Literature Search and Data Extraction

The literature included in the meta-analysis was identified by searching specific terms in Google Scholar and three weed science journals of Weed Science Society of America (WSSA) (Weed Technology, Weed Science, and Invasive Plant Science and Management) published before April 2021. The search term included “weed” or the common and scientific names of the top 10 most common and troublesome weeds among all broadleaf crops, fruits, and vegetable crops based on the 2019 WSSA National Weed Survey Dataset (Wychen 2019) and the top 10 most common and troublesome weeds among all grass crops, pasture, and turf from the 2020 WSSA National Weed Survey Dataset (Wychen 2020) in the title of the publication in conjunction with (“AND”) the search phrase (“water stress” OR “moisture stress” OR “moisture” OR “drought” OR “water reduction”) in separate queries yielding 2,384 total search hits. We included the most common and troublesome weeds in our search terms, because they are the most extensively studied weeds and their inclusion was intended to broaden the search criteria. A multistep screening protocol was adopted to identify relevant literature for this meta-analysis (Page and McKenzie 2021; Figure 1). For the literature to be included, it had to meet the following criteria: (1) water-stressed and comparative control (i.e., well-watered) treatments were investigated under the same experimental conditions; (2) water stress was quantitatively expressed using one of three metrics: solution osmotic potential (ψ_solution), soil moisture in terms of soil water potential (ψ_soil), or percent field capacity (studies using vague terms such as “drought” to denote water stress were excluded); (3) means for at least 1 of the 12 response variables were reported for both water-stress and control treatments, and these response variables include indices related to weed germination (germination/ emergence), weed growth characteristics (radicle/root length, plant height, leaf area, branches/tillers per plant, leaves per plant, total biomass, root biomass, shoot biomass, and root:shoot ratio), and seed production (inflorescences per plant and seeds per plant); (4) the weed was grown individually (i.e., in monoculture) and not in competition with the crop; and (5) water stress was maintained throughout the duration of the experiment.

A total of 86 relevant published papers were identified. From each selected paper, we extracted the following information (Table 1):

When ψ_solution or ψ_soil was reported in different units, units were standardized into a common unit of “MPa.” Depending on the test medium and metrics used to express water stress across studies, a solution with ψ_solution of “0 MPa” and soil with ψsoil of ∼ “–0.03 MPa” or “100% field capacity” were considered as comparative control treatments. If the information for given indices were reported over time, data were extracted from the last recorded observation. From each study, responses of different weed species (including distinct populations, biotypes, sex types, environmental occurrence, and seed sources) and at different water-stress levels were included as distinct observations in the database. The final data set had 1,196 observations from 86 articles published during 1970 through 2020 and spanned four continents (Asia, Australia, Europe, and North America).

Meta-analysis: Overall Water-Stress Effects

We used the natural logarithm of response ratios as effect sizes to calculate the overall effects of water stress on weed germination, growth characteristics, and seed production (Hedges et al. 1999).

where ln(RR) is the natural log of response ratios, X̄_WS and X̄_C are mean values of indices related to weed germination (germination/ emergence), weed growth characteristics (radicle/root length, plant height, leaf area, branches/tillers per plant, leaves per plant, total biomass, root biomass, shoot biomass, and root:shoot ratio), and seed production (inflorescences per plant and seeds per plant) for water-stressed and control treatments, respectively. Under severe water-stress conditions, weeds did not germinate or died. In such cases, values for given indices were reported as zero. Because ln(RR) cannot be calculated when any of the treatment mean values are zero, we substituted zero with the minimum possible values (for example, 0.1% germination for 0% germination, 0.1 for other growth variables such as plant height, leaf area, total biomass, etc.) (Thapa et al. 2018a).

