Digital soil mapping using artificial neural networks

In the context of a growing demand of high‐resolution spatial soil information for environmental planning and modeling, fast and accurate prediction methods are needed to provide high‐quality digital soil maps. Thus, this study focuses on the development of a methodology based on artificial neural networks (ANN) that is able to spatially predict soil units. Within a test area in Rhineland‐Palatinate (Germany), covering an area of about 600 km², a digital soil map was predicted. Based on feed‐forward ANN with the resilient backpropagation learning algorithm, the optimal network topology was determined with one hidden layer and 15 to 30 cells depending on the soil unit to be predicted. To describe the occurrence of a soil unit and to train the ANN, 69 different terrain attributes, 53 geologic‐petrographic units, and 3 types of land use were extracted from existing maps and databases. 80% of the predicted soil units (n = 33) showed training errors (mean square error) of the ANN below 0.1, 43% were even below 0.05. Validation returned a mean accuracy of over 92% for the trained network outputs. Altogether, the presented methodology based on ANN and an extended digital terrain‐analysis approach is time‐saving and cost effective and provides remarkable results.     

利用人工神经网络以及相关地形属性绘制数字土壤地图

Detailed soil surveys involve costly and time-consuming work and require expert knowledge. Since soil surveys provide information to meet a wide range of needs, new methods are necessary to map soils quickly and accurately. In this study, multilayer perceptron artificial neural networks (ANNs) were developed to map soil units using digital elevation model (DEM) attributes. Several optimal ANNs were produced based on a number of input data and hidden units. The approach used test and validation areas to calculate the accuracy of interpolated and extrapolated data. The results showed that the system and level of soil classification employed had a direct effect on the accuracy of the results. At the lowest level, smaller errors were observed with the World Reference Base (WRB) classification criteria than the Soil Taxonomy (ST) system, but more soil classes could be predicted when using ST (7 soils in the case of ST vs. 5 with WRB). Training errors were below 11% for all the ANN models applied, while the test error (interpolation error) and validation error (extrapolation error) were as high as 50% and 70%, respectively. As expected, soil prediction using a higher level of classification presented a better overall level of accuracy. To obtain better predictions, in addition to DEM attributes, data related to landforms and/or lithology as soil-forming factors, should be used as ANN input data.     

Effect of spatial scale of soil data on estimates of soil ecosystem services: Case study in 100 km2 area in France

Over the last decade, the ecosystem services (ESs) framework has been increasingly used to support mapping and assessment studies for sustainable land management purposes. Previous analysis of practical applications has revealed the significance of the spatial scale at which input data are obtained. This issue is particularly problematic with soil data that are often unavailable or available only at coarse scales or resolutions in various part of the world. In this context, four soil-based ecosystem services, namely biomass provision, water provision, global climate regulation, and water quality regulation, are assessed using three conventional soil maps at the 1:1,000,000, 1:250,000 and 1:50,000 scales. The resulting individual and joint ES maps are then compared to examine the effects of changing the spatial scale of soil data on the ES levels and spatial patterns. ES levels are finally aggregated to landforms, land use, or administrative levels in order to try to identify the determinants of the sensitivity of ES levels to change in the scale of input soil data. Whereas the three soil maps turn out to be equally useful whenever ESs levels averaged over the whole 100 km² territory are needed, the maps at the 1:1,000,000 and 1:250,000 induced biases in the assessment of ESs levels over spatial units smaller than 100 and 10 km², respectively. The simplification of the diversity and spatial distribution of soils at the two coarsest scales indeed resulted in local differences in ES levels ranging from several 10 to several 100%. Identification of the optimal representation of soil diversity and distribution to obtain a reliable representation of ESs spatial distribution is not straightforward. The ESs sensitivity to scale effect is indeed context-specific, variable among individual ESs, and not directly or simply linked with the soil typological diversity represented in soil maps. Forested and natural lands in the study area appear particularly sensitive to soil data scales as they occupy marginal soils showing very specific ESs signatures.     

