Wei Shangguan

SoilGrids250m: Global gridded soil information based on machine learning

This paper describes the technical development and accuracy assessment of the most recent and improved version of the SoilGrids system at 250m resolution (June 2016 update). SoilGrids provides global predictions for standard numeric soil properties (organic carbon, bulk density, Cation Exchange Capacity (CEC), pH, soil texture fractions and coarse fragments) at seven standard depths (0, 5, 15, 30, 60, 100 and 200 cm), in addition to predictions of depth to bedrock and distribution of soil classes based on the World Reference Base (WRB) and USDA classification systems (ca. 280 raster layers in total). Predictions were based on ca. 150,000 soil profiles used for training and a stack of 158 remote sensing-based soil covariates (primarily derived from MODIS land products, SRTM DEM derivatives, climatic images and global landform and lithology maps), which were used to fit an ensemble of machine learning methods—random forest and gradient boosting and/or multinomial logistic regression—as implemented in the R packages ranger, xgboost, nnet and caret. The results of 10–fold cross-validation show that the ensemble models explain between 56% (coarse fragments) and 83% (pH) of variation with an overall average of 61%. Improvements in the relative accuracy considering the amount of variation explained, in comparison to the previous version of SoilGrids at 1 km spatial resolution, range from 60 to 230%. Improvements can be attributed to: (1) the use of machine learning instead of linear regression, (2) to considerable investments in preparing finer resolution covariate layers and (3) to insertion of additional soil profiles. Further development of SoilGrids could include refinement of methods to incorporate input uncertainties and derivation of posterior probability distributions (per pixel), and further automation of spatial modeling so that soil maps can be generated for potentially hundreds of soil variables. Another area of future research is the development of methods for multiscale merging of SoilGrids predictions with local and/or national gridded soil products (e.g. up to 50 m spatial resolution) so that increasingly more accurate, complete and consistent global soil information can be produced. SoilGrids are available under the Open Data Base License.

A global soil data set for earth system modeling

We developed a comprehensive, gridded Global Soil Dataset for use in Earth System Models (GSDE) and other applications. The GSDE provides soil information, such as soil particle-size distribution, organic carbon, and nutrients, and quality control information in terms of confidence level at 30″ × 30″ horizontal resolution and for eight vertical layers to a depth of 2.3 m. The GSDE is based on the Soil Map of the World and various regional and national soil databases, including soil attribute data and soil maps. We used a standardized data structure and data processing procedures to harmonize the data collected from various sources. We then used a soil type linkage method (i.e., taxotransfer rules) and a polygon linkage method to derive the spatial distribution of the soil properties. To aggregate the attributes of different compositions of a mapping unit, we used three mapping approaches: the area-weighting method, the dominant soil type method, and the dominant binned soil attribute method. The data set can also be aggregated to a lower resolution. In this paper, we only show the vertical and horizontal variations of sand, silt and clay contents, bulk density, and soil organic carbon as examples of the GSDE. The GSDE estimates of global soil organic carbon stock to the depths of 2.3, 1, and 0.3 m are 1922.7, 1455.4, and 720.1 Gt, respectively. This newly developed data set provides more accurate soil information and represents a step forward to advance earth system modeling.

Development of a China Dataset of Soil Hydraulic Parameters Using Pedotransfer Functions for Land Surface Modeling

AbstractThe objective of this study is to develop a dataset of the soil hydraulic parameters associated with two empirical soil functions (i.e., a water retention curve and hydraulic conductivity) using multiple pedotransfer functions (PTFs). The dataset is designed specifically for regional land surface modeling for China. The authors selected 5 PTFs to derive the parameters in the Clapp and Hornberger functions and the van Genuchten and Mualem functions and 10 PTFs for soil water contents at capillary pressures of 33 and 1500 kPa. The inputs into the PTFs include soil particle size distribution, bulk density, and soil organic matter. The dataset provides 12 estimated parameters and their associated statistical values. The dataset is available at a 30 × 30 arc second geographical spatial resolution and with seven vertical layers to the depth of 1.38 m. The dataset has several distinct advantages even though the accuracy is unknown for lack of in situ and regional measurements. First, this dataset utilize...