The bulk of the studies included in the meta-analysis did not report information that denotes within-study variabilities such as standard deviation (SD), standard error (SE), or the coefficient of variation (CV). Individual effect sizes could not be weighted by sampling variances as suggested by Hedges and Olkin (1985). Therefore, we weighted individual effect sizes based on experimental replications using the following equation (Adams et al. 1997):

where w_i is the weight for ith effect size, N_WS and N_C are the number of replications for water-stressed and control treatments, respectively.

More than one effect size was calculated from studies that reported results from multiyear experiments, and that tested multiple weed populations/biotypes and multiple water-stress intensities. This could lead to dependencies among effect sizes. Therefore, we modeled various sources of dependencies in effect sizes within and across studies by creating a multilevel mixed-effects meta-analytic model in the R nlme package (Pinheiro and Bates 2022; Thapa et al. 2018a, 2018b; Van den Noortgate et al. 2013). In this model, effect sizes were considered as a fixed effect, study/year/weed biotype/common controls were nested as random effects, and w_i values were included as weighting factors. Due to lack of actual measures of sampling variances, a cluster-based robust variance estimator was used to estimate robust SEs for mean effect sizes using the clubSandwich package in R (Pustejovsky 2022). Robust SEs were used to calculate 95% confidence intervals (CIs) for weighted mean effect sizes, that is, the natural log of response ratios [ln(RR)]. The overall water-stress effect on various indices related to weed germination, growth characteristics, and seed production was considered significant when the 95% CIs did not overlap zero (P < 0.05). For ease of interpretation, the mean effect sizes and their associated 95% CIs are exponentially back-transformed to the percentage change in responses using the following equation:

where ln(RR) is the mean effect size for each index.

Figure 1.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses; Page and McKenzie 2021) flow diagram highlighting the selection procedure of 86 scientific published papers included in the meta-analysis.

Moderator or Subgroup Analysis: Effect of Weed Types, Families, and Water-Stress Intensity

A moderator analysis was conducted to determine whether or not the overall mean water-stress effects determined in this study were influenced by potential covariates. Covariates that were investigated included weed types (broadleaf vs. grass), families (Amaranthaceae, Asteraceae, Convolvulaceae, Fabaceae, Rubiaceae, and Poaceae), and the level of water stress. For this particular analysis on weed germination and seedling radicle length, we used studies in which water stress was expressed as ψ_solution, that is, studies conducted using PEG or D-mannitol solutions. Pot studies using soil as a test medium were not included due to a small number of pair-wise comparisons. For weed germination, ψ_solution is categorized into seven subgroups ranging from low to severe water stress: 0 to –0.2, –0.2 to –0.4, –0.4 to –0.6, –0.6 to –0.8, –0.8 to –1.0, –1.0 to –1.4, and <–1.4 MPa. Shrestha et al. (2018) used exorbitantly greater levels of ψ_solution (i.e., up to –5.56 MPa); therefore, it was excluded from the moderator analysis on weed germination. For seedling radicle length, ψsolution is categorized into five subgroups: 0 to –0.2, –0.2 to –0.4, –0.4 to –0.6, –0.6 to –1.0, and <–1.0 MPa. To investigate the moderating effect of water-stress intensity on indices related to weed growth characteristics and seed production, we only used pot studies that used soil as a test medium and expressed water stress in terms of “percent field capacity.” We categorized effect sizes into three subgroups based on water stress: severe water stress (<30% field capacity), moderate water stress (30% to 60% field capacity), and low water stress (>60% field capacity). Due to small number of pair-wise comparisons, we did not use any studies that expressed water stress in terms of ψ_soil in any of the moderator analyses. Similarly, the germination response of broadleaf versus grass weeds to water stress was assessed.

