Reviewing the essential roles of remote phenotyping, GWAS and explainable AI in practical marker-assisted selection for drought-tolerant winter wheat breeding.

Marker-assisted selection (MAS) plays a crucial role in crop breeding improving the speed and precision of conventional breeding programmes by quickly and reliably identifying and selecting plants with desired traits. However, the efficacy of MAS depends on several prerequisites, with precise phenotyping being a key aspect of any plant breeding programme. Recent advancements in high-throughput remote phenotyping, facilitated by unmanned aerial vehicles coupled to machine learning, offer a non-destructive and efficient alternative to traditional, time-consuming, and labour-intensive methods. Furthermore, MAS relies on knowledge of marker-trait associations, commonly obtained through genome-wide association studies (GWAS), to understand complex traits such as drought tolerance, including yield components and phenology. However, GWAS has limitations that artificial intelligence (AI) has been shown to partially overcome. Additionally, AI and its explainable variants, which ensure transparency and interpretability, are increasingly being used as recognised problem-solving tools throughout the breeding process. Given these rapid technological advancements, this review provides an overview of state-of-the-art methods and processes underlying each MAS, from phenotyping, genotyping and association analyses to the integration of explainable AI along the entire workflow. In this context, we specifically address the challenges and importance of breeding winter wheat for greater drought tolerance with stable yields, as regional droughts during critical developmental stages pose a threat to winter wheat production. Finally, we explore the transition from scientific progress to practical implementation and discuss ways to bridge the gap between cutting-edge developments and breeders, expediting MAS-based winter wheat breeding for drought tolerance.

Learning vs. understanding: When does artificial intelligence outperform process-based modeling in soil organic carbon prediction?

In recent years, machine learning (ML) algorithms have gained substantial recognition for ecological modeling across various temporal and spatial scales. However, little evaluation has been conducted for the prediction of soil organic carbon (SOC) on small data sets commonly inherent to long-term soil ecological research. In this context, the performance of ML algorithms for SOC prediction has never been tested against traditional process-based modeling approaches. Here, we compare ML algorithms, calibrated and uncalibrated process-based models as well as multiple ensembles on their performance in predicting SOC using data from five long-term experimental sites (comprising 256 independent data points) in Austria. Using all available data, the ML-based approaches using Random forest and Support vector machines with a polynomial kernel were superior to all process-based models. However, the ML algorithms performed similar or worse when the number of training samples was reduced or when a leave-one-site-out cross validation was applied. This emphasizes that the performance of ML algorithms is strongly dependent on the data-size related quality of learning information following the well-known curse of dimensionality phenomenon, while the accuracy of process-based models significantly relies on proper calibration and combination of different modeling approaches. Our study thus suggests a superiority of ML-based SOC prediction at scales where larger datasets are available, while process-based models are superior tools when targeting the exploration of underlying biophysical and biochemical mechanisms of SOC dynamics in soils. Therefore, we recommend applying ensembles of ML algorithms with process-based models to combine advantages inherent to both approaches.

Semantic segmentation of plant roots from RGB (mini-) rhizotron images-generalisation potential and false positives of established methods and advanced deep-learning models.

BACKGROUND: Manual analysis of (mini-)rhizotron (MR) images is tedious. Several methods have been proposed for semantic root segmentation based on homogeneous, single-source MR datasets. Recent advances in deep learning (DL) have enabled automated feature extraction, but comparisons of segmentation accuracy, false positives and transferability are virtually lacking. Here we compare six state-of-the-art methods and propose two improved DL models for semantic root segmentation using a large MR dataset with and without augmented data. We determine the performance of the methods on a homogeneous maize dataset, and a mixed dataset of > 8 species (mixtures), 6 soil types and 4 imaging systems. The generalisation potential of the derived DL models is determined on a distinct, unseen dataset. RESULTS: The best performance was achieved by the U-Net models; the more complex the encoder the better the accuracy and generalisation of the model. The heterogeneous mixed MR dataset was a particularly challenging for the non-U-Net techniques. Data augmentation enhanced model performance. We demonstrated the improved performance of deep meta-architectures and feature extractors, and a reduction in the number of false positives. CONCLUSIONS: Although correction factors are still required to match human labelled root lengths, neural network architectures greatly reduce the time required to compute the root length. The more complex architectures illustrate how future improvements in root segmentation within MR images can be achieved, particularly reaching higher segmentation accuracies and model generalisation when analysing real-world datasets with artefacts-limiting the need for model retraining.