A 3D Canopy Radiative Transfer Model for Global Climate Modeling: Description, Validation, and Application

AbstractThe process of solar radiative transfer at the land surface is important to energy, water, and carbon balance, especially for vegetated areas. Currently the most commonly used two-stream model considers the plant functional types (PFTs) within a grid to be independent of each other and their leaves to be horizontally homogeneous. This assumption is unrealistic in most cases. To consider canopy three-dimensional (3D) structural effects, a new framework of 3D canopy radiative transfer model was developed and validated by numerical simulations and shows a good agreement. A comparison with the two-stream model in the offline Community Land Model (CLM4.0) shows that an increase of canopy absorption mainly happens with sparse vegetation or with multilayer canopies with a large sun zenith angle θsun and is due to increases of the ground and sky shadows and of the optical pathlength because of the shadow overlapping between bushes and canopy layers. A decrease of canopy absorption occurs in densely vegeta...

Age‐dependent forest carbon sink: Estimation via inverse modeling

Forests have been recognized to sequester a substantial amount of carbon (C) from the atmosphere. However, considerable uncertainty remains regarding the magnitude and time course of the C sink. Revealing the intrinsic relationship between forest age and C sink is crucial for reducing uncertainties in prediction of forest C sink potential. In this study, we developed a stepwise data assimilation approach to combine a process-based Terrestrial ECOsystem Regional model, observations from multiple sources, and stochastic sampling to inversely estimate carbon cycle parameters including carbon sink at different forest ages for evergreen needle-leaved forests in China. The new approach is effective to estimate age-dependent parameter of maximal light-use efficiency (R2 = 0.99) and, accordingly, can quantify a relationship between forest age and the vegetation and soil C sinks. The estimated ecosystem C sink increases rapidly with age, peaks at 0.451 kg C m−2 yr−1 at age 22 years (ranging from 0.421 to 0.465 kg C m−2 yr−1), and gradually decreases thereafter. The dynamic patterns of C sinks in vegetation and soil are significantly different. C sink in vegetation first increases rapidly with age and then decreases. C sink in soil, however, increases continuously with age; it acts as a C source when the age is less than 20 years, after which it acts as a sink. For the evergreen needle-leaved forest, the highest C sink efficiency (i.e., C sink per unit net primary productivity) is approximately 60%, with age between 11 and 43 years. Overall, the inverse estimation of carbon cycle parameters can make reasonable estimates of age-dependent C sequestration in forests.

Soil diversity as affected by land use in China: consequences for soil protection.

Rapid land-use change in recent decades in China and its impact on terrestrial biodiversity have been widely studied, particularly at local and regional scales. However, the effect of land-use change on the diversity of soils that support the terrestrial biological system has rarely been studied. Here, we report the first effort to assess the impact of land-use change on soil diversity for the entire nation of China. Soil diversity and land-use effects were analyzed spatially in grids and provinces. The land-use effects on different soils were uneven. Anthropogenic soils occupied approximately 12% of the total soil area, which had already replaced the original natural soils. About 7.5% of the natural soil classes in China were in danger of substantial loss, due to the disturbance of agriculture and construction. More than 80% of the endangered soils were unprotected due to the overlook of soil diversity. The protection of soil diversity should be integrated into future conservation activities.

Particle-Size Distribution Models for the Conversion of Chinese Data to FAO/USDA System

We investigated eleven particle-size distribution (PSD) models to determine the appropriate models for describing the PSDs of 16349 Chinese soil samples. These data are based on three soil texture classification schemes, including one ISSS (International Society of Soil Science) scheme with four data points and two Katschinskis schemes with five and six data points, respectively. The adjusted coefficient of determination r 2, Akaikes information criterion (AIC), and geometric mean error ratio (GMER) were used to evaluate the model performance. The soil data were converted to the USDA (United States Department of Agriculture) standard using PSD models and the fractal concept. The performance of PSD models was affected by soil texture and classification of fraction schemes. The performance of PSD models also varied with clay content of soils. The Anderson, Fredlund, modified logistic growth, Skaggs, and Weilbull models were the best.

