Soil inventory data is urgently needed for agricultural management, geochemical baseline mapping, and environmental monitoring, but laboratory soil analysis is labor-intensive and time-consuming. Therefore, spectroscopic instruments have been increasingly used to collect soil information. Once calibrated, many soil properties can be measured or estimated through a single scan.

My Masters and PhD projects focused on developing and evaluating soil spectroscopic methods (visible near-infrared – vis-NIR and portable X-ray fluorescence – pXRF) and sensor data fusion methods to estimate soil physiochemical properties at the field to farm scales (Zhang et al., 2017; Zhang and Hartemink, 2019). With an increasing nationwide mid-infrared (MIR) spectral dataset, my Post-doc project developed soil MIR libraries for accurate assessment of soil microbial community compositions (PLFA) and bacterial abundance and diversity (16S rRNA) and studied their variations across ecoclimatic domains in the USA (Zhang et al., 2022). Moreover, MIR spectra contain many signals related to diagnostic soil horizons, so I also developed a MIR-based method to objectively classify soil orders and horizons (Zhang et al., 2021).

By collaborating with Dr. ‪Gafur Gözükara, we evaluated geochemical signature of loess and terra rossa horizons in Wisconsin soils and used vis-NIR and pXRF data to distinguish them with a decision tree method (Gozukara et al., 2021), and evaluated the predictability of various soil properties from vis-NIR and pXRF in the loess soils (Gozukara et al., 2022), and predicted soil properties and soil quality index using vis-NIR and pXRF data in an Inceptisol (Gozukara et al., 2022). By collaborating with Dr. Viachesla Adamchuk, we investigated the in-situ prediction of soil properties from the MIR spectrometer (Ji et al., 2016).

I currently lead a USDA-NRCS funded project (2022-2025) aiming at creating a nationwide MIR web platform for NRCS staff and soil scientists to rapidly and automatically assess soil health conditions from MIR spectra. More results will be shared in the near future.

With the enrichment of soil data from spectral estimates and advances in spatial statistics and machine learning, soil maps can be generated with a higher accuracy and finer resolution for precision agriculture and soil process modeling, termed as "digital soil mapping".

In my MS research, I combined spectra-derived soil properties with geophysical instruments (gamma-ray radiation, electromagnetic induction, digital elevation model) to map soil physiochemical properties across an 11-ha agricultural field in 3D (0 to 1 m depth at every 10-cm interval) using geostatistics and machine learning methods (Zhang et al., 2020). In my PhD research, to quantify the soil variability in a slightly larger farm (330-ha) and a highly heterogeneous landscape, I evaluated the data fusion of different remote sensing data (hyperspectral imaging, Sentinel-1, Lidar DEM) to map surface soil properties (0–10 cm) (Zhang et al., 2021). Optimized sampling design can reduce the sampling costs, so I compared commonly used sampling designs in terms of their spatial and feature space coverage and the accuracy of 3D soil mapping (Zhang et al., 2022). As part of my PhD research, I particularly quantified short-range variation of soil texture and total carbon in the 330-ha farm and investigate its vertical and horizontal distribution for different geomorphic units (unglaciated parts, outwash plain, end moraine, and a former lacustrine bed) (Zhang and Hartemink, 2021).

Additionally, Dr. Asim Biswas and I completed a review on sampling designs for validating digital soil maps (Biswas and Zhang, 2018). By collaborating with Dr. Jingyi Huang, we estimated and mapped soil organic carbon (SOC) stock changes over 150 years in Wisconsin and investigated the effects of climate and land use change (Huang et al., 2019). I also worked with a big team on developing the global 100-m soil moisture maps (Huang et al., 2020).

With the support of Dr. Marvin Beatty Fellowship, I have worked on soil variability (particularly soil thickness, carbonate, carbon fractions) across the contiguous USA (CONUS) and investigated the different environmental and human driving factors.

It is essential to quantify the temporal variability and soil change and understand the responses of soil formation and processes to climate change and human activities for soil process modeling and prediction and sustainable soil management.

My Post-doc research studied a vital soil property (i.e., soil thickness) and used long-term (67 years) observations of soil profiles to investigate its spatial and temporal distribution across the US and its natural and human causes (Zhang et al., in prep.). The effects of topsoil thickness on ecosystem productivity and its resilience to climate change was also evaluated under different ecosystems and climatic regions across the US (Zhang et al., in prep.).

Additionally, by collaborating with Dr. ‪Gafur Gözükara, we evaluated soil formation along a chronosequence formed from a former lakebed and a biosequence under different land use types in Turkey (Gozukara et al., 2021). By collaborating with Dr. Sérgio Henrique Godinho Silva and Dr. Nilton Curi, we studied a 4.5-m deep Oxisol in Brazil and its formation and variation of geochemical properties using proximal sensing techniques (Mancini et al., 2021). By collaborating with Dr. Raquel de Castro Portes and Dr. Diogo Spinola, we are currently studying Spodosol formation after the glacier retreat and changes of geochemical properties and carbon fractions. I will involve the MIR spectroscopy to investigate the changes of important spectral signatures (organic functional groups and clay minerals) during this process.

My research interest has been recently shifted towards soil change, where I am interested in modeling soil temporal variatiability and its response to climate change and different human activities/disturbance.

Soil profile and its characterization are used for studying soil formation, soil processes, soil morphology, soil properties, and soil classification. Traditional description of soil profiles is often subjective and qualitative. With the development of technologies and computational methods, several quantitative methods are being used to study soil profiles and their horizons.

During my MSc and PhD research, different quantitative methods have been developed to characterize soil profile characteristics and advance the knowledge of digital pedology. A sigmoid depth function was developed based on the pedological and management characteristics of soil profile to characterize the vertical variation of soil pH (Zhang et al., 2017). Sampling designs were compared for SOC stock assessment in soil profiles (Zhang and Hartemink, 2017). Digital soil mapping techniques were used to map the profile wall of an Alfisol with digital images to study its 2D variation, soil weathering stages, and quantitative horizon delineation (Zhang and Hartemink, 2019). Automated soil horizon delineation methods were proposed using vis-NIR and pXRF data (Zhang and Hartemink, 2019) and digital image analysis (Zhang and Hartemink, 2019). Image analysis was used to quantify the content and shape characteristics of coarse fragments in an Alfisol developed from loess over outwash (Zhang et al., 2019).

Additionally, we published a review paper on soil horizon variation (Hartemink et al., 2020) and a book chapter on digital soil morphometrics (Zhang et al., 2022) with many collaborators. By collaborating with Dr. Nataliya Kirillova, we developed color calibration methods of digital images on moist soils (Kirillova et al., 2021). By collaborating with Thibaut Simon and Dr. Christian Walter, we investigated the relationships between color coordinates and soil physiochemical properties of sandy soils (Simon et al., 2020).