Separate mean effect sizes and robust SEs were calculated for each subgroup using each one as a sole covariate in the original multilevel mixed-effects meta-analytic model described earlier. To safeguard against experiment-wise type I errors, 99% CIs were calculated for the subgroup analysis. The mean water-stress effect for each subgroup was considered significant (P < 0.01) if their 99% CIs leave out zero and significantly different if their 99% CIs did not overlap with one another. A four-parameter logistic model was fit to determine the quantitative relationship between water stress (expressed as ψ_solution) and mean water stress effect on moderating variables such as germination and seedling radicle length for grass versus broadleaf weeds:

where ln(RR) is the mean effect size for each subgroup, c is the lower asymptote, d is the higher asymptote, b is the slope at the inflection point, ψ_solution is the solution osmotic potential, and ψ_solution.50 is the solution osmotic potential at the inflection point (i.e., ψ_solution that produces a response midway between c and d).

Table 1.

Summary of 86 published articles included in the meta-analysis.

Table 1.

(Continued)

Table 1.

(Continued)

Table 1.

(Continued)

Publication Bias and Sensitivity Analysis

As mentioned earlier, many studies included in this meta-analysis did not report sampling variances to create meaningful funnel plots. Therefore, publication bias was investigated indirectly by visualizing the distribution of individual effect sizes for each of the indices using density plots (Basche and DeLonge 2017; Thapa et al. 2018a). To create these density plots, we excluded the imputed effect sizes, that is, effect sizes in which weed response under water-stressed conditions was zero and was replaced with the minimum possible value. We also performed sensitivity analysis to identify studies that may have influenced results and hence, test the robustness of the overall effect size estimates obtained in this meta-analysis (Philibert et al. 2012). Overall effect sizes and their corresponding CIs for each of the indices were repeatedly calculated using the Jackknife sensitivity analysis procedure. Our approach involves rerunning the same multilevel mixed-effects multi-analytic model as described earlier, with each individual study excluded from the data set every time.

Database Description

The 86 studies included in the meta-analysis were conducted during the previous five decades (1970 through 2020) in nine countries across four continents: Asia (China, Iran, and the Philippines), Australia (Australia), Europe (Denmark, Greece, and the United Kingdom), and North America (Canada and the United States). More than one-third of the studies (79%; n = 68) were conducted in three countries: the United States (36%; n = 31), Australia (24%; n = 21), and the Philippines (19%; n = 16). China, Canada, and Iran had six, five, and four studies, respectively, whereas each European country (Denmark, Greece, and the United Kingdom) had one study.

Across all studies, a total of 102 weed species belonging to 24 families were investigated for their response to water stress (Supplementary Table  S1). Most of the studies investigated water-stress effects on broadleaf weeds (n = 62) followed by grass weeds (n = 23). Only one study by Hoveland and Buchanan (1973) investigated both broadleaf and grass weed species (Supplementary Table  S1). Most of the broadleaf weed species belonged to Asteraceae (n = 22), followed by Fabaceae (n = 9), Convolvulaceae (n = 5), Amaranthaceae (n = 4), and Rubiaceae (n = 4). Similarly, the investigated grass weed species belonged to the family Poaceae (n = 24). Among weed species, hemp sesbania [Sesbania herbacea (Mill.) McVaugh] was the most frequently investigated species in four studies, followed by junglerice [Echinochloa colona (L.) Link] and redroot pigweed (Amaranthus retroflexus L.), both of which were investigated three times. All other weed species were investigated once, except eight weed species that were investigated twice: crowfootgrass [Dactyloctenium aegyptium (L.) Willd.], green foxtail [Setaria viridis (L.) P. Beauv.], itchgrass [Rottboellia cochinchinensis (Lour.) W.D. Clayton], ivyleaf morningglory (Ipomoea hederacea Jacq.), pitted morningglory (Ipomoea lacunosa L.), ragweed parthenium (Parthenium hysterophorus L.), slender amaranth (Amaranthus viridis L.), and tall morningglory [Ipomoea purpurea (L.) Roth].

Figure 2.

Overall water-stress effects on weed germination/emergence, growth characteristics, and seed production. The vertical black dashed line represents zero effect. The black dots are overall mean effect sizes, and the black lines are 95% confidence intervals (CIs). The values in parentheses are the number of observations followed by the number of studies for each pair-wise comparison. The mean effect sizes were considered significantly different when their 95% CIs did not include zero.

