We developed a framework to estimate high-resolution spatiotemporal soil moisture (monthly, annual, and seasonal) and temperature-moisture regimes. Our approach uses the Newhall simulation model (NSM) which we fully describe in the Larger Citation. For our analyses, we developed and used open-source software (spatial_nsm) relying on Python^TM^ that was translated from jNSM software (v. 1.6.1; U.S. Department of Agriculture 2016)---a java implementation of the NSM relies on aspatial climate stations. Our software allows for spatial estimates, supports additional parameters to inform the model, and improves upon elements of the originating software.
Briefly, the NSM is an accounting system of water movement in a vertical soil profile and characterizes the soil moisture and soil temperature conditions (Newhall 1972, Van Wambeke 1982, Newhall and Berdanier 1996, Van Wambeke 2000) using monthly precipitation (total), monthly air temperature (mean), and available water capacity (AWC; characteristic of soil properties defining potential to retain water) as data inputs. The model uses the Thornthwaite-Matter-Sellers PET equation (Thornthwaite 1948, Thornthwaite and Mather 1955, Sellers 1965) to reflect energy available for extracting moisture (see Larger Citation; supplemental S8), but other methods can be exchanged within spatial_nsm.
The daily, enumerated simulation of soil moisture movement (NSM) is based on a two-dimensional diagram of the soil profile (Van Wambeke 2000; supplemental S9). The profile extends from the ground surface to a maximum depth of 200 cm (depth defined and characterized by the range of depth corresponding to AWC estimates). For example, clay materials can support an AWC of 200 mm with a soil depth of 80 cm, while sandy loam with an AWC of 200 mm might be as deep as 200 cm (National Soil Survey Center 1998, Van Wambeke 2000). The original NSM estimates soil temperature and moisture regimes (STMRs) to characterize the soil-climate for plant productivity. The STMR classifications described by Natural Resources Conservation Service (NRCS) in Soil Taxonomy (Soil Survey Staff 1999) and Keys to Soil Taxonomy (Soil Survey Staff 2014) were the foundation for classifying soil-climate regimes.
The spatial_nsm is a spatial implementation (data inputs and outputs reflect raster surfaces) of the Newhall model. We modified several components and provided additional tools to assess trends and seasonality. These modifications broadly included the following: 1) implementing a dichotomous approach of the classification (Paetzold 1990) that can reliably produce soil temperature-moisture classifications; 2) providing methods for including air-soil monthly temperature offsets; 3) modularizing spatial_nsm that allows for integration of alternative potential evapotranspiration (PET) methods; 4) accounting for timing and redistribution of snowmelt (given data availability), which improves estimates for when and where water can infiltrate soil; and 5) calculating soil moisture, trends, and seasonality (see Larger Citation for details).
For our study, we applied the model across the western United States using monthly climate averages (1981 -- 2010) to understand whether soil-climate can enhance our understanding of ecological potential and risk. We demonstrated that soil moisture or soil moisture trends correlated significantly with vegetation patterns, including sagebrush, annual herbaceous plant cover, bare ground, and fire occurrence. Because our framework has the flexibility to assess dynamic climate conditions (historical, contemporary, or projected), we can begin to improve our knowledge of changing spatiotemporal biotic patterns. These spatial resources are intended to provide tools to managers and researchers for assessing risk (invasives and fire), improving estimates of vegetation patterns, and informing prioritization of habitat management and expected restoration outcomes.
Developing a spatially explicit soil-climate model involved several steps: 1) identifying and processing spatial input data, 2) processing and collecting data to enhance model, 3) executing models on a high-performance cluster, and 4) identifying key post-analyses products and evaluating soil-climate estimates (Figure 1). For step one, we acquired spatial data representing monthly precipitation and temperature (Prism Climate Group 2015), daily snow deposition (National Operational Hydrologic Remote Sensing Center 2004), soil physical properties (Polaris; Chaney et al. 2016, Chaney et al. 2019), monthly averages from climate stations (Arguez et al. 2010, Arguez et al. 2012), and daily soil conditions from the soil climate analysis network (SCAN; U.S. Department of Agriculture 2020). See the Larger Citation for descriptions of data sources in supplemental (S1; Table S1). Importantly, users of our software are not restricted to these data sources, which are generically described in Data Dependencies. For step two, we used additional data to enhance our model, including the SCAN database to define monthly ambient-temperature offsets, snow cover to account for attenuation of potential evapotranspiration (PET) and offset soil temperature, and organic matter to inform the classification of temperature regimes.