基于地表昼夜温差的平原区土壤质地模糊聚类与预测制图

The use of landscape covariates to estimate soil properties is not suitable for the areas of low relief due to the high variability of soil properties in similar topographic and vegetation conditions.A new method was implemented to map regional soil texture (in terms of sand,silt and clay contents) by hypothesizing that the change in the land surface diurnal temperature difference (DTD) is related to soil texture in case of a relatively homogeneous rainfall input.To examine this hypothesis,the DTDs from moderate resolution imagine spectroradiometer (MODIS) during a selected time period,i.e.,after a heavy rainfall between autumn harvest and autumn sowing,were classified using fuzzy-c-means (FCM) clustering.Six classes were generated,and for each class,the sand (> 0.05 mm),silt (0.002-0.05 mm) and clay (< 0.002 mm) contents at the location of maximum membership value were considered as the typical values of that class.A weighted average model was then used to digitally map soil texture.The results showed that the predicted map quite accurately reflected the regional soil variation.A validation dataset produced estimates of error for the predicted maps of sand,silt and clay contents at root mean of squared error values of 8.4%,7.8% and 2.3%,respectively,which is satisfactory in a practical context.This study thus provided a methodology that can help improve the accuracy and efficiency of soil texture mapping in plain areas using easily available data sources.     

Mapping soil pH levels across Europe: An analysis of LUCAS topsoil data using random forest kriging (RFK)

Soil pH affects food production, pollution control and ecosystem services. Mapping soil pH levels, therefore, provides policymakers with crucial information for developing sustainable soil use and management policies. In this study, we used the LUCAS 2015 TOPSOIL data to map soil pH at a European level. We used random forest kriging (RFK) to build a predictive model of spatial variability of soil pH, as well as random forest (RF) without co-kriging and boosted regression trees (BRT) modelling techniques. Model accuracy was evaluated using a ten-fold cross-validation procedure. While we found that all methods accurately predicted soil pH, the accuracy of the RFK method was best with regression performance metrics of: R² = 0.81 for pH (H₂O) and pH (CaCl₂); RMSE = 0.59 for pH (H₂O) and RMSE = 0.61 in pH (CaCl₂); MAE = 0.41 for pH (H₂O) and MAE = 0.43 in pH (CaCl₂). Dominant explanatory variables in the RF and BRT modelling were topography and remote sensing variables, respectively. The generated maps broadly depicted similar spatial patterns of soil pH, with an increasing gradient of soil pH from north to south Europe, with the highest values mainly concentrated along the Mediterranean coast. The mapping could provide spatial reference for soil pH assessment and dynamic monitoring.     

Application of digital soil mapping methods for identifying salinity management classes based on a study on coastal central China

In coastal China, there is an urgent need to increase land for agriculture. One solution is land reclamation from coastal tidelands, but soil salinization poses a problem. Thus, there is need to map saline areas and identify appropriate management strategies. One approach is the use of digital soil mapping. At the first stage, auxiliary data such as remotely sensed multispectral imagery can be used to identify areas of low agricultural productivity due to salinity. Similarly, proximal sensing instruments can provide data on the distribution of soil salinity. In this study, we first used multispectral QuickBird imagery (Bands 1–4) to provide information about crop growth and then EM38 data to indicate relative salt content using measurements of apparent soil electrical conductivity (EC_a) in the horizontal (EC_h) and vertical (EC_v) modes of operation. Second, we used a fuzzy k‐means (FKM) algorithm to identify three salinity management zones using the normalized difference vegetation index (NDVI), EC_h and EC_v/EC_h. The three identified classes were statistically different in terms of auxiliary and topsoil properties (e.g. soil organic matter) and more importantly in terms of the distribution of soil salinity (EC_e) with depth. The resultant three classes were mapped to demonstrate that remote and proximally sensed auxiliary data can be used as surrogates for identifying soil salinity management zones.     