Deep soil exploration vs. topsoil exploitation: distinctive rooting strategies between wheat landraces and wild relatives.

AIMS: Diversity of root systems among genetic resources can contribute to optimize water and nutrient uptake. Topsoil exploitation vs. deep soil exploration represent two contrasting ideotypes in relation to resource use. Our study reveals how rooting patterns changed between wheat wild progenitors and landraces in regard to these ideotypes. METHODS: Root (partitioning, morphology, distribution, elongation, anatomy) and shoot traits (dry-matter, leaf area, assimilation) of durum landraces, wild emmer and wild einkorn from Iran, Syria, Turkey and Lebanon were phenotyped using the GrowScreen-Rhizo platform. Distinctive rooting patterns were identified via principal component analysis and relations with collection site characteristics analyzed. RESULTS: Shoot trait differentiation was strongly driven by seed weight, leading to superior early vigor of landraces. Wild progenitors formed superficial root systems with a higher contribution of lateral and early-emerging nodal axes to total root length. Durum landraces had a root system dominated by seminal axes allocated evenly over depth. Xylem anatomy was the trait most affected by the environmental influence of the collection site. CONCLUSIONS: The durum landrace root system approximated a deep soil exploration ideotype which would optimize subsoil water uptake, while monococcum-type wild einkorn was most similar to a topsoil exploiting strategy with potential competitive advantages for subsistence in natural vegetation. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s11104-020-04794-9.

Root System Phenotying of Soil-Grown Plants via RGB and Hyperspectral Imaging.

Phenotyping root systems provide essential information for plant breeding, particularly aiming for better abiotic stress resistance. Rhizobox systems provide a field-near growth environment for in situ imaging of root systems in soil. A protocol for RGB and hyperspectral imaging of rhizobox-grown plants is presented that enables gathering of root structural (morphology, architecture) as well as functional (water content, decomposition) information. The protocol exemplifies the setup of a root phenotyping platform combining low-cost RGB with advanced short-wave infrared hyperspectral imaging. For both types of imaging approach, the essential steps of an image analysis pipeline are provided to retrieve biological information on breeding-relevant traits from the imaging datasets.

Characterization of Cover Crop Rooting Types from Integration of Rhizobox Imaging and Root Atlas Information.

Plant root systems are essential for sustainable agriculture, conveying resource-efficient genotypes and species with benefits to soil ecosystem functions. Targeted selection of species/genotypes depends on available root system information. Currently there is no standardized approach for comprehensive root system characterization, suggesting the need for data integration across methods and sources. Here, we combine field measured root descriptors from the classical Root Atlas series with traits from controlled-environment root imaging for 10 cover crop species to (i) detect descriptors scaling between distant experimental methods, (ii) provide traits for species classification, and (iii) discuss implications for cover crop ecosystem functions. Results revealed relation of single axes measures from root imaging (convex hull, primary-lateral length ratio) to Root Atlas field descriptors (depth, branching order). Using composite root variables (principal components) for branching, morphology, and assimilate investment traits, cover crops were classified into species with (i) topsoil-allocated large diameter rooting type, (ii) low-branched primary/shoot-born axes-dominated rooting type, and (iii) highly branched dense rooting type, with classification trait-dependent distinction according to depth distribution. Data integration facilitated identification of root classification variables to derive root-related cover crop distinction, indicating their agro-ecological functions.

Hyperspectral imaging: a novel approach for plant root phenotyping.

