation KW - Predictions KW - Scaling up Y1 - 2025 U6 - https://doi.org/10.3897/ibe.1.148200 SN - 3033-0947 VL - 1 SP - 1 EP - 18 PB - Pensoft Publishers CY - Sofia ER - TY - GEN A1 - Guo, Tongtian A1 - Guo, Meiqi A1 - Ryo, Masahiro A1 - Rillig, Matthias C. A1 - Liu, Nan A1 - Zhang, Yingjun T1 - Ungulate herbivory affects grassland soil biota β‐diversity and community assembly via modifying soil properties and plant root traits T2 - New phytologist : international journal of plant science N2 - Summary Ungulate herbivory, a widespread and complex disturbance, shapes grassland biodiversity and functions primarily through three mechanisms: defoliation, trampling, and excreta return. However, the specific effects of these mechanisms on soil biodiversity and community assembly remain unclear. We conducted a 4‐yr factorial experiment in the Eurasian steppe to investigate how defoliation, trampling, and excreta return influence soil bacterial, fungal, and nematode β‐diversity and community assembly under moderate‐ and high‐density ungulate grazing scenarios. Our findings reveal that herbivores affect soil biota through multiple pathways at different grazing intensities. Specifically, selective defoliation in the moderate‐density scenario promoted stochastic community assembly of nematodes and fungi by increasing the specific root length of plant communities. Excreta return encouraged stochastic bacterial communities by carbon input, while urine‐induced acidification and elevated ammonium levels promoted environmental filtering of bacteria and nematodes. In the high‐density scenario, non‐selective defoliation and heavy trampling created harsher soil conditions, reducing bacterial and nematode β‐diversity via habitat filtering and diminishing association of soil biota with plant roots. This study explored how different components of ungulate behaviour influence soil community assembly and highlighted the crucial role of root traits in mediating soil biota responses, providing insights into the mechanisms of soil biodiversity maintenance under complex disturbances. KW - Aboveground–belowground interactions KW - Community assembly KW - Grassland KW - Root traits KW - Soil microbe KW - Soil nematodes KW - Ungulate herbivory KW - β-diversity Y1 - 2025 U6 - https://doi.org/10.1111/nph.70199 SN - 0028-646X VL - 247 IS - 1 SP - 281 EP - 294 PB - Wiley CY - Oxford ER - TY - GEN A1 - Schiller, Josepha A1 - Stiller, Stefan A1 - Ryo, Masahiro T1 - Artificial intelligence in environmental and earth system sciences : explainability and trustworthiness T2 - Artificial intelligence review : an international science and engineering journal N2 - Explainable artificial intelligence (XAI) methods have recently emerged to gain insights into complex machine learning models. XAI can be promising for environmental and Earth system science because high-stakes decision-making for management and planning requires justification based on evidence and systems understanding. However, an overview of XAI applications and trust in AI in environmental and Earth system science is still missing. To close this gap, we reviewed 575 articles. XAI applications are popular in various domains, including ecology, engineering, geology, remote sensing, water resources, meteorology, atmospheric sciences, geochemistry, and geophysics. XAI applications focused primarily on understanding and predicting anthropogenic changes in geospatial patterns and impacts on human society and natural resources, especially biological species distributions, vegetation, air quality, transportation, and climate-water related topics, including risk and management. Among XAI methods, the SHAP and Shapley methods were the most popular (135 articles), followed by feature importance (27), partial dependence plots (22), LIME (21), and saliency maps (15). Although XAI methods are often argued to increase trust in model predictions, only seven studies (1.2%) addressed trustworthiness as a core research objective. This gap is critical because understanding the relationship between explainability and trust is lacking. While XAI applications continue to grow, they do not necessarily enhance trust. Hence, more studies on how to strengthen trust in AI