A total of five red clover germplasm accessions (F3474, F4079, F3514, F2507 & F3594), one elite pre-breeding red clover top-cross line (JB4), one commercial red clover cultivar (Grasslands Relish) and one white clover cultivar (Grasslands Quartz) used in this experiment were sourced from the Margot Forde Germplasm Centre (MFGC) ( Table 1 ). The five germplasm accessions were selected based on their place of origin and information on their collection sites. An accession is a distinct collection of uniquely identified seed, which is maintained as a part of a germplasm collection (National Academies of Sciences, Engineering, and Medicine, 1991). The elite line JB4 was selected due to its genetic background, it was produced from a top-cross involving plants from the cultivar Colenso and elite germplasm accessions. An elite line is defined as a group of plants with similar traits that are well adapted to environmental conditions in a targeted region. Which will be used for further breeding or released as a cultivar (National Academies of Sciences, Engineering, and Medicine, 1991; Falk, 2010). Relish and Quartz were selected as superior cultivar comparators, the latter added for interspecific comparison. Cultivars are a collection of plants with distinguishable characters created by breeders to meet specific breeding objectives (National Academies of Sciences, Engineering, and Medicine, 1991; Bernardo, 2002). From here onwards all accessions, elite lines and cultivars will be referred to as populations. A population is a collection of plants with similar backgrounds, individuals within potentially interbreed with one another (National Academies of Sciences, Engineering, and Medicine, 1991).

A randomized complete block design comprised of five replicates and three treatments containing 120 plants was used for this experiment. Each pot contained one plant. Six-liter plastic pots were filled with a soil mixture made with 75:25 ratio of Templeton silt loam soil and mortar sand. Additional nutrients were added for plant health and soil structure ( Supplementary Information ). This mixture is free draining to allow for the least resistant root growth and simpler separation of roots at final harvest (Ballizany et al., 2012). A further 108 potted red clover plants were placed next to the outer rows of pots, as a buffer, to minimized border effects. Plants were established in root trainers for 63 days and then transplanted into the pots and allowed to establish for 30 days before experiment initiation. The experiment ran for 52 days from treatment initiation between January 16th, 2021, and March 8th, 2021, at the AgResearch research farm glasshouse. The mean temperature was 24.4°C, average relative humidity 66% and daily solar radiation 19.5 MJ/m2.

All plants were watered to 15% volumetric soil moisture content (VMC) for 30 days to establish the experiment before initiation. Throughout the experiment the control plants maintained this status through watering by hand. In the water deficit stress treatment, plants were kept 3-4% above the wilting point of 5% VMC. This was measured gravimetrically, and the calculated amount of water was applied every 2-3 days (Hofmann et al., 2003). Plant populations were exposed to a minimum of 30 days water deficit stress. Waterlogging stress was applied by submerging the roots in a double pot system. The water level was kept at 2-3cm above pot height by encompassing the pot in a bucket filled with water. This system stopped the drainage of excess water, and the pots were topped up with water to ensure adequate submergence when needed (Simova-Stoilova et al., 2012). VMC was 15%, 5% and 50% for the control, water deficit, and waterlogged treatments, respectively. Plants were supported with bamboo poles when needed, prostrate growing accessions were not staked up to avoid false height measurements. The red clover populations were exposed to waterlogging for the first five days of the experiment between January 16^th- 20^th 2021. When plants showed signs of stress through wilting the waterlogging treatment was stopped, and plants were allowed to recover. The waterlogging treatment for these plants was then reintroduced four days before the experiment finished. The white clover cultivar Quartz remained submerged for 52 days, as it showed no signs of wilting stress during that period.