Effect of Water Stress on Weed Germination

Averaged across pair-wise observations, water stress reduced weed seed germination/emergence by 90% (95% CI = –92% to –86%; Figure 2). This effect of water stress on weed germination/ emergence is likely because more than one-third of the observations (i.e., n = 276 out of 759) were exposed to severe water stress (ψ_solution >–0.6 MPa), resulting in >97% germination inhibition. Seed imbibition is required for germination, and hydration levels vary by plant species (Hegarty 1978), although understanding of these levels in weeds is limited (Pérez-Fernández et al. 2000). We observed 86% to 95% inhibition in the germination of Amaranthaceae, Asteraceae, Convolvulaceae, Fabaceae, Rubiaceae, and Poaceae weed families (Figure 3). Although differences were nonsignificant, Asteraceae was the least responsive family with 86% (99% CI = –93% to –72%) germination inhibition, while Amaranthaceae was the most responsive family with 95% (99% CI = –99% to –65%) germination inhibition due to water stress.

Plant functional groups respond differently to moisture availability (Emanuel et al. 2007; Manzoni et al. 2011), as evidenced by greater negative responses of grasses than broadleaves to water stress (Emanuel et al. 2007). Overall, germination of grass weeds was inhibited by 93% (99% CI = –97% to –83%) compared with 90% (99% CI = –93% to –84%) for broadleaf weeds (Figure 3). Mackie et al. (2019) also noted a greater impact of summer drought on grasses than forbs in their experiments across eight sites. Similarly, we observed a general trend of grass weeds being more negatively affected than broadleaf weeds across water-stress levels (Figure 4). However, water-stress effects between broadleaf and grass weeds were not significantly different due to their overlapping 99% CIs across ψ_solution subgroups.

Figure 3.

Overall water-stress effects on germination/emergence of grass and broadleaf weeds (top) and six weed families—Asteraceae, Fabaceae, Convolvulaceae, Amaranthaceae, Rubiaceae, and Poaceae (bottom). The vertical black dashed line represents zero effect. The black dots are overall mean effect sizes, and the black lines are 99% confidence intervals (CIs). The values in parentheses are the number of observations followed by the number of studies for each pair-wise comparison. The mean effect sizes were considered significantly different when their 99% CIs did not include zero.

At low water stress (i.e., ψ_solution >–0.2 MPa), the mean water-stress effect on germination of broadleaf weeds was nonsignificant (mean = –18%, 99% CI = –41% to 14%), whereas grass weeds showed a negative effect (mean = –51%, 99% CI = –76% to –0.3%). At moderate water stress, the germination of broadleaves significantly decreased by 53% (99% CI = –69% to –29%) at ψ_solution between –0.2 to –0.4 MPa and by 84% (99% CI = –92% to –70%) at ψ_solution between –0.4 to –0.6 MPa. Similarly, the germination of grass weeds decreased by 73% (99% CI = –88% to –39%) at ψ_solution between –0.2 to –0.4 MPa and by 92% (99% CI = –98% to –67%) at ψ_solution between –0.4 to –0.6 MPa. At severe water stress, germination of broadleaf and grass weeds decreased by 97% (99% CI = –99% to –93%) and 98% (99% CI = –100% to –87%), respectively, at ψ_solution between –0.6 to –0.8 MPa. With further decrease in ψ_solution above –0.8 MPa, germination of both broadleaves and grasses was completely inhibited (i.e., >99% inhibition). We further modeled the effect of water stress on weed germination by fitting a four-parameter logistic model between mean effect sizes and mean ψ_solution for each subgroup (Figure 4). The fitted model coefficients are presented in Table 2. For both broadleaf and grass weed types, weed germination decreased with a decrease in ψ_solution (Figure 4). This indicates that the adverse effects of water stress on weed germination increased with increasing water stress.