applications are critically needed. Finally, this review underlines the recommendation of developing a “human-centered” XAI framework that incorporates the distinct views and needs of multiple stakeholder groups to enable trustworthy decision-making. KW - Trustworthy artificial intelligence KW - User-centered artificial intelligence KW - Explainable artificial intelligence KW - Interpretable machine learning KW - Model interpretation KW - Decision-making Y1 - 2025 U6 - https://doi.org/10.1007/s10462-025-11165-2 SN - 1573-7462 VL - 58 IS - 10 SP - 1 EP - 23 PB - Springer Netherlands CY - Dordrecht ER - TY - GEN A1 - Zhu, Yanjie A1 - Meidl, Peter A1 - Li, Huiying A1 - Bi, Mohan A1 - Ryo, Masahiro A1 - Rillig, Matthias C. T1 - Concurrent anthropogenic stressors affect plant–soil systems with different plant diversity levels T2 - New phytologist N2 - Summary Plant diversity strongly influences ecosystem functioning. Due to human activities, ecosystems are increasingly threatened by the co‐occurrence of numerous anthropogenic pressures, but how they respond to this multifaceted phenomenon is poorly documented, and what role plant diversity plays in this process has not been investigated so far. Here, plant–soil systems with different plant diversity levels (3 vs 9 species) were subjected to an increasing number of anthropogenic stressors (0, 1, 2, 5, and 8). Results show that soil properties and functions were directionally driven by stressor number, irrespective of plant diversity level, and plant functional group evenness declined continuously along the stressor number gradient. The impact of stressors on plant–soil systems varied depending on plant diversity, and when plant diversity was higher, concurrent stressors may have interacted more to affect plant–soil systems. Notably, increasing the stressor number tended to diminish the effects of plant diversity. This study represents a first attempt to address the effect of plant diversity under multi‐stressor combinations and highlights the importance of emphasizing plant–soil systems in the research field of multifactorial global change. We also suggest that efforts should be made to reduce the number of coacting stressors when managing plant–soil ecosystems. KW - Biodiversity KW - Ecosystem functioning KW - Global change KW - Interactive effect KW - Plant community KW - Plant functional group evenness KW - Sampling effect Y1 - 2025 U6 - https://doi.org/10.1111/nph.70275 SN - 0028-646X VL - 247 IS - 4 SP - 1897 EP - 1911 PB - Wiley CY - Oxford ER - TY - GEN A1 - Palka, Marlene A1 - Nendel, Claas A1 - Weiß, Lucas A1 - Schiller, Josepha A1 - Jänicke, Clemens A1 - Gaviria, Juliana Arbeláez A1 - Ryo, Masahiro T1 - Cropping history, agronomic rules, and commodity prices shape crop rotations across Central Europe T2 - Agricultural systems N2 - Context Crop rotations provide agronomic benefits over monocropping, such as enhanced nitrogen supply, improved weed and pest control, and higher yields. Although the theoretical understanding of optimal rotations has advanced, little is known about their real-world implementation and the factors influencing rotation decisions on large scales. Objective Understanding these factors is key for projecting future cropping patterns, refining agricultural policy, and improving crop models that often oversimplify rotation practices. This study identifies the drivers influencing operational crop rotations across Central Europe and projects future cropping patterns in the region. Methods We analyse over 16 million field-year combinations from Germany, Austria, and the Czech Republic. Using a random forest algorithm, we determine feature importance and apply a novel machine learning approach that incorporates uncertainty in farmers' decision-making to provide a potential outlook on cropping patterns until 2070. Results and Conclusions Historical cropping patterns, agronomic practices, and legume commodity prices significantly shaped crop rotations across the region. Projections indicate a substantial increase in legume cultivation over the coming decades, with implications for nitrogen budgets, dietary transitions, and in-silico upscaling. Significance Rather than optimizing rotations, this study