Statistical analysis was carried out using the statistical software package Genstat 18th edition (VSN International, 2021) The unbalanced analysis of variance (ANOVA) using the regression model statistical method was selected for all trait analysis due to the incomplete block effect from plants dying during the experiment. After accounting for block effects, the ANOVA examined the levels of population and of water levels (control, water deficit and waterlogging) as the two main treatment factors, followed by the examination of the interaction of these two main factors. Raw trait data was either square root or log transformed to correct skewed data to conform to normality. Fisher’s unprotected least significant difference (LSD) test was used to separate treatment means wherever the ANOVA showed a significant treatment or interaction effect (P< 0.05). Interspecies differences were obtained using the comparisons contrast function within ANOVA in Genstat. Pattern analysis, a combination of cluster analysis and principal component analysis (PCA), was conducted using means generated from ANOVA, for traits with significant (P< 0.05) differences among populations using DeltaGen (Jahufer and Luo, 2018).

Prior to pattern analysis the data were standardized to remove scaling effects. The optimum number of entry groups from cluster analysis was determined according to Cooper and DeLacy (1994). The resulting biplot is a graphical summary of population by trait associations. Trait associations are presented through directional vectors, a positive relationship between traits is shown if vectors are separated by less than 90°. The results of the clustering analysis are superimposed on the biplot through assigning each group of populations a different color. Quartz was not included in the cluster analysis for water deficit and waterlogging to not affect the differences among the red clover populations. F2507 was not included in the cluster analysis for waterlogging as it did not survive the waterlogging treatment. Means and standard errors produced from ANOVA were used to generate column graphs.

The average trait expression of each of the eight populations across water treatments is summarized in Table 2 . Both F3594 and Relish were classified as large leaved, tall erect populations with large root dry mass and at the high end of the scale for the other morphological traits. Both JB4 and F3474 averaged medium measurements for most of the traits. JB4 had a tall erect growth habit similar to F3594 and Relish but smaller root dry mass, leaf thickness, and leaf size. Populations F3514, F2507, and F4079 all measured as the smallest or lowest for all of the traits. These Populations were characterized as being small-leaved, with low biomass and prostrate growth. While trait expressions changed within each population across water treatments ( Tables 3 , 4 ), there were no changes among populations. This indicated that the relative performance of populations that expressed the smallest or lowest for each trait under control conditions did not change under both water stress treatments.

Compared to control conditions, under water deficit and waterlogging, there were decreases in total dry matter (41% and 30%, respectively), root dry mass (19% and 30%, respectively), leaf thickness (50% and 28%, respectively), plant height (15% and 9%, respectively), and leaf number (50 and 34%, respectively); ( Figures 1A, C , 2A, C ). Under waterlogging both petiole length (10%) and leaf size (8%) increased, in contrast under water deficit both decreased (25% for petiole length, 11% for leaf size) ( Figures 2A , 3B ). Plant physiology and water status traits were affected by the different treatments, both chlorophyll content (5%) and solute potential (5%) increased under waterlogging but decreased under water deficit (2%, 23%, respectively) ( Figures 2B, C ). Under water deficit, both runner dry matter and root to shoot ratio increased (10%, 38%, respectively), relative water content also decreased by 10% ( Figure 1B ).

Under water deficit, all populations significantly decreased total dry matter, but increased dry matter percentage ( Figure 1A , Table 3 ). Plant height and petiole length significantly decreased for all populations except for F3514 and F3594, while JB4 also retained plant height ( Figures 1C , 2A ). Both leaf thickness and leaf number were significantly decreased in all the populations except for F3474 which was able to maintain leaf number ( Figures 3A, C ). There was a significant decrease in leaf size for populations F4079, F2507, JB4 and Quartz ( Figure 3B ). Root dry mass decreased for populations F3594 and Quartz, while root to shoot ratio increased for populations F3474, F4079, F3514, F2507, F3594, JB4, and Relish ( Figure 1B , Table 3 ). While relative water content significantly decreased for Relish and Quartz and chlorophyll content significantly increased for populations F3594 and Quartz ( Table 2 ; Figure 2C ).

