A Global Metadata of the Influence of Cover Crops on Key Soil Hydraulic Properties

1. Summary

Cover crops are plants grown after cash crop harvest primarily to protect and improve soil health and water resources [1,2,3]. Cover crops represent an incremental step towards diversified crop rotations. Furthermore, CCs in combination with conservation tillage (i.e., minimal or no soil disturbance) represent a premier strategy for advancing full-stack soil health management principles in our cropping systems. For instance, CCs incorporate the remaining three of four soil health principles (i.e., maximizing soil cover, biodiversity, and the presence of living roots). Besides improving soil health and enhancing systems’ resilience, CCs provide several other agronomic and ecological services, some of which include weed suppression, soil moisture conservation, erosion control, soil N scavenging and reduced nitrate leaching losses, biological N fixation and supply, improvement in soil structural and hydraulic properties, and yield stability [4,5,6,7,8,9,10].

Among the multiple ecosystem services provisioned by CCs, improvements in soil hydraulic properties are of particular interest to enhance the long-term sustainability of our cropping systems. This is because these soil properties are key soil health indicators governing soil water movement, retention, and availability in our agricultural fields [11,12,13]. It is partly through the improvements in soil physical environment, CCs can buffer against weather extremities reducing inter-annual variability and improving yield stability over years [4,13]. The CC-associated improvements in soil hydraulic properties can be ascribed to multiple inter-connected mechanisms or pathways as illustrated in Figure 1.

Globally, numerous field studies have been conducted by independent researchers to investigate the effects of CCs on multiple specific soil hydraulic properties. However, individual field studies are limited to their specific environmental and management context, and therefore, the findings are both mixed and inconclusive. Studies have reported both positive and negative or neutral responses of CCs to a set of soil hydraulic properties. For example, while a body of evidence suggests that CCs increase the infiltration rate [14,15], some studies reported no change [16,17] or even a decrease in infiltration rate [18] with CCs compared to no-CC control. Such inconsistencies have created greater uncertainty about the generalizable effects of CCs on soil hydraulic properties and functioning.

Meta-analysis offers a powerful approach to quantitatively synthesize and contextualize all available relevant articles investigating a specific research question. With that, we conducted a meta-analysis [19] to quantify the overall effects of CCs on six key soil hydraulic properties, namely total porosity, infiltration rate, saturated hydraulic conductivity (Ksat), water retention at field capacity and permanent wilting points, and available water holding capacity (AWHC). We further investigated how soil, climate, and management influence the magnitude and direction of overall CC effects on soil hydraulic properties. The dataset presented here is the result of this meta-analysis and integrates data from individual studies investigating CC effects on soil hydraulic properties worldwide. This dataset provides a standardized and transparent resource for testing future research hypothesis, updating meta-analysis, and supporting informed management and policy decisions on CC adoption among farmers.

2. Data Description

The metadata consists of 1007 rows of data extracted from 146 peer-reviewed publications comparing CC effects on at least one of the six soil hydraulic properties considered in this meta-analysis. These soil hydraulic properties include total porosity, infiltration rate, Ksat, water retention at field capacity and permanent wilting points, and AWHC. Each row in the metadata represents a pairwise comparison between CC vs. no-CC control treatment groups which includes treatment means, standard deviations, and number of replications for both groups. Each pairwise observation is further associated with a broad set of ancillary variables which include bibliographic, geographic, soil, climate, and management (Table 1). The geographic variables (e.g., latitude and longitude) were included to fill the missing soil and climate data using publicly available databases (e.g., SSURGO and ISRIC SoilGrids for soil; PRISM, WorldClim, and Köppen-Grieger Classification for climate). Similarly, soil, climate, and management were included as either categorical or continuous variables. This facilitates moderator and meta-regression analysis to assess how soil, climate, and management shape the overall CC response on each of the individual soil hydraulic properties. Description of column names and abbreviations used in the metadata is detailed in Table 1.

The geographical distribution of the studies included in the metadata is shown in Figure 2. For all soil hydraulic properties, the majority of the studies included in this metadata originated from North America followed by Europe, Asia, and South America. Studies from both Africa and Oceania continents were poorly represented in the metadata. Studies were evenly distributed across all soil textural classes or groups.

The box and whisker plots presented in Figure 3 depicts the distributions of responses for CC vs. no-CC groups. We observed considerable variations in the observed response across studies for all soil hydraulic properties. In most cases, the observed response was slightly higher in CC as compared to no-CC controls. This indicates that the CC-associated improvements in soil hydraulic properties are highly consistent and have been observed globally across studies.