identifies key drivers of operational crop rotations in Central Europe. The findings provide the basis for large-scale simulations that represent cropping patterns more realistically. To the best of our knowledge, the data set compiled here is the most extensive yet analysed in the context of operational crop rotation management. KW - Crop rotation KW - Machine learning KW - Farmer decision-making KW - Central Europe Y1 - 2026 U6 - https://doi.org/10.1016/j.agsy.2025.104522 SN - 0308-521X VL - 231 SP - 1 EP - 14 PB - Elsevier BV CY - Amsterdam ER - TY - GEN A1 - Raza, Ahsan A1 - Ryo, Masahiro A1 - Ghazaryan, Gohar A1 - Baatz, Roland A1 - Main-Knorn, Magdalena A1 - Inforsato, Leonardo A1 - Nendel, Claas T1 - Predicting regional-scale groundwater levels at high spatial resolution using spatial random forest models T2 - International journal of applied earth observation and geoinformation N2 - Groundwater is an important resource for sustainable crop growth and agricultural productivity, and groundwater level (GWL) fluctuations directly influence agroecosystems. However, GWL data for most regions is unavailable because of spatial scarcity and discontinuous groundwater observations. In recent studies, machine learning techniques have proven to be a useful tool for producing GWL estimates, but these techniques may not necessarily take into account distances and observations from the nearest points to unmonitored sites as covariates. In this study, we explored the potential of these additional spatial covariates by implementing Random Forest for Spatial Interpolation (RFSI) to generate high-resolution (1 km²) GWL estimates on a monthly scale. To evaluate the effectiveness of the RFSI model, we performed a thorough comparison of the RFSI model with a conventional Random Forest (RF) model, a Random Forest for spatial data (RFsp) and a Support Vector Machine (SVM). The framework was implemented using GWL observation data for the period 2001 to 2022 in Brandenburg, Germany. The RFSI model exhibited the highest predictive accuracy for GWL predictions at testing points, achieving an R² value of 0.86, surpassing the performance of RF, RFsp and SVM with R² values of 0.75, 0.72 and 0.67, respectively. Incorporating nearby-well information substantially improved 1 km GWL predictions: RFSI achieved the lowest RMSE (3.77 m), as compared to RF (4.85 m), RFsp (5.04 m), and SVM (5.59 m). Notably, among the models tested, RF, RFsp, and SVM showed notable inaccuracies, substantially underestimating groundwater levels (GWL) in deeper wells (>10 m) and overestimating levels in shallow wells (1–10 m). In contrast, RFSI performed better, with smaller overestimations (up to ~1.1 m) and less severe underestimations (up to ~-3.9 m). Furthermore, the RFSI model demonstrated significantly higher computational efficiency compared to RFsp and SVM, particularly for large training datasets and high-resolution prediction mapping. KW - Machine learning KW - Spatial interpolation KW - Data integration KW - Random forest for spatial interpolation KW - Groundwater level Y1 - 2025 U6 - https://doi.org/10.1016/j.jag.2025.104918 SN - 1569-8432 SN - 1872-826X VL - 144 SP - 1 EP - 19 PB - Elsevier BV CY - Amsterdam ER - TY - GEN A1 - Rivera‐Palacio, Juan C. A1 - Bunn, Christian A1 - Ryo, Masahiro T1 - Smartphone‐based monitoring identifies the importance of farm size and soil type for coffee tree productivity at a large geographic scale T2 - Journal of sustainable agriculture and environment N2 - Smartphone‐based monitoring has been increasingly applied to coffee crops for multiple tasks, such as predicting coffee tree productivity. However, its implementation remains limited to small‐scale use, typically at the individual plant level. At larger scales, such as the farm level, its application is largely unexplored. Moreover, it is unclear whether the use of smartphone‐based monitoring can help identifying key factors driving coffee tree productivity such as climate, soil, and management characteristics. To address these challenges, we investigate coffee tree productivity at the farm level and its key driving factors using smartphone‐based monitoring and explainable artificial intelligence (xAI), and compare the results