Under waterlogging, total dry matter and root dry mass decreased for all the red clover populations while Quartz only had a significant reduction in total dry matter ( Figure 1A ). Leaf size increased for populations F3474 and F4079 ( Figure 3B ). Leaf thickness decreased in populations F4079, F3594, JB4 and Relish ( Figure 3A ). Plant height decreased for populations F3474, F3594 and Relish, while increasing for Quartz ( Figure 1C ). Petiole length increased significantly for population F3594 ( Figure 2A ). In all the red clover populations leaf number significantly decreased except for population F3514 ( Figure 3C ). Root to shoot ratio significantly decreased for population F3514 and significantly increased for Quartz ( Figure 1B ). In all the red clover populations chlorophyll content significantly decreased except for in population F3474 ( Figure 2C ).

Under water deficit, there were three significant differences between red clover and white clover responses. Leaf size and plant height decreased 10% and 11%, respectively for red clover and 29% and 43% for white clover, respectively. Relative water content was reduced 7% and 22% for red clover and white clover, respectively ( Table 3 ). Under waterlogging there were large significant differences between red clover and white clover for total dry matter (39%, 2% decreases, respectively), root dry mass (83% decrease, 1% increase, respectively), leaf number (50% decrease, 13% increase, respectively), leaf thickness (37% and 3% decreases, respectively) and root to shoot ratio (24% decrease, 67% increase, respectively). There were lesser significant differences between red clover and white clover for dry matter percentage (1% increase, 11% decrease, respectively), petiole length (6% and 22% increases, respectively) and relative water content (1% and 10% decreases, respectively) ( Table 4 ).

The principal component analysis for water deficit response in the seven red clover populations showed that the first two principal components explained 69.3% of the variation in the data. The traits and their positive to negative associations with each other are indicated by the directional vectors ( Figure 4 ). Differences between populations were derived from hierarchical cluster analysis. The seven red clover populations were clustered into five groups, indicated in color. Group 1; F3514 had large reductions in total biomass, leaf number and leaf thickness under water deficit. Group 2; F3594, JB4 and Relish had large reductions in leaf thickness and leaf number, along with a small change in leaf size, petiole length and plant height. Group 3; F2507 had high reductions in leaf size, plant height and petiole length. Group 4; F4079 had low reductions in total biomass, leaf thickness, leaf size and leaf number. It also had a large reduction in petiole length. Group 5; F3474 had lower reductions in total biomass, a high increase in dry matter percentage.

The principal component analysis for waterlogging response in six red clover populations showed that the first two principal components explained 74.9% of the variation in the data. The relationships among the traits are shown by the associations among the directional vectors color ( Figure 5 ). Differences between populations were derived from hierarchical cluster analysis. The red clover populations were clustered into groups, indicated in Group 1; F3514 had a high reduction in root dry mass but was able to maintain similar measurements for the other traits. Group 2; F3474 and F4079 had high reductions in total biomass but had low reduction in root dry mass. Group 3; F3594 and Relish had low reductions in biomass, leaf thickness, leaf size and plant height. They also had a medium reduction in root dry mass. Group 4; JB4 had high reductions in total biomass, leaf thickness, leaf size and plant height.

The principal component analysis for the combined waterlogging and water deficit stress plant response showed that the first two principal components explained 69.1% of the variation in the data ( Figure 6 ). There were four groups of traits with multiple significant relationships with other traits. These groups can be identified as aboveground plant biomass traits (leaf number, total dry matter, chlorophyll content and leaf thickness), leaf response traits (petiole length and plant height), water status traits (relative water content and root to shoot ratio), and water deficit traits (runner dry matter percentage, root dry mass, leaf size and solute potential). The seven populations were clustered into groups, indicated in color which aligned with the four groups identified above ( Figure 6 ). Group 1; F3514 and JB4 had high reductions in total biomass, root dry mass and number of leaves. Group 2; F4079 and F3474 had high reductions in plant height, petiole length and root dry mass. Group 3; Quartz had the lowest reductions in total biomass and number of leaves. It also increased chlorophyll content. Group 4; F3594 and Relish increased in petiole length and had medium reductions in total biomass. F2507 was not included in this analysis as it did not survive the waterlogging treatment.

