VNIR and MIR spectroscopy of PLFA-derived soil microbial properties and associated soil physicochemical characteristics in an experimental plant diversity gradient

Highlights

• VNIR/MIR allows accurate calibration of soil physicochemical properties.

• PLFA-derived microbial biomasses can be estimated with fair accuracy.

• PLFA estimates mediated by correlations with spectrally active soil constituents.

• Soil state (dry vs. field-moist) strongly impacts spectral correlation patterns.

• Soil spectroscopy could detect plant diversity effects on microbial biomasses.

Abstract

Improving our understanding of the functions and processes of soil microbial communities and their interactions with the physicochemical soil environment requires large amounts of timely and cost-efficient soil data, which is difficult to obtain with routine laboratory-analytical methods. Soil spectroscopy with portable visible-to-near infrared (VNIR) and mid-infrared (MIR) instruments can fill this gap by facilitating the rapid acquisition of biotic and abiotic soil information.

In this study, we evaluated the capabilities of VNIR and MIR spectroscopy to analyze soil physicochemical and microbial properties in a long-term grassland biodiversity-ecosystem functioning experiment. Soil samples were collected at the Jena Experiment (Jena, Germany) and measured with portable VNIR and MIR spectrometers in field-moist condition to determine their potential for on-site data collection and analysis. Reference data to calibrate spectroscopic models were acquired with routine analytical methods, including PLFA extractions of microbial biomarkers. We further collected reference VNIR and MIR data on pre-treated soils (dried and finely ground) to assess the anticipated impact of field measurements on spectroscopic calibrations.

MIR spectra allowed more accurate estimates of soil physicochemical and microbial properties than VNIR data on pre-treated samples. For soils in field condition, MIR calibrations were more accurate for physicochemical properties, but VNIR data gave significantly better estimates of microbial properties. Combined VNIR/MIR estimates achieved the most accurate estimation results for all soil properties in each case.

Soil physicochemical properties could be estimated from VNIR/MIR data with high accuracy (R² = 0.72–0.99) on pre-treated soil samples, whereas the results for soil microbial properties were more moderate (R² = 0.66–0.72). On field-moist soils, estimation accuracies decreased notably for organic and inorganic carbon (ΔRMSE = 52–72%), improved slightly for soil texture (ΔRMSE = 4–7%) and decreased slightly for microbial properties (ΔRMSE = 4–9%). The VNIR/MIR estimates derived from soils in field condition were sufficiently accurate to detect experimental plant treatment effects on organic carbon, as well as bacterial and fungal biomass. We further found that spectroscopic estimates of soil microbial properties were primarily enabled through indirect correlations with spectrally active soil constituents, i.e., associations between soil microbial properties and the physicochemical soil environment.

Our findings highlight the capacity of VNIR and MIR spectroscopy to analyze the physicochemical soil environment, including potential on-site data collection and analysis on soils in field condition, and indicate that VNIR/MIR data can estimate soil microbial properties when soil physicochemical properties shape the distribution of soil microbial communities.

Introduction

Soil microorganisms play a crucial role in the maintenance of soil ecosystem functioning, health and fertility through the decomposition of plant material, the biogeochemical cycling of nutrients, and the stabilization of soil aggregates (Chaparro et al., 2012; Bardgett and van der Putten, 2014; Fierer, 2017). The soil microbiome, in turn, is influenced by above- and belowground inputs linked to the composition of plant communities (Lange et al., 2015; Dassen et al., 2017) and the abiotic soil environment, characterized by various soil physicochemical properties, including soil pH, organic carbon quality and quantity, soil moisture, nitrogen and phosphorus, and soil texture and structure (Fierer, 2017; Xia et al., 2020). At larger spatial scales, soil microbial diversity and abundance is primarily driven by environmental factors such as climate, topography and land cover (Xue et al., 2018). Accordingly, soil microbial properties can exhibit spatial variation at multiple nested scales (Ettema and Wardle, 2002), with microbial community structure exhibiting as much variation within a single local soil sample as there is between different biomes thousands of kilometres apart (Eilers et al., 2012; Ramirez et al., 2014). Analyzing this variation may prove essential in further understanding soil microbial-plant interactions (Classen et al., 2015), larger-scale biogeographic patterns of microbial diversity (Xue et al., 2018; Yang et al., 2019) or the response of soil microbial communities to climate change (Jansson and Hofmockel, 2020).