Effect of Water Stress on Weed Growth Characteristics

Averaged across pair-wise comparisons, water stress negatively affected belowground weed growth characteristics (Figure 2). Water stress, on average, decreased seedling radicle/root length by 77% (95% CI = –89% to –54%) and root biomass by 44% (95% CI = –63% to –13%). A more intense effect of water stress on seedling radicle/root length was likely due to the use of PEG or D-mannitol solutions, with 35% of the observations (n = 19 out of 54) being exposed to severe water-stress (i.e., ψ_solution >–0.6 MPa) conditions exhibiting > 97% inhibition (Figure 4). As ψ_solution decreases, water stress increases, causing seedling radicle length to decrease progressively. At low to moderate water stress of ψ_solution between 0 to –0.4 MPa, the mean decrease in seedling radicle length was not significantly different from zero. However, further decrease in ψ_solution below –0.4 MPa resulted in a reduction in seedling radicle length compared with a no water stress condition. For instance, seedling radicle length decreased by 65% (99% CI = –82% to –31%) at ψ_solution between –0.4 to –0.6 MPa, by 97% (99% CI = –99% to –84%) at ψ_solution between –0.6 to –1.0 MPa, and by 99% (99% CI = –100% to –86%) at ψsolution below –1.0 MPa. Although the mean water-stress effects for each subgroup were not significantly different from zero, root biomass decreased from a mean positive effect of 2% (99% CI = –28% to 44%) at low (i.e., soil moisture >60% field capacity) to a mean negative effect of 39% (99% CI = –77% to 60%) and 69% (99% CI = –92% to 15%) at moderate (i.e., soil moisture at 30% to 60% field capacity) and severe (i.e., soil moisture <30% field capacity) water stress, respectively (Figure 5). The results suggest that belowground weed growth characteristics were negatively impacted by water stress, and the magnitude of the effect intensified with increasing water stress.

Water stress reduced most aboveground weed growth characteristics (Figure 2). Averaged across pair-wise comparisons, water stress decreased plant height by 36% (95% CI = –54% to –11%), leaf area by 45% (95% CI = –66% to –10%), leaves per plant by 28% (95% CI = –39% to –16%), and shoot biomass by 45% (95% CI = –58% to –29%). Although not statistically different, water stress decreased branches/tillers per plant by 27% (95% CI = –49% to 4%). Results were consistent with the findings from a recent meta-analysis by Sun et al. (2020), who reported a decrease in plant growth characteristics under water stress: for instance, they observed an overall decrease in plant dry biomass by 29% due to water stress.

Results from the moderator analysis indicated that the negative effects of water stress on aboveground weed growth characteristics intensified with increasing water stress (Figure 5). As soil moisture, expressed as percent field capacity, became more deficit, we observed a progressive reduction in the mean effect sizes for indices related to aboveground weed growth characteristics. For example, the mean effect on weed shoot biomass decreased from a nonsignificant effect of –15% (99% CI = –40% to 21%) at low (i.e., soil moisture > 60% field capacity) to a significant effect of –39% (99% CI = –56% to –14%) at moderate (i.e., soil moisture at 30% to 60% field capacity) and –61% (99% CI =-73% to –43%) at severe (i.e., soil moisture < 30% field capacity) water stress. Similarly, we found a nonsignificant effect of low water stress on other aboveground weed growth characteristics, including plant height, leaf area, leaves per plant, and branches/tillers per plant. Growth indices were reduced at moderate and severe water stress: plant height by –24% (99% CI = –34% to –12%) and –37% (99% CI = –46% to –26%), leaf area by –43% (99% CI = –60% to –18%) and –44% (99% CI = –57% to –26%), leaves per plant by –30% (99% CI = –48% to –5%) and –47% (99% CI = –60% to –29%), and branches/tillers per plant by –23% (99% CI = –38% to –4%) and –52% (99% CI = –85% to 52%), respectively. Taken altogether, the adverse effects of soil moisture limitations on weed growth intensified with increasing water stress, as reported in a recent meta-analysis (Sun et al. 2020). Growth reduction is not only a direct effect of water stress but also an important adaptive mechanism (Skirycz and Inzé 2010). Plants rapidly inhibit their growth at the onset of water stress before gradually recovering and adapting to stressed conditions (Skirycz and Inzé 2010). Additionally, plants have similar multiple adaptive responses, such as generating antioxidants (Nayyar and Gupta 2006), regulating hormones (Peleg and Blumwald 2011), inducing stress proteins (Poolman et al. 2002), and improving water-use efficiency by increasing root ducts (Lee et al. 2016). Hence, although water stress will reduce weed growth, timely established weeds that can utilize soil moisture from the early onset of precipitation or soil water storage can be more competitive owing to lower resource competition (Hanson 2015).