BACKGROUND: Root phenotyping aims to characterize root system architecture because of its functional role in resource acquisition. RGB imaging and analysis procedures measure root system traits via colour contrasts between roots and growth media or artificial backgrounds. In the case of plants grown in soil-filled rhizoboxes, where the colour contrast can be poor, it is hypothesized that root imaging based on spectral signatures improves segmentation and provides additional knowledge on physico-chemical root properties. RESULTS: Root systems of Triticum durum grown in soil-filled rhizoboxes were scanned in a spectral range of 1000-1700 nm with 222 narrow bands and a spatial resolution of 0.1 mm. A data processing pipeline was developed for automatic root segmentation and analysis of spectral root signatures. Spectral- and RGB-based root segmentation did not significantly differ in accuracy even for a bright soil background. Best spectral segmentation was obtained from log-linearized and asymptotic least squares corrected images via fuzzy clustering and multilevel thresholding. Root axes revealed major spectral distinction between center and border regions. Root decay was captured by an exponential function of the difference spectra between water and structural carbon absorption regions. CONCLUSIONS: Fundamentals for root phenotyping using hyperspectral imaging have been established by means of an image processing pipeline for automated segmentation of soil-grown plant roots at a high spatial resolution and for the exploration of spectral signatures encoding physico-chemical root zone properties.

Faba Bean Cultivation - Revealing Novel Managing Practices for More Sustainable and Competitive European Cropping Systems.

Faba beans are highly nutritious because of their high protein content: they are a good source of mineral nutrients, vitamins, and numerous bioactive compounds. Equally important is the contribution of faba bean in maintaining the sustainability of agricultural systems, as it is highly efficient in the symbiotic fixation of atmospheric nitrogen. This article provides an overview of factors influencing faba bean yield and quality, and addresses the main biotic and abiotic constraints. It also reviews the factors relating to the availability of genetic material and the agronomic features of faba bean production that contribute to high yield and the improvement of European cropping systems. Emphasis is to the importance of using new high-yielding cultivars that are characterized by a high protein content, low antinutritional compound content, and resistance to biotic and abiotic stresses. New cultivars should combine several of these characteristics if an increased and more stable production of faba bean in specific agroecological zones is to be achieved. Considering that climate change is also gradually affecting many European regions, it is imperative to breed elite cultivars that feature a higher abiotic-biotic stress resistance and nutritional value than currently used cultivars. Improved agronomical practices for faba bean crops, such as crop establishment and plant density, fertilization and irrigation regime, weed, pest and disease management, harvesting time, and harvesting practices are also addressed, since they play a crucial role in both the production and quality of faba bean.

Root traits of European Vicia faba cultivars-Using machine learning to explore adaptations to agroclimatic conditions.

Faba bean (Vicia faba L.) is an important source of protein, but breeding for increased yield stability and stress tolerance is hampered by the scarcity of phenotyping information. Because comparisons of cultivars adapted to different agroclimatic zones improve our understanding of stress tolerance mechanisms, the root architecture and morphology of 16 European faba bean cultivars were studied at maturity. Different machine learning (ML) approaches were tested in their usefulness to analyse trait variations between cultivars. A supervised, that is, hypothesis-driven, ML approach revealed that cultivars from Portugal feature greater and coarser but less frequent lateral roots at the top of the taproot, potentially enhancing water uptake from deeper soil horizons. Unsupervised clustering revealed that trait differences between northern and southern cultivars are not predominant but that two cultivar groups, independently from major and minor types, differ largely in overall root system size. Methodological guidelines on how to use powerful ML methods such as random forest models for enhancing the phenotypical exploration of plants are given.

RGB and Spectral Root Imaging for Plant Phenotyping and Physiological Research: Experimental Setup and Imaging Protocols.

Better understanding of plant root dynamics is essential to improve resource use efficiency of agricultural systems and increase the resistance of crop cultivars against environmental stresses. An experimental protocol is presented for RGB and hyperspectral imaging of root systems. The approach uses rhizoboxes where plants grow in natural soil over a longer time to observe fully developed root systems. Experimental settings are exemplified for assessing rhizobox plants under water stress and studying the role of roots. An RGB imaging setup is described for cheap and quick quantification of root development over time. Hyperspectral imaging improves root segmentation from the soil background compared to RGB color based thresholding. The particular strength of hyperspectral imaging is the acquisition of chemometric information on the root-soil system for functional understanding. This is demonstrated with high resolution water content mapping. Spectral imaging however is more complex in image acquisition, processing and analysis compared to the RGB approach. A combination of both methods can optimize a comprehensive assessment of the root system. Application examples integrating root and aboveground traits are given for the context of plant phenotyping and plant physiological research. Further improvement of root imaging can be obtained by optimizing RGB image quality with better illumination using different light sources and by extension of image analysis methods to infer on root zone properties from spectral data.