半干旱沙区土类/亚类的遥感调查制图方法

传统土壤调查制图存在低时效性、低精度等问题。为了解决半干旱沙区土壤遥感调查制图问题,该文以科尔沁左翼后旗为例,基于野外实地调查和专家知识分析了半干旱沙区土壤类型分布特征与环境因素之间的关系,并探讨了基于多时相Landsat8 OLI影像数据的半干旱沙区土类/亚类遥感调查制图方法。结果表明:利用多时相Landsat8 OLI影像数据提取的归一化差异水体指数(modified normalized difference water index,MNDWI)、盐分指数(salt index,SI)、归一化差异湿度指数(normalized difference moisture index,NDMI)、归一化差异植被指数(normalized difference vegetation index,NDVI)等环境信息,可实现对沼泽土、盐碱土、草甸土、风沙土及其亚类等半干旱沙区主要土壤类型的遥感调查制图。应用本文提出的半干旱区土类/亚类遥感调查制图方法对科左后旗进行土壤遥感调查制图和精度验证,总体精度约为72.84%,Kappa系数为0.667 8。该方法可为半干旱沙区数字土壤调查制图提供思路和参考。     

Digital soil mapping of available water content using proximal and remotely sensed data

Two‐thirds of all irrigated agriculture in Australia is undertaken within the Murray–Darling Basin. However, climate change predictions for this region suggest rainfall will decrease. To maintain profitability, more will need to be done by irrigators with less water. In this regard, irrigators need to be aware of the spatial distribution of the available water content (AWC) in the root‐zone (i.e. 0.0–0.90 m). To reduce the cost, digital soil mapping (DSM) techniques are being used to map soil properties related to AWC (e.g. soil texture). The purpose of this study was to create a DSM of the AWC at the district scale. This is achieved by determining AWC by the difference between laboratory measured permanent wilting point (PWP) and field capacity (FC) and using pressure plate apparatus. The PWP and FC data are coupled to remote (i.e. gamma‐ray spectrometry) and proximal (i.e. EM38 and EM34) sensed data and two trend surface parameters. Using a hierarchical spatial regression (HSR), we predict PWP and FC across the areas of Warren and Trangie in the lower Macquarie valley, Australia. The reliability of the DSM of PWP and FC were compared using prediction precision (RMSE – root mean square error) and bias (ME – mean error). The best results were achieved using EM38‐v, EM34‐20, eU and eTh. The DSM map of AWC is consistent with known Pedoderms and provides a basis for agricultural water management.     

Digital mapping of cation exchange capacity using genetic programming and soil depth functions in Baneh region,Iran

This study evaluates the performances of a combination of genetic programming and soil depth functions to map the three-dimensional distribution of cation exchange capacity (CEC) in a semiarid region located in Baneh region, Iran. Using the conditioned Latin hypercube sampling method, the locations of 188 soil profiles were selected, which were then sampled and analyzed. In general, results showed that equal-area quadratic splines had the highest R², 89%, in fitting the vertical CEC distribution compared to power and logarithmic functions with R² of 81% and 84%, respectively. Our findings indicated some auxiliary variables had more influence on the prediction of CEC. Normalized difference vegetation index (NDVI) had the highest correlation with CEC in the upper two layers. However, the most important auxiliary data for prediction of CEC in 30–60 cm and 60–100 cm were topographic wetness index and profile curvature, respectively. Validation of the predictive models at each depth interval resulted in R² values ranging from 66% (0–15 cm) to 19% (60–100 cm). Overall, results indicated the topsoil can be reasonably well predicted; however, the subsoil prediction needs to be improved. We can recommend the use of the developed methodology in mapping CEC in other parts in Iran.     