with those obtained from manual monitoring at the farm level. We used a multimodal data set composed of satellite data (soil and climate), smartphone‐based monitoring (coffee tree productivity), and management characteristics (area, shade trees, and farm shape). The results showed that smartphone‐based monitoring reached a of R ² = 0.84 in predicting coffee tree productivity at the farm level. The xAI results revealed that both smartphone‐based and manual monitoring approaches identified the coffee cultivation area (greater than 13 ha) and soil texture (sandy, clay loam) as the most important variables influencing coffee tree productivity at farm level. The analysis also indicated that shade trees do not significantly affect coffee tree productivity. These findings suggest that smartphone‐based monitoring can serve as a reliable and scalable alternative to manual monitoring for evaluating coffee tree productivity at the farm level. KW - Citizen science KW - Coffee KW - Explainable artificial intelligence KW - Monitoring KW - Soil Y1 - 2025 U6 - https://doi.org/10.1002/sae2.70111 SN - 2767-035X VL - 4 IS - 4 SP - 1 EP - 9 PB - Wiley CY - Hoboken, NJ ER - TY - GEN A1 - Safonova, Anastasiia A1 - Stiller, Stefan A1 - Yordanov, Momchil A1 - Ryo, Masahiro T1 - Self-supervised learning outperforms supervised learning for crop classification by annotating only 5% of images T2 - Precision agriculture N2 - Purpose One of the most pervasive Artificial Intelligence (AI) methodologies utilized in the domain of agriculture for image-based classification purposes is Supervised Learning (SL). However, SL depends on a large amount of annotation effort and is susceptible to overfitting to the given prediction task. Self-Supervised Learning (SSL) is a novel training paradigm with the potential to address these issues, while its potential has not been investigated in the agriculture domain. This paper presents the initial experimental investigation and comparison of SL and SSL for the classification of agricultural images in the context of limited samples. Methods We used an agricultural subset of the Land Use and Cover Area Frame Survey (LUCAS) dataset serving as a case study. In total, it comprised 1,000 images for each of the 10 crops: common wheat, barley, oats, maize, potatoes, sugar beet, sunflower, rape and turnip rape, soya, and temporary grassland. For SL, we trained popular and frequently used Convolutional Neural Network (CNN) architectures such as VGG16, Inception, ResNet-18/50, SqueezeNet, ResNeXt-50, MobileNet-V2, ShuffleNet, EfficientNet-V2, and ConvNeXt Tiny with and without data augmentations. For SSL, the best-performing CNN architectures (ResNet-18, ResNet-50, and ResNeXt-50) were further tested. The architectures were pre-trained with the VICReg algorithm (Variance Invariance Covariance Regularization) and fine-tuned successively using supervision for crop type classification. Results Our results demonstrate that the SSL models can distinguish crop types (common wheat, barley, oats, maize, potatoes, sugar beet, sunflower, rape, soya, and grassland) even without labels based solely on morphological features and organize them into three semantically meaningful visual groups: cereal-like and grassland crops, upright broadleaf crops, and low-growing broadleaf crops. The fine-tuned models, particularly ResNeXt-50, achieved superior performance compared to any of the SLs. Notably, we show that the fine-tuned SSL models outperformed the best-performing SL models by using only 5% of the labeled training data for fine-tuning, corresponding to a small and balanced subset of the training split. Conclusion These findings highlight the potential of SSL for improving classification efficiency and generalization under limited data availability conditions in agriculture applications, providing a viable path toward more efficient agricultural monitoring systems. KW - Supervised learning KW - Self-supervised learning KW - VICReg KW - LUCAS dataset KW - Image classification KW - Small data problem KW - Agricultural crop Y1 - 2025 U6 - https://doi.org/10.1007/s11119-025-10302-9 SN - 1385-2256 VL - 27 IS - 1 SP - 1 EP - 31 PB - Springer Science and Business Media LLC CY - Dordrecht ER -