Evaluating common land model energy fluxes using FLUXNET data

Given the crucial role of land surface processes in global and regional climates, there is a pressing need to test and verify the performance of land surface models via comparisons to observations. In this study, the eddy covariance measurements from 20 FLUXNET sites spanning more than 100 site-years were utilized to evaluate the performance of the Common Land Model (CoLM) over different vegetation types in various climate zones. A decomposition method was employed to separate both the observed and simulated energy fluxes, i.e., the sensible heat flux, latent heat flux, net radiation, and ground heat flux, at three timescales ranging from stepwise (30 min) to monthly. A comparison between the simulations and observations indicated that CoLM produced satisfactory simulations of all four energy fluxes, although the different indexes did not exhibit consistent results among the different fluxes. A strong agreement between the simulations and observations was found for the seasonal cycles at the 20 sites, whereas CoLM underestimated the latent heat flux at the sites with distinct dry and wet seasons, which might be associated with its weakness in simulating soil water during the dry season. CoLM cannot explicitly simulate the midday depression of leaf gas exchange, which may explain why CoLM also has a maximum diurnal bias at noon in the summer. Of the eight selected vegetation types analyzed, CoLM performs best for evergreen broadleaf forests and worst for croplands and wetlands.摘 要陆面模式描述陆地表面与大气之间物质, 能量和动量交换过程, 因其可以为大气环流模式提供准确的下垫面条件而对全球气候变化研究非常重要. 随着人们对天气预报和气候预测精度的需求越来越高, 用观测资料验证和改进陆面模式也逐渐成为提高模式模拟能力的一种重要手段. 本研究利用覆盖不同气候区不同植被类型的 20 个 FLUXNET 站点观测数据检验通用陆面模式(Common Land Model, CoLM)的能量平衡模拟情况. 在对比分析观测和模拟的感热, 潜热, 净辐射和地表热通量时, 采用分解方法将每个观测值或模拟值分解为月平均, 日平均残差和逐时残差三个部分. 综合相关系数(R), 均方根误差(RMSE)和标准化误差(Nbias)对不同时间尺度和不同植被类型的能量通量分析, 结果发现 CoLM 能够很好地模拟 4 个地表能量通量, 其中模拟净辐射最好, 模拟潜热通量好于感热通量; 模式能够很好地模拟能量通量的季节变化, 但在有明显干湿季的站点低估潜热通量, 这可能与模式对干季土壤水的模拟存在缺陷有关; CoLM 在夏季中午产生较大模拟误差, 这很可能是因为 CoLM 不能准确模拟光合午睡现象; 就不同的植被类型而言, CoLM 在所选的 8 种植被类型中的常绿针叶林表现最好.

Incorporating root hydraulic redistribution and compensatory water uptake in the Common Land Model: Effects on site-level and global land modeling

Treatment of plant water uptake through the roots remains a significant issue in land surface models. Most current land surface models calculate the root water uptake (RWU) by extracting soil water in different soil layers based on the relative soil water availability and the root fraction of each layer within the rooting zone. This approach is also used as the default in the Common Land Model (CoLM). This approach often significantly underestimates plant transpiration during dry periods. Therefore, more realistic RWU functions are needed in land surface models. In this study, the modified CoLM with root hydraulic redistribution (HR) and compensatory water uptake (CWU) was evaluated against the CoLM with the default approach by comparing the observed and simulated latent and sensible heat fluxes observed from three sites that experience seasonal drought over the measured periods. We found that the CoLM using the default RWU significantly underestimated latent heat fluxes and overestimated the sensible heat fluxes over dry periods, whereas those biases were significantly reduced by the CoLM with HR and CWU functions. We also ran global offline simulations using the revised CoLM to evaluate the performance of these alternative RWU functions on the global scale. Compared with the estimated latent heat fluxes from the FLUXNET-MTE model product, CoLM with HR and CWU functions significantly improved the estimated latent heat fluxes over the Amazon, Southern Africa and Central Asia during their dry seasons. Therefore, we recommend the implementation of HR and CWU in land surface models.

Reexamination and further development of two‐stream canopy radiative transfer models for global land modeling

Four representative two-stream canopy radiative transfer models were examined and intercompared using the same configuration. Based on the comparison results, two modifications were introduced to the widely used Dickinson-Sellers model and then incorporated into the Community Land Model (CLM4.5). The modified model was tested against Monte-Carlo simulations and produced significant improvements in the simulated canopy transmittance and albedo values. In direct comparison with MODIS albedo data, the modified model shows good performance over most snow/ice-free vegetated areas, especially for regions that are covered by dense canopy. The modified model shows seasonally dependent behavior mainly in the near-infrared band. Thus, the improvements are not present in all seasons. Large biases are still noticeable in sparsely vegetated areas, in particular for the snow/ice covered regions, that is possibly related to the model, the land surface input data, or even the observations themselves. Further studies focusing on the impact of the seasonal changes in leaf optical properties, the parameterizations for snow/ice covered regions and the case of sparsely vegetated areas, are recommended.