Skip to main content

How and to what extent have degraded grasslands recovered after ecological restoration in China: A meta-analysis?

Cite this dataset

Su, Jishuai et al. (2021). How and to what extent have degraded grasslands recovered after ecological restoration in China: A meta-analysis? [Dataset]. Dryad.


  1. A range of measures for restoration have been tested on degraded grasslands in China. However, the factors controlling recovery to natural grassland conditions and the overall effectiveness of restoration remain unclear.
  2. We synthesized data from 365 studies for 103 grassland sites, with 21 variables related to plant community and soil properties. We used meta-analysis to evaluate the effectiveness of ecological restoration and the influence of restoration approach, grassland type, soil depth, climatic conditions, and the duration of restoration programmes.
  3. Ecological restoration showed larger enhancements on plant community structure and biomass compared with plant diversity and soil properties, and the recovery of soil properties was substantially slower than the recovery of vegetation. Specifically, aboveground biomass showed faster recovery than belowground biomass, but plant species richness did not achieve full recovery to natural conditions. Soil water content, soil carbon and nitrogen content, and soil carbon storage did not recover fully to natural conditions, although they were improved to a greater extent in topsoil than in subsoil. In contrast, soil nutrient availability and microbial biomass generally recovered to the level of natural grasslands, and their restoration efforts tended to increase with soil depth, possibly due to the more enhanced belowground biomass in subsoil. There was no general significant difference in restoration efforts between passive and active restorations. The duration of restoration programmes, grassland type, and climatic conditions modulated the effectiveness of grassland restoration.
  4. Synthesis and applications. Our study showed that vegetation recovered faster than soil in degraded grasslands after ecological restoration. Given that passive and active restoration regimes achieved similar recovery, we recommend that passive restoration as the most cost-effective option for restoring degraded grasslands where a spontaneous recovery process is possible. We also propose that a restoration network is initiated to facilitate the development of grassland recovery strategies, informed by restoration approaches, grassland types, and climatic conditions.


Data compilation of restoration studies in degraded grasslands

The ISI Web of Science ( and the China National Knowledge Infrastructure (CNKI) institution were searched for peer-reviewed literatures before 6 July 2020, using the following search term combinations: (grassland OR steppe OR meadow OR pasture OR rangeland) AND (restor* OR rehabilitat* OR regenerat* OR establish*) AND (China). From these studies, we collected information about the effects of restoration measures on plant and soil variables in degraded grasslands in China.

We selected peer-reviewed literatures based on five criteria. (1) The studies included field measurements, thus we excluded studies derived from laboratory incubation, remote sensing research, and model simulation. (2) The study sites were in degraded grasslands in China. (3) Details about the restoration regime (duration and methods) were explicitly mentioned in the literature. (4) The studies included details about sample size and mean of target variables from restored and degraded grasslands. (5) The studies included at least one of the following variables (which formed part of this meta-analysis); namely changes in plant community biomass, structure, diversity, and physical, chemical and biological properties of soil in the grasslands after restoration.

We found 208 relevant publications about studies conducted in 103 degraded grassland sites. We collected 21 variables related to plant community and soil properties. These plant variables included aboveground biomass (AGB), litter biomass (LitterB), belowground biomass (BGB), the ratio of BGB to AGB (BAratio), plant height, plant coverage, plant density, plant species richness, and the Shannon-Wiener diversity index (Shannon). Soil variables included soil bulk density (BD), soil water content (SWC), pH, soil organic C content (SOC), soil total nitrogen content (TN), soil total phosphorus content (TP), soil available nitrogen content (AN), soil available phosphorus content (AP), soil organic C storage (SOCS), soil total nitrogen storage (NS), soil microbial biomass C content (MBC), and soil microbial biomass nitrogen (MBN).

The WebPlotDigitizer 4.1 ( was used to digitize and extract raw data from the studies. We recorded the following information about the study sites: detailed name, latitude, longitude, mean annual temperature (MAT), mean annual precipitation (MAP), grassland type, restoration measures employed, duration of the restoration, and soil depths of BGB and soil variables. We extracted climatic characteristics from the global database at if MAT and/or MAP data were missing in the literature, based on latitude and longitude coordinates.


