Field-based soil-texture estimates could replace laboratory analysis
Introduction
Soil texture influences nearly all soils processes, either directly or indirectly. The texture can be determined with common protocols in the laboratory by particle size analyses but also in the field: Trained soil scientists can do this by feel, assessing the granularity, the mealiness and the cohesiveness of the sample. This is done by testing the plasticity, shininess, detectability of sand grains visually and by deforming the sample between one's fingers (Sponagel et al., 2005) (‘Fingerprobe’). The laboratory methods are standard methods to estimate soil particle size distribution. However, they are time-consuming and therefore more expensive than a soil texture-by-feel test; in particular due to pre-treatments of the samples to remove organic matter and salts. Even though there is traditional knowledge on the quality of field based texture estimates, a systematic evaluation of the magnitude of the uncertainty of field based texture estimates has never been conducted and published. In the course of the German Agricultural Soil Inventory 3896 soil samples were taken from 728 soil profiles and analyzed for their texture both in the field and in the laboratory, so that the accuracy of the field-based estimates could be assessed.
Earlier studies on the variability of field- and laboratory-based texture analyses focus on the percentage of samples for which both analyses yield the same texture class. Foss et al. (1975) for example, compared the results of field textures to those of laboratory analysis. For 598 samples taken from soils in the Coastal Plain of Maryland, USA, 50% were assigned the right texture class during the field texture rating according to the USDA textural classification (using 12 different classes). The accuracy was the highest for the texture classes sand, clay, sandy loam and silt loam, which the authors attribute to the larger portion of the textural triangle that is taken up by these classes. In cases where the texture class was not estimated correctly, high amounts of coarse fragments, free iron, organic matter or fine or coarse sands were often present in the sample. Other analyses of the accuracy of field based texture estimations of the USDA soil texture classes by Post et al. (1986); Rawls and Pachepsky (2002) and Pachepsky et al. (2006) showed that the overall texture class was correctly assigned in 46%, 39.8% and 28.4% of all cases in the three studies respectively. However, for most applications, such as soil models and ecological studies, texture classes are not required as input variable, but mass fraction of sand, silt and clay.
So far there are no published studies on the accuracy of field based texture class estimates to derive the texture mass fractions, thus the soil particle size distribution. Soil particle size distributions with the mass fractions of clay, silt and sand are basic parameters to characterize soils, and are used as input variables in most soil models. While earlier studies focus on the percentage of correctly assigned texture classes in the field, for modeling purposes the accuracy of the estimate is more important than the attribution of the correct texture class. Thus, the aim of this study is to assess the variance of the field based texture class estimates and their precision in determining a soil's particle size distribution.
Section snippets
Materials and methods
In the course of the German Agricultural Soil Inventory 3896 soil samples were taken at 728 sites throughout Germany in the Federal states Lower Saxony, Rhineland Palatinate, Mecklenburg Western Pomerania, Saarland, Bremen and Hamburg between 2011 and 2014 (Bach et al., 2011). These samples were taken by eight soil scientists in five to seven different depth increments between 0 cm and 200 cm depth from a soil pit. Only mineral soil samples with soil organic matter content below 30% were included
Accuracy of field-based estimates
The mean differences (RMSD) between field based estimation and laboratory measurement of the soil textures were 87 ± 1.26 g kg− 1 (mean ± standard error) for the sand content, 83 ± 1.29 g kg− 1 for the silt content, and 40 ± 0.81 g kg− 1 for the clay content in a given sample. As the samples contained on average 610, 280 and 110 g kg− 1 of sand, silt and clay, respectively, the relative error of texture-by-feel was 14%, 30% and 36% for estimates of the sand, silt and clay content.
The bias is the mean systematic
Conclusions
It has been shown that the soil texture of a given sample can be estimated with relatively high precision in the field with the manual texture-by-feel method. Between 43 and 48% of the deviation of texture-by-feel estimates could be attributed to the fact that only texture classes and not fractions were estimated in the field. The accuracy could be even better if the mass fractions were to be estimated directly and not just the soil texture class, as this would diminish the part of the
References (9)
- et al.
The German agricultural soil inventory: sampling design for a representative assessment of soil organic carbon stocks
Procedia Environ. Sci.
(2011) - et al.
Hydropedology and pedotransfer functions
Geoderma
(2006) - et al.
Testing the accuracy of field textures
Soil Sci. Soc. Am. Proc.
(1975) Extension of nakagawa & schielzeth's R2GLMM to random slopes models
Methods Ecol. Evol.
(2014)
Cited by (49)
Potential impacts of seasonal and altitudinal changes on enzymatic peat decomposition in the High Andean Paramo region of Ecuador
2023, Science of the Total EnvironmentPost-LGM environments and foragers on the move: New data from the lower Altmühl Valley (Franconian Jura, SE Germany)
2022, Journal of Human Evolution