Table 2.

Parameter estimates and SEs from a four-parameter logistic model fit to effect sizes for germination and seedling radicle length of broadleaf and grass weed species under water-stress gradients.^a

Figure 4.

The log response ratio for germination and seedling radicle length of broadleaf (green dots/line) and grass (red dots/line) weed species as a function of water-stress intensity. Water stress increased as solution osmotic potential (ψ_solution) decreased and vice versa. The subgroups for germination are 0 to –0.2, –0.2 to –0.4, –0.4 to –0.6, –0.6 to –0.8, –0.8 to –1.0, –1.0 to –1.4, and <–1.4 MPa, while the subgroups for radicle length are 0 to –0.2, –0.2 to –0.4, –0.4 to –0.6, –0.6 to –1.0, and <–1.0 MPa. Only ψ_solution-based studies were used in this analysis. For each subgroup, the solid dots and lines represent mean effect sizes and their corresponding 99% confidence intervals (CIs). The mean effect sizes were considered significantly different when their 99% CIs did not include zero. Similarly, the water-stress effects were significantly different for each subgroup and among weed types only when their 99% CIs did not overlap with one another. The fitted lines represent a four-parameter logistic regression model, and the coefficients of the models are presented in Table 2.

Averaged across pair-wise comparisons, water stress reduced total weed biomass by 61% (95% CI = –77% to –35%; Figure 2). Although not significantly different from zero, root: shoot ratio was the only index that increased under water stress (mean = 19%; 95% CI = –13% to 63%; Figure 2). A moderator analysis further indicated that the positive effect on root:shoot ratio was mostly observed when soil moisture was maintained above 60% field capacity, that is, at low water stress (mean = 27%, 99% CI = –0.2% to 61%; Figure 5). Even under moderate and severe water stress, that is, when soil moisture was below 60% field capacity, the root:shoot ratio of weeds remained unaffected. These results indicate root growth is generally less sensitive to water stress relative to shoot growth (Sharp and Davies 1989). Plants allocate a greater portion of assimilated dry matter to roots under water stress (Delfin et al. 2021; Xu et al. 2015), and the resultant increase in rooting depth allows for water extraction from deeper layers, maintaining a higher root water influx for longer durations (Chaves et al. 2002). Osmotic adjustment (Saab et al. 1992), higher soluble sugars and dry matter in roots (Xu et al. 2015), increased cell wall loosening ability (Hsiao and Xu 2000), and water stress–induced abscisic acid and ethylene (Sharp and LeNoble 2002; Spollen et al. 2000) are the primary mechanisms assuring greater root resilience under water stress relative to shoots. Considering plant adaptive mechanisms and the abilities of weeds to extract more water and tolerate water stress, weeds are thus expected to further intensify water-stressed conditions for crops (Griffin et al. 1989; Patterson and Flint 1982; Stuart et al. 1984).