基于生态系统服务和PSO-SOFM神经网络的中亚水土热资源匹配分区

水土热资源匹配度分区研究对于区域农业规划具有重要意义。中亚地区长期以来缺乏合理的水土热资源管理,已引发了一系列资源环境问题,严重威胁该地区农业生产。目前的研究也较少关注中亚水土热资源匹配分区模式。本研究利用遥感数据,通过量化4种主要生态系统服务(植被固碳、土壤保持、水源供给与涵养及生物多样性保护)的时空分布特征,结合PSO-SOFM(particleswarm optimization,PSO;self-organizing featuremap,SOFM)神经网络模型实现中亚水土热资源匹配度分区,并利用Spearman秩相关分析探索不同匹配度分区与生态环境因子的关系,应用偏相关分析确定气温和降水量对中亚地区生态系统服务的影响。结果表明,中亚生态系统服务总体呈东南高、西北低的空间格局,沿山地—绿洲—荒漠方向递减。在2000—2015年间,各类生态系统服务均有不同程度变化,其中植被固碳和土壤保持呈显著下降的面积占整个中亚的84.81%和84.82%;水源供给与涵养以及生物多样性保护服务呈显著下降的面积较少,占比分别为69.48%和19.8%,且这两种生态系统服务在个别地区有增加趋势。PSO-SOFM神经网络模型在中亚水土热资源匹配度分区中表现良好,根据生态系统服务值空间模式,中亚水土热资源匹配度可被划为5大类21个子类分区。在空间尺度,各类匹配度分区之间生态系统服务值有显著差异,降水是影响生态系统服务和匹配度高低的重要限制因子,而气温和土壤因素影响较弱;在时间尺度,降水和各生态系统服务值间呈显著正相关关系的范围更广,而气温对生态系统服务值有显著影响的区域主要集中在哈萨克斯坦北部草地—半荒漠生态敏感区、中亚荒漠生态脆弱区、中亚中部半荒漠生态敏感区以及巴特赫兹—卡拉比尔半荒漠生态敏感区等地。而在其他区域,气温和降水量并非决定生态系统服务值高低的主要因素,生态系统服务值的变化可能与土地开发利用模式有关。结合不同匹配度分区的生态地理条件,本研究可为中亚地区水土资源开发利用、农牧业发展以及生态环境保护提供有用信息。     

Nature,extent and severity of soil erosion in upland Scotland

This paper examines the nature, extent and severity of soil erosion in upland Scotland, an area subject to land management pressures typical of temperate maritime uplands. Erosion features were classified and their area measured from aerial photographs. Some 12 per cent of the upland area sampled was subject to some form of erosion, which is very similar to the percentage for Europe as a whole. The most significant erosion category was peat erosion, with 6 per cent of the area being affected; a figure which increased to 20 per cent in one sub-region. Peat erosion varied in severity, with the most severe erosion being in areas of eastern Scotland with the greatest land management pressures. Gully erosion of slopes on mineral soils was found in almost 5 per cent of the area sampled, particularly in sub-regions with large amplitude of relief. Debris flow/cone features and screes were less extensive and largely found at higher altitudes. Footpath erosion was mapped in popular mountain areas, but overall the mean length of eroded footpaths was less than that of large gullies. There was little evidence of spatial linkages between erosion of mineral soils and land management at the scale of the survey. Separating the roles of upland management and extreme rainfall events in the inception of erosion remains a key issue in the study of accelerated erosion in temperate maritime upland areas. Some guidance for management is presented.     

Soil S availability in upland pastures of NE Scotland: relationship of extractable soil S and soil respiration to soil and site characteristics

Abstract. The mean extractable sulphur (S) concentration in 315 upland topsoil samples collected in 1988/89 from beneath pasture in NE Scotland was 13 μg S g⁻¹ (range 2–77 μg S g⁻¹). More than two thirds of the samples had S concentrations less than that acceptable for productive soils. Continued decreases in atmospheric S inputs may have increased this proportion subsequently. The analysis of herbage S also indicated that two-thirds of the samples were below 0.2% S. A 'respirometric index', namely CO₂ produced during cellulose decomposition without added S as a percentage of that produced with added S, was significantly less than 100% in a quarter of the soils. Results of three different extraction procedures suggested that sulphate in the soils was present mainly as free plus adsorbed rather than precipitated forms. Soil extraction identified a significant non-sulphate S fraction, presumably organic S. The variability in extractable S stemmed from a combination of geographical, depositional and local site and soil factors. Extractable S was significantly correlated with soil organic matter content and inversely with soil pH and together these factors explained 37% of the variability. While significant differences in mean concentrations between geographical area, soil association and drainage status were evident, no trends could be observed between the major soil subgroups or with altitude.     