Root architecture simulation improves the inference from seedling root phenotyping towards mature root systems.

Root phenotyping provides trait information for plant breeding. A shortcoming of high-throughput root phenotyping is the limitation to seedling plants and failure to make inferences on mature root systems. We suggest root system architecture (RSA) models to predict mature root traits and overcome the inference problem. Sixteen pea genotypes were phenotyped in (i) seedling (Petri dishes) and (ii) mature (sand-filled columns) root phenotyping platforms. The RSA model RootBox was parameterized with seedling traits to simulate the fully developed root systems. Measured and modelled root length, first-order lateral number, and root distribution were compared to determine key traits for model-based prediction. No direct relationship in root traits (tap, lateral length, interbranch distance) was evident between phenotyping systems. RootBox significantly improved the inference over phenotyping platforms. Seedling plant tap and lateral root elongation rates and interbranch distance were sufficient model parameters to predict genotype ranking in total root length with an RSpearman of 0.83. Parameterization including uneven lateral spacing via a scaling function substantially improved the prediction of architectures underlying the differently sized root systems. We conclude that RSA models can solve the inference problem of seedling root phenotyping. RSA models should be included in the phenotyping pipeline to provide reliable information on mature root systems to breeding research.

Phenotyping: Using Machine Learning for Improved Pairwise Genotype Classification Based on Root Traits.

Phenotyping local crop cultivars is becoming more and more important, as they are an important genetic source for breeding - especially in regard to inherent root system architectures. Machine learning algorithms are promising tools to assist in the analysis of complex data sets; novel approaches are need to apply them on root phenotyping data of mature plants. A greenhouse experiment was conducted in large, sand-filled columns to differentiate 16 European Pisum sativum cultivars based on 36 manually derived root traits. Through combining random forest and support vector machine models, machine learning algorithms were successfully used for unbiased identification of most distinguishing root traits and subsequent pairwise cultivar differentiation. Up to 86% of pea cultivar pairs could be distinguished based on top five important root traits (Timp5) - Timp5 differed widely between cultivar pairs. Selecting top important root traits (Timp) provided a significant improved classification compared to using all available traits or randomly selected trait sets. The most frequent Timp of mature pea cultivars was total surface area of lateral roots originating from tap root segments at 0-5 cm depth. The high classification rate implies that culturing did not lead to a major loss of variability in root system architecture in the studied pea cultivars. Our results illustrate the potential of machine learning approaches for unbiased (root) trait selection and cultivar classification based on rather small, complex phenotypic data sets derived from pot experiments. Powerful statistical approaches are essential to make use of the increasing amount of (root) phenotyping information, integrating the complex trait sets describing crop cultivars.

Identification of Water Use Strategies at Early Growth Stages in Durum Wheat from Shoot Phenotyping and Physiological Measurements.

Modern imaging technology provides new approaches to plant phenotyping for traits relevant to crop yield and resource efficiency. Our objective was to investigate water use strategies at early growth stages in durum wheat genetic resources using shoot imaging at the ScreenHouse phenotyping facility combined with physiological measurements. Twelve durum landraces from different pedoclimatic backgrounds were compared to three modern check cultivars in a greenhouse pot experiment under well-watered (75% plant available water, PAW) and drought (25% PAW) conditions. Transpiration rate was analyzed for the underlying main morphological (leaf area duration) and physiological (stomata conductance) factors. Combining both morphological and physiological regulation of transpiration, four distinct water use types were identified. Most landraces had high transpiration rates either due to extensive leaf area (area types) or both large leaf areas together with high stomata conductance (spender types). All modern cultivars were distinguished by high stomata conductance with comparatively compact canopies (conductance types). Only few landraces were water saver types with both small canopy and low stomata conductance. During early growth, genotypes with large leaf area had high dry-matter accumulation under both well-watered and drought conditions compared to genotypes with compact stature. However, high stomata conductance was the basis to achieve high dry matter per unit leaf area, indicating high assimilation capacity as a key for productivity in modern cultivars. We conclude that the identified water use strategies based on early growth shoot phenotyping combined with stomata conductance provide an appropriate framework for targeted selection of distinct pre-breeding material adapted to different types of water limited environments.

