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.

提高我们对土壤微生物群落的功能和过程及其与物理化学土壤环境的相互作用的理解需要大量及时且具有成本效益的土壤数据，而常规实验室分析方法很难获得这些数据。便携式可见近红外 (VNIR) 和中红外 (MIR) 仪器的土壤光谱可以通过促进生物和非生物土壤信息的快速获取来填补这一空白。

在这项研究中，我们评估了 VNIR 和 MIR 光谱在长期草地生物多样性-生态系统功能实验中分析土壤理化和微生物特性的能力。在耶拿实验（德国耶拿）收集土壤样品，并在现场潮湿条件下使用便携式 VNIR 和 MIR 光谱仪进行测量，以确定它们在现场数据收集和分析方面的潜力。校准光谱模型的参考数据是通过常规分析方法获得的，包括微生物生物标志物的 PLFA 提取。我们进一步收集了预处理土壤（干燥和细磨）的参考 VNIR 和 MIR 数据，以评估现场测量对光谱校准的预期影响。

与预处理样品的 VNIR 数据相比，MIR 光谱可以更准确地估计土壤理化和微生物特性。对于田间条件下的土壤，MIR 校准对理化特性更准确，但 VNIR 数据对微生物特性的估计要好得多。在每种情况下，组合 VNIR/MIR 估计对所有土壤特性都获得了最准确的估计结果。

土壤理化性质可以从 VNIR/MIR 数据中以高精度 (R ²  = 0.72-0.99) 对预处理的土壤样品进行估计，而土壤微生物性质的结果则较为温和 (R ² = 0.66–0.72）。在田间潮湿的土壤上，有机碳和无机碳的估计精度显着下降（ΔRMSE = 52–72%），土壤质地略有改善（ΔRMSE = 4–7%），微生物特性略有下降（ΔRMSE = 4–9%） ）。来自田间条件下土壤的 VNIR/MIR 估计值足够准确，可以检测实验性植物处理对有机碳以及细菌和真菌生物量的影响。我们进一步发现，土壤微生物特性的光谱估计主要是通过与光谱活性土壤成分的间接相关性来实现的，即土壤微生物特性与土壤物理化学环境之间的关联。

我们的研究结果突出了 VNIR 和 MIR 光谱分析土壤物理化学环境的能力，包括潜在的现场数据收集和现场条件下的土壤分析，并表明当土壤理化性质影响土壤时，VNIR/MIR 数据可以估计土壤微生物性质。土壤微生物群落分布。