The reduction of total dry matter, leaf number, petiole length and plant height under water deficit conditions is a consistent phenomenon highlighted in the literature (Peterson et al., 1992; White and Hodgson, 1999; Vaseva et al., 2011; Scoffoni et al., 2014). The severity of the reduction varies on individual plants as various plant organs respond differently (Wright and Westoby, 2002; Aroca, 2012). By reducing aboveground biomass under water deficit, root growth is prioritized over producing total dry matter. This provides adequate water for root growth from the limited supply until the plant can resume normal growth post water deficit. This results in an increase of root to shoot ratio. (White and Hodgson, 1999). All red clover populations prioritized maintaining root dry mass under water deficit by sacrificing shoot growth as evident by the increasing root-shoot ratio in our experiment. An increase in root to shoot ratio can be attributed to water deficit resistance, with active investment in root growth or maintenance the plant is able to access sufficient available water without fully depleting stored energy reserves (Aroca, 2012).

Under waterlogged conditions, the reduction of plant height, leaf number and total shoot dry matter has been found in red clover and other Trifolium species (Simova-Stoilova et al., 2012; Enkhbat et al., 2021). This reduction is an indication of the lack of tolerance to waterlogging (Bowley et al., 1984). In contrast, plants that tolerate waterlogging elongate shoot growth above the water level to maximize leaf area exposure, ensuring root aeration through the intake of oxygen and carbon fixation (Parent et al., 2008). Waterlogging tolerant species such as white clover are likely to increase root to shoot ratio due to the shift to the less efficient anaerobic energy production pathway needing more roots to support the same amount of growth (Striker, 2012a). Petiole length increased amongst waterlogged populations, this appeared to be the plants trying to efficiently intercept the most amount of light possible with fewer leaves by elongating the petiole above the water level (Black et al., 2009; Huber et al., 2009). Unlike the red clover populations, the white clover cultivar Quartz was able to elongate shoot growth above the water level and increase root to shoot ratio to support similar growth under the less efficient anaerobic energy production pathway.

For plants, developing organs such as leaves is a large investment and construction costs for each varies amongst species. The lifespan of these organs must allow the plant to cover initial investment costs and gain additional energy to promote plant growth (Westoby et al., 2002). When exposed to water stress how a plant invests the available resources changes, including what to invest into and at what scale. Other factors such as water level, light level and temperature can influence this investment. Leaf thickness and size decreased in water deficit plants, reflecting results shown in other clover studies (Huber et al., 2009; Ballizany et al., 2012; Scoffoni et al., 2014). Maintaining leaf thickness and size comes at a great cost to plants. However, by switching to producing fewer smaller leaves requires less energy and reduces transpiration water loss. The trade-off towards thicker leaves is an increase in durability and longevity (Westoby et al., 2002). While decreased leaf size and thickness is advantageous under restricted water availability it does reduce the agronomic productivity of a plant (Aroca, 2012). As in previous red clover studies (Mommer and Visser, 2005; Huber et al., 2009), the waterlogged plants produced larger and thinner leaves. In ideal conditions, plants that produce thinner leaves can increase biomass quicker by producing new leaves at faster rate. Under waterlogging, plants produced thinner leaves, and in previous studies this has been related to reductions in transpiration and photosynthesis (Huber et al., 2009; Striker, 2012b).

Root structure plays an important role in the storage of energy, response to water stress and subsequent recovery. As expected, both treatments caused a reduction in root dry mass. Similar observations on root structure were made by Christie and Martin (1999) and Aroca (2012). The differences in root dry mass ( Table 4 ) under waterlogging conditions suggest that the white clover root system is better suited for these conditions, as Quartz was the only population that maintained root dry mass. Following water stress, the amount of reserve carbohydrates depleted during this period impacts the longevity and recovery of a plant. During water stress, plants switch to a less efficient anaerobic energy production pathway. Carbohydrates are used to keep up with energy demands. These are limited and so must be used efficiently and effectively to maintain growth. Plants that use less carbohydrate reserve or can continue to add to reserves during this period are able to resume pre-stress growth with more speed and vigor. Plants with very depleted reserves recover more slowly post-stress (Striker, 2012a). It can be suggested that under waterlogged conditions the red clover populations depleted their root reserves quickly resulting in slower recovery, whereas Quartz was able to continue growth close to pre-stress levels. The decreases in root dry mass ( Table 3 ) under water deficit conditions suggest that the depletion of root reserves varied between populations.