Effect of Water Stress on Weed Seed Production

Water stress decreased weed seed production or fecundity. Averaged across pair-wise comparisons, water stress decreased inflorescences per plant by 48% (95% CI = –49% to –46%) and seeds per plant by 64% (95% CI = –73% to –52%) relative to the unstressed condition (Figure 2). A moderator analysis indicates that both inflorescences and seeds per plant decreased with increasing water stress (Figure 5). The inflorescences per plant decreased by 76% (99% CI = –81% to –71%) at severe water stress of <30% field capacity to 34% (99% CI = –53% to –8%) at moderate water stress of 30% to 60% field capacity, and 20% (99% CI = –38% to 2.5%) at low water stress of >30% field capacity. Similarly, the seeds per plant decreased by 88% (99% CI = –95% to –70%) at severe water stress of <30% field capacity to 50% (99% CI = –65% to –31%) at moderate water stress of 30% to 60% field capacity, and 27% (99% CI = –43% to –7%) at low water stress of >30% field capacity. These results suggest that weeds can continue to produce flowers and seeds to some extent under water-stressed conditions. When water is limited, plants often shorten their vegetative growth and accelerate to rapid flowering and seed production to attain senescence (Bernal et al. 2011; Franks et al. 2007; Sherrard and Maherali 2006).

Figure 5.

The log response ratio for weed growth characteristics (plant height, leaf area, branches/tillers per plant, leaves per plant, root biomass, shoot biomass, and root:shoot ratio) and seed production (inflorescences per plant and seeds per plant) as a function of water-stress intensity. Water stress increased as soil moisture (% field capacity) decreased and vice versa. The green and red dots represent broadleaf and grass weed species, respectively. The solid black points and the lines represent mean effect sizes and their 99% confidence intervals (CIs) for low (>60%), moderate (30%–60%), and severe (<30% field capacity) water-stress subgroups. The mean effect sizes were considered significantly different when their 99% CIs did not include zero. Similarly, the water-stress effects were significantly different for each subgroup and among weed types only when their 99% CIs did not overlap with one another.

Publication Bias and Sensitivity Analysis

The distribution of individual effect sizes for various indices related to weed germination, growth characteristics, and seed production is shown as density plots in Figure 6. All indices had narrow distributions and slightly offset from zero, indicating a negative effect, except root:shoot ratio, which showed a slightly positive effect of water stress. Nonetheless, density plots for all indices show nearly symmetrical distribution, indicating no publication bias for any of the indices considered in the meta-analysis (Light and Pillemer 1984; Sterne and Harbord 2004).

Sensitivity analysis identified a few influential studies for some of the indices investigated (Figures 7 and 8). For example, an exclusion of Zollinger and Kells (1991) from the data set increased overall effect size estimates from –36% (95% CI = –54% to –11%) to –23% (95% CI = –29% to –15%) for plant height and from –45% (95% CI = –66% to –10%) to –33% (95% CI = –46% to –17%) for leaf area. Similarly, with the exclusion of Chauhan (2013), the magnitude of overall effect size estimates increased from –27% (95% CI = –49% to 4%) to –15% (95% CI = –28% to –0.2%) for branches/tillers per plant and from –28% (95% CI = –39% to –16%) to –15% (95% CI = –28% to –0.2%) for leaves per plant. This was likely because these studies reported drastic impacts of water stress on weed growth characteristics; for example, Chauhan (2013) observed a 56% reduction in leaf area at 12.5% field capacity compared with the control (i.e., 100% field capacity). Therefore, the exclusion of these studies caused a deviation in the overall effect size estimates. In contrast, the exclusion of Chadha et al. (2019) decreased the magnitude of the overall effect size estimates for weed total biomass by 9% from –61% (95% CI = –77% to –35%) to –70% (95% CI = –72% to –68%). However, we determined that the main conclusions of this meta-analysis are robust, because (1) overall effect sizes for all other indices (germination/emergence, radicle/root length, inflorescences per plant, seeds per plant, root biomass, shoot biomass, and root:shoot ratio) were not sensitive to any given study, (2) overall effect sizes estimated using the Jackknife procedure after excluding influential studies fall within 95% CI of their original overall effect size estimates, and (3) drastic effects of severe water stress on plants are not uncommon, and these effects have caused the resultant change in magnitude of the aforementioned overall effect sizes.