Can diversity in root architecture explain plant water use efficiency? A modeling study.

Drought stress is a dominant constraint to crop production. Breeding crops with adapted root systems for effective uptake of water represents a novel strategy to increase crop drought resistance. Due to complex interaction between root traits and high diversity of hydrological conditions, modeling provides important information for trait based selection. In this work we use a root architecture model combined with a soil-hydrological model to analyze whether there is a root system ideotype of general adaptation to drought or water uptake efficiency of root systems is a function of specific hydrological conditions. This was done by modeling transpiration of 48 root architectures in 16 drought scenarios with distinct soil textures, rainfall distributions, and initial soil moisture availability. We find that the efficiency in water uptake of root architecture is strictly dependent on the hydrological scenario. Even dense and deep root systems are not superior in water uptake under all hydrological scenarios. Our results demonstrate that mere architectural description is insufficient to find root systems of optimum functionality. We find that in environments with sufficient rainfall before the growing season, root depth represents the key trait for the exploration of stored water, especially in fine soils. Root density, instead, especially near the soil surface, becomes the most relevant trait for exploiting soil moisture when plant water supply is mainly provided by rainfall events during the root system development. We therefore concluded that trait based root breeding has to consider root systems with specific adaptation to the hydrology of the target environment.

Dissection of drought response of modern and underutilized wheat varieties according to Passioura's yield-water framework.

Trait-based breeding is essential to improve wheat yield, particularly when stress adaptation is targeted. A set of modern and underutilized wheat genotypes was examined in a 2-year field experiment with distinct seasonal water supply. Yield formation and drought response strategies were analyzed in relation to components of Passioura's yield-water framework based on phenological, morphological, physiological, and root characteristics. Limited water supply resulted in 60% yield loss and substantially lower water use (37%), water use efficiency (32.6%), and harvest index (14%). Phenology and root length density were key determinants of water use. Late flowering underutilized wheat species with large root system and swift ground coverage showed greatest water use. Leaf chlorophyll concentration and stomata conductance were higher in modern cultivars, supporting their high biomass growth and superior water use efficiency. While, lower chlorophyll concentration and stomata conductance of underutilized wheats indicated a water saving strategy with an intrinsic limitation of potential growth. Harvest index was strongly dependent on phenology and yield components. Optimized flowering time, reduced tillering, and strong grain sink of modern cultivars explained higher harvest index compared to underutilized wheats. Cluster analysis revealed the consistent differentiation of underutilized and modern wheats based on traits underlying Passioura's yield-water framework. We identified physiological and root traits within modern cultivars to be targeted for trait-based crop improvement under water-limited conditions. High capacity of water use in underutilized genetic resources is related to yield-limiting phenological and morphological traits, constraining their potential role for better drought resistance. Still some genetic resources provide adaptive features for stress resistance compatible with high yield as revealed by high harvest index under drought of Khorasan wheat.

A statistical approach to root system classification.

Plant root systems have a key role in ecology and agronomy. In spite of fast increase in root studies, still there is no classification that allows distinguishing among distinctive characteristics within the diversity of rooting strategies. Our hypothesis is that a multivariate approach for "plant functional type" identification in ecology can be applied to the classification of root systems. The classification method presented is based on a data-defined statistical procedure without a priori decision on the classifiers. The study demonstrates that principal component based rooting types provide efficient and meaningful multi-trait classifiers. The classification method is exemplified with simulated root architectures and morphological field data. Simulated root architectures showed that morphological attributes with spatial distribution parameters capture most distinctive features within root system diversity. While developmental type (tap vs. shoot-borne systems) is a strong, but coarse classifier, topological traits provide the most detailed differentiation among distinctive groups. Adequacy of commonly available morphologic traits for classification is supported by field data. Rooting types emerging from measured data, mainly distinguished by diameter/weight and density dominated types. Similarity of root systems within distinctive groups was the joint result of phylogenetic relation and environmental as well as human selection pressure. We concluded that the data-define classification is appropriate for integration of knowledge obtained with different root measurement methods and at various scales. Currently root morphology is the most promising basis for classification due to widely used common measurement protocols. To capture details of root diversity efforts in architectural measurement techniques are essential.