Under drought conditions, relative water content decreases due to the reduced availability of water, resulting in a reduction of leaf size. A result of this is an increase in chlorophyll concentration (Ling et al., 2010). Chlorophyll content was measured here as an indicator for photosynthetic capacity of the populations, and there was a slight increase for two of the red clover water deficit populations. F3594, unlike the other red clover populations was able to actively raise chlorophyll content, indicating the photosynthetic capacity of these plants had increased ( Figure 3C ) (Palta, 1990; Ling et al., 2010). Solute potential also decreased amongst populations in response to the water deficit,

Under waterlogging, a reduction in chlorophyll content among the red clover populations indicated a decrease in photosynthetic activity, reducing three key morphological traits, total dry matter, leaf number and root dry mass. This showed a lack of tolerance to waterlogging as the populations spent stored energy but were not able to produce sufficient shoot growth. There is limited literature on both traits in red clover under waterlogging conditions, but similar results have been found in other leguminous species (Simova-Stoilova et al., 2012; Stevenson et al., 2017 ppr; Enkhbat et al., 2021; Kyu et al., 2021). Under waterlogging both traits are affected by the plants reduced uptake of oxygen and carbon dioxide due to stomatal closure and a decrease in the uptake of water and key nutrients such as nitrogen due to reduced root function (Striker, 2012b).

When developing cultivars, assessment of persistence and performance across multiple environments and stresses is crucial. This enables the identification of specific or broad adaptation amongst populations, and stress coping strategies. To our knowledge this is the first published research observing red clover germplasm populations under both water deficit and waterlogging treatments in the same controlled conditions. While no population exceeded expectation or showed full acclimatization under both treatments, four coping mechanisms were identified ( Figure 6 ). Both F3514 and JB4 had high reductions in aboveground and root dry mass resulting in low performance across treatments and lack of broad adaptation. F2507 was not included in this analysis due to perishing under waterlogging after 5 days. The steep coastal land with high rainfall and good drainage conditions where F2507 originates from appears to have led to a lack of adaptation to waterlogging. Therefore, the lack of drainage under the waterlogging treatment appeared to be the key contributor to its death.

The remaining four red clover populations were categorized into two groups, large leaved erect populations, and small leaf prostrate populations. Both groups produced similar total dry matter using different isohydric (“saver”) or anisohydric (“spender”) coping mechanisms. The larger aboveground stature of F3594 and Relish required more energy to maintain than what was produced under stress. This significantly reduced the ability to maintain root dry mass. Whereas the smaller stature F4079 and F3474 populations reduced petiole length and plant height to maintain a high root dry mass. The performance of Quartz white clover appeared to be inflated by its superior performance under waterlogging, as it did not tolerate water deficiency well, with severe reductions in all morphological traits but similar total dry matter to the other populations. Overall, Quartz was able to maintain a larger total biomass than the red clover populations, but further research is recommended to validate its adaptive ability over both stresses. The fact that Relish did not outperform the other germplasm populations under waterlogging, identifies an area to target in future breeding programs.

This study identified key morphological and physiological responses to water stress extremes amongst red clover populations, highlighting the different coping mechanisms applied by differing morphological stature populations under water deficiency and waterlogging. Root dry mass was identified across both water treatments as a key trait, a plant’s ability to maintain root dry mass was a key contributor to its overall growth and survival. The size and number of key aboveground morphological traits such leaf number, leaf size, petiole length and plant height influenced how a plant responded to the water stress. Better understanding the effects of water stress on key plant traits will enable breeders to select important ones in breeding programs. The introduction of raw germplasm for evaluation reveals new genetics for selection and will provide the basis the development of stress-adapted cultivars.