Data compilation from observational studies in natural grasslands

We searched the ISI Web of Science and the CNKI institution for published literatures about natural grasslands before 12 November 2020, using the site name and grassland type provided. We aimed to evaluate the degree of recovery of degraded grassland by comparing with natural (undisturbed) grasslands in the same study site. In this meta-analysis, natural grasslands were classified into three groups, which included undisturbed and intact grasslands, reference grasslands for degradation research, and reference grasslands as a control group for global change experiments (i.e. N deposition, global warming, precipitation changes).

Three criteria were used to select studies about natural grasslands from the literatures. (1) Studies had been conducted in the field. (2) Studies included observations of the natural state of the grassland. (3) The degraded grassland type was consistent with the natural grasslands type.

We extracted data related to the 21 variables described above. We also collected information about grassland type, sampling area, soil depth, and soil variables, aiming to match observations of restoration studies within the same site. Given the spatiotemporal variability of natural grasslands, we only included variables in our study if there were at least five observations within one site to allow for the calculation of mean values. Using these filters, we selected 157 publications related to natural grasslands for further analysis.


Data analysis

We used a natural log-transformed response ratio (RR) to assess the effects of grassland restoration and calculated the corresponding variance based the sample size.

The RR was calculated as follows (Equation 1):

                                                      RR = lnXtXc                                                                  (1)

The variance of the log RR was calculated as follows (Equation 2):

                                           varRR = Nt+NcNt Nc                                                                  (2)

where X and N represent the mean and the sample size of observations, respectively. ‘t’ and ‘c’ represent the treatment and control group, respectively. When the control group referred to degraded grasslands, RR was expressed as ln (Res/Deg), for the RR between restored and degraded grasslands. When the control group referred to natural grasslands, the RR was expressed as ln (Res/Nat), for the RR between restored and natural grasslands.

The weighted mean of the RR (RR++) for each variable was analysed using the “” function (Viechtbauer, 2010). Since many studies have more than one RR, we treated the “study” as a random factor. The effects of grassland restoration were considered significant if the 95% confidence intervals of RR++ did not overlap with zero.

Percent changes of variables were calculated as follows (Equation 3):

                                             (eRR++-1)×100%                                                                  (3)

Restoration studies included in the meta-analysis had different restoration approaches related to grassland type, soil depth, climatic conditions, and restoration duration. We treated restoration type, grassland type, and soil depth as categorical moderators to evaluate their effects on the responses of variables to grassland restoration. A QM test was conducted to estimate the significance of differences in the RRs between different levels of categorical moderators. Specifically, we classified restoration type into two groups: passive restoration and active restoration. In this meta-analysis, passive restoration was regarded as the spontaneous recovery occurring in degraded grasslands or abandoned old fields once disturbance ceased. Active restoration indicated human-assisted restoration measures, such as fertilization, irrigation, seeding, and shallow ploughing of degraded grasslands.

Grasslands were classified into five groups: alpine meadows, alpine steppe, temperate meadows, temperate steppe, and desert grasslands. Few observations were reported in unclassified grasslands, and this category was thus omitted in the further analysis of grassland type. Soil depth was divided into categories of 0-10 cm, 10-20 cm, 20-40 cm, and >40 cm. Furthermore, meta-regressions weighted by the inverse of variance were conducted to determine the dependence of the RRs of variables on climatic conditions and duration of restoration processes. Bivariate relationships among RRs of different variables were determined using the Pearson correlation analyses. All the statistical analyses were conducted in R (R Development Core Team, 2020).

Usage notes

This dataset contains basic study information from 365 studies for 103 grassland sites, with 21 variables related to plant community and soil properties.


Ministry of Science and Technology of the People's Republic of China, Award: 2017YFA0604702

National Natural Science Foundation of China, Award: 31630010

Science and Technology Transformation Project of Inner Mongolia Autonomous Region, Award: 2020CG0055