No abstract

Abstract. Traditional laboratory methods for acquiring soil information remain important for assessing key soil properties, soil functions and ecosystem services over space and time. Infrared spectroscopic modeling can link and massively scale up these methods for many soil characteristics in a cost-effective and timely manner. In Switzerland, only 10 % to 15 % of agricultural soils have been mapped sufficiently to serve spatial decision support systems, presenting an urgent need for rapid quantitative soil characterization. The current Swiss soil spectral library (SSL; n = 4374) in the mid-infrared range includes soil samples from the Biodiversity Monitoring Program (BDM), arranged in a regularly spaced grid across Switzerland, and temporally resolved data from the Swiss Soil Monitoring Network (NABO). Given that less than 2 % of the samples in the SSL originate from organic soils, we aimed to develop both an efficient calibration sampling scheme and accurate modeling strategy to estimate the soil carbon (SC) contents of heterogeneous samples between 0 and 2 m depth from 26 locations within two drained peatland regions (School of Agricultural, Forest and Food Sciences (HAFL) data set; n = 116). The focus was on minimizing the need for new reference analyses by efficiently mining the spectral information of the SSL. We used partial least square regressions (PLSRs), together with five repetitions of a location-grouped, 10-fold cross-validation, to predict SC ranging from 1 % to 52 % in the local HAFL data set. We compared the validation performance of different calibration schemes involving local models (1), models using the entire SSL combined with local samples (2), commonly referred to as spiking, and subsets of local and SSL samples optimized for the peatland target sites using the resampling local (RS-LOCAL) algorithm (3). Using local and RS-LOCAL calibrations with at least five local samples, we achieved similar validation results for predictions of SC up to 52 % (R2 = 0.93 to 0.97; bias = -0.07 to 1.65; root mean square error (RMSE) = 2.71 % to 3.89 % total carbon; ratio of performance to deviation (RPD) = 3.38 to 4.86; and ratio of performance to interquartile range (RPIQ) = 4.93 to 7.09). However, calibrations using RS-LOCAL only required five or 10 local samples for very accurate models (RMSE = 3.16 % and 2.71 % total carbon, respectively), while purely local calibrations required 50 samples for similarly accurate results (RMSE < 3 % total carbon). Of the three approaches, the entire SSL spiked with local samples for model calibration led to validations with the lowest performance in terms of R2, bias, RMSE, RPD and RPIQ. Hence, we show that a simple and comprehensible modeling approach, using RS-LOCAL together with a SSL, is an efficient and accurate strategy when using infrared spectroscopy. It decreases field and laboratory work, the bias of SSL spiking approaches and the uncertainty of local models. If adequately mined, the information in the SSL is sufficient to predict SC in new and independent study regions, even if the local soil characteristics are very different from the ones in the SSL. This will help to efficiently scale up the acquisition of quantitative soil information over space and time.

Abstract. Information on soils' composition and physical, chemical and biological properties is paramount to elucidate agroecosystem functioning in space and over time. For this purpose, we developed a national Swiss soil spectral library (SSL; n=4374) in the mid-infrared (mid-IR), calibrating 16 properties from legacy measurements on soils from the Swiss Biodiversity Monitoring program (BDM; n=3778; 1094 sites) and the Swiss long-term Soil Monitoring Network (NABO; n=596; 71 sites). General models were trained with the interpretable rule-based learner CUBIST, testing combinations of {5,10,20,50, and 100} ensembles of rules (committees) and {2, 5, 7, and 9} nearest neighbors used for local averaging with repeated 10-fold cross-validation grouped by location. To evaluate the information in spectra to facilitate long-term soil monitoring at a plot level, we conducted 71 model transfers for the NABO sites to induce locally relevant information from the SSL, using the data-driven sample selection method RS-LOCAL. In total, 10 soil properties were estimated with discrimination capacity suitable for screening (R2≥0.72; ratio of performance to interquartile distance (RPIQ) ≥ 2.0), out of which total carbon (C), organic C (OC), total nitrogen (N), pH and clay showed accuracy eligible for accurate diagnostics (R2>0.8; RPIQ ≥ 3.0). CUBIST and the spectra estimated total C accurately with the root mean square error (RMSE) = 8.4 g kg−1 and the RPIQ = 4.3, while the measured range was 1–583 g kg−1 and OC with RMSE = 9.3 g kg−1 and RPIQ = 3.4 (measured range 0–583 g kg−1). Compared to the general statistical learning approach, the local transfer approach – using two respective training samples – on average reduced the RMSE of total C per site fourfold. We found that the selected SSL subsets were highly dissimilar compared to validation samples, in terms of both their spectral input space and the measured values. This suggests that data-driven selection with RS-LOCAL leverages chemical diversity in composition rather than similarity. Our results suggest that mid-IR soil estimates were sufficiently accurate to support many soil applications that require a large volume of input data, such as precision agriculture, soil C accounting and monitoring and digital soil mapping. This SSL can be updated continuously, for example, with samples from deeper profiles and organic soils, so that the measurement of key soil properties becomes even more accurate and efficient in the near future.