Figure 4 shows the regression analyses between CC vs. no-CC groups for all six soil hydraulic properties. The majority portion of the fitted regression lines were above the 1:1 line across all panels, suggesting that soil hydraulic properties under CCs were universally higher than those under no-CC control groups. However, it should be noted that the slopes of the fitted regression lines were different. This indicates that CCs influence certain soil hydraulic property more strongly than the others.

We calculated individual effect sizes (i.e., log of the response ratios: $\mathrm{L}\mathrm{R}\mathrm{R}$) to determine the magnitude and direction of CC effects on soil hydraulic properties in each row in the metadata. For each soil hydraulic property, $\mathrm{L}\mathrm{R}\mathrm{R}\mathrm{s}$ were symmetrically distributed across both positive, negative, and neutral values indicating no evidence for publication bias. We further conducted Pearson correlation analysis among individual effect sizes for all six soil hydraulic properties (Figure 5). We found that multiple of these $\mathrm{L}\mathrm{R}\mathrm{R}\mathrm{s}$ were highly correlated suggesting that the CC-associated improvements in soil hydraulic properties are inter-connected and linked to each other. For instance, the $\mathrm{L}\mathrm{R}\mathrm{R}$ for total porosity was positively correlated with the $\mathrm{L}\mathrm{R}\mathrm{R}$ for Ksat, water retention at field capacity, water retention at permanent wilting points, and AWHC. The $\mathrm{L}\mathrm{R}\mathrm{R}$ for infiltration rate was positively correlated to the $\mathrm{L}\mathrm{R}\mathrm{R}$ for Ksat and AWHC. In addition, the $\mathrm{L}\mathrm{R}\mathrm{R}$ for field capacity was positively correlated with $\mathrm{L}\mathrm{R}\mathrm{R}$ for permanent wilting points and AWHC.

3. METHODS

3.1. Literature search and screening:

Data was collected through a systematic search and screening of peer-reviewed articles investigating CC effects on soil hydraulic properties worldwide. For this, we followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Figure 6). An initial search was conducted based on topic words (i.e., titles, abstracts, and keywords) in the Thomson Reuters ISI Web of Science database using the search terms as listed in Table 2. The search resulted in 726 articles published through 5th October 2024.

All articles were passed through the two-step screening to assess their relevance for inclusion in the metadata. This included an abstract followed by full-text screening. First screening based on abstracts resulted in a total of 268 relevant articles. Second screening based on full texts was conducted to determine if these articles passed through the pre-defined eligibility and inclusion criteria listed in Table 3. Full-text screening resulted in a total of 127 eligible articles to be included in the final metadata. Further, we found six additional articles from the Soil Health Database published by Jian et al. [20], which were not indexed in our initial search via the ISI Web of Science database. In addition, we screened references of the existing meta-analyses and systematic review articles to ensure the inclusion of any other remaining articles that were not already identified. As a result, we obtained thirteen additional articles: nine from a previous meta-analysis by Basche and DeLonge [21] and four from a review article by Haruna et al. [22]. Altogether, our final metadata included data from 146 peer-reviewed publications.

3.2. Data extraction and standardization:

We extracted soil hydrological responses and associated metadata from the studies. As response data, we recorded treatment means and measures of variability i.e., standard deviation (SD), standard error (SE), coefficient of variation (CV), or least significant difference (LSD) for all the soil hydraulic properties considered, if available. The response means were converted into standard units: total porosity (cm³ cm⁻³), infiltration rate (cm h⁻¹), Ksat (cm d⁻¹), and all water retention characteristics (i.e., field capacity, permanent wilting point, and available water holding capacity) (cm³ cm⁻³). Similarly, reported measures of variability other than SD were converted into SD using the following equations.

$$\mathrm{S}\mathrm{D}=\mathrm{S}\mathrm{E}\times \sqrt{n}$$

(1)

$$\mathrm{S}\mathrm{D}=\mathrm{C}\mathrm{V}\times \mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}$$

(2)

$$\mathrm{S}\mathrm{E}=\left({\displaystyle \frac{\mathrm{L}\mathrm{S}\mathrm{D}}{{\mathrm{t}}_{(0.975,\mathrm{n})\sqrt{2bn}}}}\right)$$

(3)

where ‘t’ is the t-test value, ‘n’ is the sample size, and ‘b’ is the number of blocks or replications.