Efficient soil policy and management require detailed information on the distribution and diversity of soil microbial properties; however, such data is often not available at appropriate spatial and temporal scales (Guerra et al., 2020, 2021). Current routine laboratory approaches (e.g., wet chemical analyses, fatty acid profiling) may be too laborious or costly to obtain physicochemical and microbial soil data at the necessary or desired spatial and temporal sampling density for such analyses (McBratney et al., 2006; Viscarra Rossel et al., 2011). Soil diffuse reflectance spectroscopy in the visible/near-infrared (VNIR) and mid-infrared (MIR) can complement routine analytical methods by providing a faster and cheaper alternative method for collecting quantitative soil data (Nocita et al., 2015; Seybold et al., 2019). Diffuse reflectance spectroscopy is sensitive to the presence and abundance of organic and inorganic molecule bonds in the near- and mid-infrared and electronic transitions in the visible electromagnetic range (Hunt, 1977; Nguyen et al., 1991). Through the interactions of electromagnetic radiation with the soil matrix, the diffuse reflectance spectrum provides extensive information on the chemical composition of the soil (Parikh et al., 2014). The quantitative determination of soil properties with VNIR and MIR spectroscopy requires chemometric calibrations, usually with partial least squares regression (PLSR), to link the measured reflectance signal with the soil property of interest (Stenberg et al., 2010). Soil properties at additional sampling locations within the calibration domain can then be estimated from VNIR/MIR reflectance measurements without further conventional analyses.

Soil physicochemical properties (e.g., organic carbon (OC), nitrogen (N), soil carbonates (inorganic carbon, IC), clay minerals and sand (quartz) content) can generally be estimated with high accuracy as they are directly linked to several relevant absorption bands in the VNIR and MIR regions (Viscarra Rossel et al., 2006; Kuang et al., 2012; Soriano-Disla et al., 2014). In contrast, VNIR/MIR calibrations of soil microbial and biological properties have been discussed more controversially (Soriano-Disla et al., 2014; Ludwig et al., 2015). As microbial biomass in mineral soils represents only about ∼5% of total soil organic matter, it seems unlikely to observe a directly related spectral signal or pattern in the VNIR or MIR (Soriano-Disla et al., 2014). Some studies, although significantly fewer than for soil physicochemical properties, have nevertheless shown that (micro)biological soil properties can be estimated spectroscopically with moderate to high accuracy (Soriano-Disla et al., 2014; Ludwig et al., 2015). These include microbial biomasses derived from PLFAs (Rinnan and Rinnan, 2007; Zornoza et al., 2008), the fungal biomarker ergosterol and microbial soil carbon (Terhoeven-Urselmans et al., 2008; Heinze et al., 2013; Vohland et al., 2017), and soil microbial biomass from 16S rRNA gene quantification (Rasche et al., 2013). VNIR/MIR estimation models have been hypothesized to rely predominantly on indirect correlations between soil biological and physicochemical properties that can be captured through a direct spectral response of, e.g., total organic matter or soil texture parameters (Zornoza et al., 2008; Ludwig et al., 2015). In this context, recently published studies have shown that much of the variation in soil bacterial abundance and diversity at different scales can be modelled by VNIR and MIR soil reflectance through its capability to characterize the soil habitat in its overall mineral and organic composition (Yang et al., 2019; Ricketts et al., 2020).

Against this background, the recent introduction of portable, high-performance MIR spectrometers (Forrester et al., 2015; Soriano-Disla et al., 2017; Hutengs et al., 2018), in conjunction with established portable VNIR instruments (Stevens et al., 2006; Kusumo et al., 2008; Terhoeven-Urselmans et al., 2008; Kuang and Mouazen, 2011), opens the opportunity for fast and cost-effective VNIR/MIR analysis of soil physicochemical and microbial properties at the local scale, e.g., in the framework of globally distributed ecological experiments or soil monitoring. First field studies with MIR measurements on field-moist, untreated soil samples have confirmed the potential for on-site analyses of soil OC (Hutengs et al., 2019) and particle size distribution (Janik et al., 2020), with handheld MIR instruments potentially allowing more accurate OC estimates than VNIR instruments (Hutengs et al., 2019). The application of portable MIR instruments to evaluate more soil physicochemical and microbial properties, alone or together with VNIR instruments, thus merits further investigation.

In this study, we examined the potential of VNIR/MIR reflectance spectroscopy with portable instruments to analyze soil microbial properties coupled with soil physicochemical characteristics in the framework of a long-term grassland biodiversity-ecosystem functioning (BEF) experiment. We aimed to address two key issues as a fundamental prerequisite for the possible integration of VNIR/MIR analyses with portable instruments in soil microbiological and ecological research, including (i) the capability of VNIR/MIR calibrations to estimate PLFA-derived soil microbial properties and soil physicochemical characteristics in field-moist condition; (ii) whether recently introduced handheld MIR instruments bring an additional benefit for the analyses of the aforementioned soil properties, compared to VNIR instruments alone. In addition, we explored the predictive mechanisms that underlie VNIR/MIR models of soil microbial properties – indirect, i.e., mediated by spectrally active soil constituents such as OC and soil texture, vs. signals directly linked to soil microbial biomass – as these have important implications for VNIR/MIR model robustness and the ability to generalize across soil types and environmental conditions. Moreover, we aimed to explore if VNIR/MIR reflectance spectroscopy is sufficiently sensitive to detect local plant diversity and community effects on soil microbial properties.