干旱内陆河流域生态系统服务空间权衡与协同作用分析

为了测度生态系统服务的空间权衡/协同关系,该文以干旱内陆河流域典型地区嘉峪关-酒泉地区为研究区,对其2000年和2010年食物供给、碳储存、水源涵养和土壤保持4种生态系统服务物质量进行定量测算.利用相关分析法和空间热点制图等方法,分析了生态系统服务时空变化,从县域尺度和区域尺度上分析了生态系统服务之间的权衡协同关系,识别了某种服务的物质量或价值量极高的热点区.结果表明:研究区单位面积食物供给的高值区分布在中东部的肃州区及嘉峪关市,碳储存表现为东部及南部较高的分布格局;土壤保持在2000年和2010年变化不大,呈现出从西到东逐渐增大的分布趋势;单位面积水源涵养值由2000年的0.40mm/(m2·a)提高到2010年的0.99mm/(m2·a).县域尺度上,各种生态系统服务之间大多为协同关系,其中食物-碳和水源-碳之间的协同程度较高;区域尺度上,食物供给和土壤保持、水源涵养和土壤保持生态系统服务间存在此消彼长的权衡关系.4种生态系统服务的值均未超过各自平均值的0类服务区和只有1种生态系统服务的值超过其所对应平均值的1类服务区分别占总面积的58.50%和25.20%.该研究结果可为制定差别化的区域发展与生态保护双赢政策提供科学参考.     

基于高分5号影像的东北典型黑土区土壤分类

高精度的土壤分类及制图结果有助于更好地制定土地环境保护和土地资源利用策略。为探究星载高光谱影像实现区域尺度高精度土壤分类及制图的可能性,该研究获取东北黑土区拜泉县、明水县共计4幅高分5号（GF-5）星载高光谱遥感影像。首先,将原始反射率数据（Original Reflectance,OR）进行包络线去除处理获得去包络线数据（Continuum Removal,CR）;其次,对OR和CR进行主成分分析（Principal Component Analysis,PCA）处理,分别得到反射率主成分信息（OR-PCA）和去包络线主成分信息（CR-PCA）,并在OR-PCA和CR-PCA的基础上结合地形因子（Terrain,TA）。最后,OR、CR、OR-PCA、CR-PCA、OR-PCA-TA、CR-PCA-TA分别作为输入量结合随机森林分类模型,进行土壤分类并实现数字土壤制图。结果表明：1）包络线去除法可有效地提高星载高光谱土壤分类精度,与OR相比,CR的总精度提高了5.48%,Kappa系数提高了0.12。2）PCA可有效地降低高光谱数据的冗余性,提高模型的运算效率以及分类精度;与CR作为输入量相比,CR-PCA的土壤分类总精度提高了3.67%,Kappa系数提高了0.02。3）TA的引入显著提升了土壤分类精度,以CR-PCA-TA作为输入量的土壤分类精度最高,总精度为81.61%,Kappa系数为0.72,实现了高精度的土壤分类模型及土壤制图。研究结果可为大范围、高精度的土壤分类及制图提供新的思路。     

基于特征选择和GA-BP神经网络的多源遥感农田土壤水分反演

土壤水分是影响水文、生态和气候等环境过程的重要参数,而微波遥感是农田地表土壤水分测量的重要手段之一.针对微波遥感反演农田地表土壤水分受植被覆盖影响较大的问题,该研究提出了一种基于特征选择和GA-BP神经网络(Genetic Algorithm-Back Propagation neural network)的多源遥感农...     