When the data were available for multiple soil depths, we recorded data for uppermost soil depth. Similarly, when multiple time-series measurements for a response were reported per site-year, we considered only the latest data available for the cropping cycle. In this dataset, total porosity was calculated as the summation of micro-, meso-, and macro-porosity when studies reported them separately. Similarly, we considered volumetric moisture content at water potential between –10 and –33 kPa as the field capacity, and that at -1500kPa as the permanent wilting point. When AWHC was not explicitly reported, we calculated it as the difference between the reported water retention at the field capacity and the permanent wilting point.

The associated metadata for each pair-wise observation in the repository included bibliographic, geographic, soil, climate, and management information, as listed and detailed in Table 1. When soil and climate data were not reported in original publications, we extracted them from publicly available data products using the site’s latitudinal and longitudinal coordinates. For US locations, missing soil and climate data were obtained from the Soil Survey Geographic Database (SSURGO) API and PRISM Climate Group, respectively. For non-US locations, missing soil and climate data were retrieved from the ISRIC SoilGrids Rest API and WorldClim v 2.1 Climate Data, respectively. Following data extraction, categorical moderator variables were grouped into pre-defined factor levels or classes (Table 4). This aids in data harmonization among individual studies, thereby facilitating moderator analysis in a meta-analysis.

3.3. Effect size calculation:

We further calculated the effect size of CCs for each response variable, i.e., total porosity, infiltration rate, Ksat, water retention at field capacity and permanent wilting points, and AWHC. Effect sizes were calculated for each pair-wise observation as the natural logarithm of the response ratios, as described in Equation 4.

$$\mathrm{L}\mathrm{R}\mathrm{R}=ln\left({\displaystyle \frac{{\overline{X}}_{\mathrm{C}\mathrm{C}}}{{\overline{X}}_{\mathrm{n}\mathrm{o}-\mathrm{C}\mathrm{C}}}}\right)$$

(4)

where ‘$\mathrm{L}\mathrm{R}\mathrm{R}$’ is the natural logarithm of the response ratio and represents individual effect sizes for each pair-wise observation in the metadata; ‘${\overline{X}}_{\mathrm{C}\mathrm{C}}$’ and ‘${\overline{X}}_{\mathrm{n}\mathrm{o}-\mathrm{C}\mathrm{C}}$’ are the mean values of the response variables for CC and no-CC groups, respectively. Positive values of $\mathrm{L}\mathrm{R}\mathrm{R}$ indicate improvements in soil hydraulic properties under CCs as compared to no-CC controls and vice-versa.

4. User Notes:

4.1. Value of the data:

• The dataset provides a comprehensive synthesis of the effects of CCs on key soil hydraulic properties, namely total porosity, infiltration rate, saturated hydraulic conductivity (Ksat), water retention at field capacity and permanent wilting point, and available water holding capacity (AWHC).
• Continuous data are converted into standardized units; categorical variables are clearly defined and grouped. This facilitates easy reuse of data for meta-analysis and modeling purposes.
• The dataset aids in identifying knowledge gaps and guiding future research areas (e.g., CC studies investigating soil hydraulic properties are under-represented in dry and tropical regions).
• The dataset can be used to foster cross-disciplinary research collaboration in soil science, agronomy, hydrology, and climate change adaptation, and can be used to make informed management and policy decisions regarding cover cropping.

4.2. Limitations:

• Depending on the response variables, a measure of variability (e.g., SD) was reported in the original publications in only 10-29% of the pair-wise observations included in this metadata. To impute the missing SDs, we first estimate the bootstrapped mean coefficient of variation (CV) based on the reported data and then multiplied it by their respective treatment means in each pairwise observation. Our sensitivity analysis indicated that the bootstrapping-based imputing approach applied in this metadata to estimate missing SDs was robust.
• The dataset contains only a limited number of pairwise observations for certain factor levels or classes within a specific categorical moderator variable. For example, data from dry and tropical climate zones as well as residue removal category were underrepresented in the dataset. This results in huge uncertainty in the observed CC response, thereby restricting our ability to make robust inference in these cases. However, this also points to direction for future research avenues.
• In some cases, the dataset does not contain data on all moderator variables considered in this repository. This is because information regarding these moderator variables was not always reported in the original publications. For example, only a limited number of pairwise observations in the dataset were associated with data on CC biomass at the time of termination. Despite limited reporting, we found CC biomass at termination as the most important moderator regulating the overall CC response on multiple soil hydraulic properties. This point to specific data caveats that should be considered while collecting in-field measurements and ultimately publishing findings in similar research areas.