Figure 6.

Density plots depicting the distribution of the individual effect sizes for all 12 response variables considered in this meta-analysis: (A) weed seed germination/emergence; (B) radicle/root length, plant height, and leaf area; (C) branches/tillers per plant, leaves per plant, inflorescences per plant, and seeds per plant; and (D) total biomass, root biomass, shoot biomass, and root:shoot ratio.

Lessons Learned, Evidence Gaps, and Future Research Considerations

The meta-analysis based on 1,196 observations from 86 studies accomplished in this study is the first to assess the integral and quantitative response of 102 weed species to water stress. This is also the first study to evaluate the holistic effect of water stress on 12 response variables associated with germination, growth, and seed production of weeds, and concurrently differentiate germination response of grasses and broadleaf weeds. We found a generally negative response of weeds to water stress, and our findings underscore and strengthen the previously held notion that water stress inhibits plant growth and performance. The germination of grass weeds might be slightly more sensitive to water stress compared with broadleaf weeds. Moreover, weed germination is completely inhibited at ψ_solution below –0.8 MPa, and a minimum of –0.09 MPa for grass weeds and –0.32 MPa for broadleaf weeds is required to inhibit their germination by half. Similarly, a minimum of –0.50 MPa is required to reduce seedling radicle length of broadleaf weeds by half. Plant height demonstrates inhibition by about one-fourth, inflorescences per plant by one-third, and seeds per plant by one-half under moderate water stress of 30% to 60% field capacity. In general, weed fecundity was found to be suppressed to a larger degree than growth morphology under water stress. For instance, weed root biomass and shoot biomass were inhibited by 61% to 69%, whereas fecundity (inflorescences per plant and seeds per plant) was inhibited by 76% to 88% under severe water stress of <30% field capacity relative to unstressed (i.e., 100% field capacity) conditions. Our findings that weeds will germinate, survive, grow, and reproduce and will continue to be competitive and problematic in managed agronomic systems, even under intense drought or water-stressed conditions, cannot be ignored. As cropping systems continue to experience extreme weather events more than at any time in the past, future studies should investigate, evaluate, and promote the adoption of multiple diverse strategies aimed at effectively managing weeds under water-limited conditions as an integral component of integrated weed management programs. This study identifies distinctive, adaptive behavior (i.e., ability to compete for water) of weeds that can inform predictions of emergence, survival rate, and possible shifts in weed communities in water-limited regions and periods. Finally, quantitative information gathered in this meta-analysis will be helpful in modeling and/or predicting multidimensional responses of weeds to water stress.

Figure 7.

Results from the sensitivity analysis depicting variations in the overall effect size estimates (mean ± 95% confidence intervals [CIs]) of water-stress effects on (A) weed germination/emergence, (B) seedling radicle/root length, (C) plant height, and (D) leaf area when a particular study is omitted from the analysis. The vertical black solid and dashed lines represent overall effect sizes (mean ± 95% CIs) with all studies included.

This current meta-analysis identified critical gaps in the existing evidence base and provides directions for reporting data standards and future research avenues:

Figure 8.

Results from the sensitivity analysis depicting variations in the overall effect size estimates (mean ± 95% confidence intervals [CIs]) of water-stress effects on (A) branches/tillers per plant, (B) leaves per plant, (C) inflorescences per plant, (D) seeds per plant, (E) total biomass, (F) root biomass, (G) shoot biomass, and (H) root:shoot ratio, when a particular study is omitted from the analysis. The vertical black solid and dashed lines represent overall effect sizes (mean ± 95% CIs) with all studies included.