Monitoring of sustainable land management using remotely sensed vegetation cover and variable tolerable soil erosion targets across New South Wales,Australia

The monitoring of land management practices is vital for the protection of soil resource and the environment. Here, we present a new approach to monitor sustainable land management using widely available remotely sensed data, such as MODIS fractional vegetation cover data. The method is based on the concept of maintaining sufficient vegetation cover to prevent hillslope water erosion beyond tolerable soil erosion targets. The targets were based on long-term natural erosion rates plus a small constant and are spatially and temporally variable, not static as in most reported studies to date. Where vegetation cover is more than that required to prevent nontolerable erosion under normal conditions for that calendar month, the site (pixel) is deemed to be managed sustainably. Monthly indices are then combined to form a yearly sustainable land management index (SLMI), presented as raster maps with a spatial resolution of 100 m. We explored this new method through case studies over New South Wales (NSW), Australia, over the period 2010 to 2021, with a particular examination of 2020. Results were further stratified by land uses and natural resource management regions, which revealed useful data and trends. The method is offered as an example of the potential use of readily available vegetation cover data to quantitatively assess and monitor levels of sustainable land management across landscapes. We believe it overcomes the limitations of previous methods to monitor vegetation cover and land management from remote-sensed data alone. Other users are encouraged to adapt the broad approach to meet local requirements.     

Mapping soil erosion risk in Rondônia,Brazilian Amazonia: using RUSLE,remote sensing and GIS

This article discusses research in which the authors applied the Revised Universal Soil Loss Equation (RUSLE), remote sensing, and geographical information system (GIS) to the maping of soil erosion risk in Brazilian Amazonia. Soil map and soil survey data were used to develop the soil erodibility factor (K), and a digital elevation model image was used to generate the topographic factor (LS). The cover‐management factor (C) was developed based on vegetation, shade, and soil fraction images derived from spectral mixture analysis of a Landsat Enhanced Thematic Mapper Plus image. Assuming the same climatic conditions and no support practice in the study area, the rainfall–runoff erosivity (R) and the support practice (P) factors were not used. The majority of the study area has K values of less than 0·2, LS values of less than 2·5, and C values of less than 0·25. A soil erosion risk map with five classes (very low, low, medium, medium‐high, and high) was produced based on the simplified RUSLE within the GIS environment, and was linked to land use and land cover (LULC) image to explore relationships between soil erosion risk and LULC distribution. The results indicate that most successional and mature forests are in very low and low erosion risk areas, while agroforestry and pasture are usually associated with medium to high risk areas. This research implies that remote sensing and GIS provide promising tools for evaluating and mapping soil erosion risk in Amazonia. Copyright © 2004 John Wiley & Sons, Ltd.     

Probing soil physical and biological resilience data from a broad sampling of arable farms in Scotland

Physical and biological soil stabilities (i.e. resistance and resilience) were measured on a range of arable farms across eastern Scotland under a range of management practices, with the objective of using a geographically restricted set of soils under similar land use to detect any underlying associations between soil stability, management factors and soil properties. Data were analysed using a combination of a stepwise fixed effects model selection within a linear mixed‐model framework (LMM) and neural network analysis using a Kohonen self‐organising map (KSOM). In general, physical and biological measures of stability were associated with both physical and biological soil properties, particularly bulk density, water retention characteristics, soil carbon and bacterial community structure. A strength of KSOM is its ability to fit more flexible models than the linear relationships of LMM. However, a weakness is that it does not have the ability of LMM to model the sampling design, which is likely to lead to overstating statistical significance. Consequently, KSOM identified more significant associations between soil properties and stability than LMM, while the latter identified significant associations at the between‐farm level. The high‐level land management decisions of farm type (conventional, organic, integrated), crop type or underlying soil type were not associated with stability at this regional scale, thus indicating that the effects of different management practices between farms were overridden by the soil properties on each farm. Management decisions on improving soil stability therefore need to be taken at the individual field scale.