Figure 2: Comparisons of the 1.6 m ISM means between (a) HRRR and in situ values, (b) CPC and in situ values, and (c) HRRR and CPC values. Note, (c) only shows results from the same locations where in situ data are available for consistency. (d-g) shows comparisons of VSM between the HRRR and in situ data for 4 different level combinations: (d) surface in HRRR to 5 cm in the in situ data, (e) 5 cm for both, (f) 10 cm for both, and (g) 100 cm for both. The gray, horizontal line represents the mean difference for the entire dataset, while red horizontal lines represent the mean difference over the five quintiles. Horizontal gray and red lines are solid if their associated data differences pass statistical significance using a paired Student’s t-test at the 99.9th percentile.
Figure 2 includes the in situ ISM estimates within the comparisons in order to better quantify the differences between the three datasets. When compared to both in situ ISM (Figure 2a) and CPC ISM (Figure 2c), HRRR is wetter in L00-20 (+101% and +97%, respectively, when taking the mean percentage difference for all locations in L00-20) and drier in L80-100 (-34% and -13%, respectively). Even though there are fewer moisture quintiles with statistically significant differences between the CPC and in situ ISMs, the longitude locations show large regional differences between the CPC and in situ datasets, even within a given quintile range. For example, L40-60and L60-80 show minimal differences between the mean CPC and in situ ISMs (Figure 2b, red lines), but at eastern locations the CPC ISMs are generally larger (Figure 2b., yellow and green dots) and at western locations the CPC ISMs are generally smaller (Figure 2b, blue dots). As was also shown in Figure 1, Figure 2 demonstrates clear regional relationships to these ISM differences between the datasets.
Despite larger, systematic, moisture-regime-dependent differences between HRRR and in situ observations (Figure 2a), HRRR aligns more closely to the in situ observation than does the CPC for some locations. These occurrences are typically associated with local soil characteristics or topography, which are not captured in the coarser climate division regions used in the CPC product. For example, based on comparisons with the in situ ISM, the HRRR produces similarly dry soil moisture conditions within the Sand Hills region of Nebraska (e.g., -101.4 W, 42.1 N in Figure 1). The temporally averaged mean ISM for this location is 138 mm and 165 mm for the in situ and HRRR data, respectively, as compared to 372 mm in the CPC data (2-3x larger).
4.2 VSM Mean Comparisons at Varying Depths
To better understand the differences in ISM data, the VSM data at different depths are compared in the HRRR and in situ data (Figure 2d-g). Note, that the VSM data are also separated into quintiles from driest to wettest in a similar way to the ISM data, and this procedure is done for each vertical level. Therefore, locations may fall into different quintiles for different vertical levels. The same trend of the HRRR being wetter in L00-20 (i.e., the driest regions) and dryer in L80-100 (i.e., the wettest regions) is present at all depths. There are generally smaller differences in the middle quintiles. However, at the lowest depths (100 cm below ground; Fig 2g), the differences between the HRRR and in situ VSMs have larger magnitudes, especially for the driest and wettest regimes (L00-20 difference of +0.08 and L80-100difference of -0.18). Near the surface (Figure 2d), the driest 40% of the regions have relatively small differences (± 0.02), although significant dry biases are present for the wetter regions (-0.11). The VSM data demonstrate that the deeper soil layers (i.e., 100 cm below ground) are the primary drivers of the ISM trends. Shallower soil layers have similar trends yet smaller differences as compared to those at deeper levels and have a smaller contribution to the ISM differences. It is important to note that the magnitude of both ISM and VSM mean differences vary throughout the year, and the evolution of these differences as a function of month and season are provided in the supporting information document.
5 Soil Moisture Variance
5.1 ISM Standard Deviation Comparisons
Understanding the temporal variability (e.g., standard deviations) in soil moisture allows for an assessment on whether models are accurately capturing the processes that result in soil moisture changes. Maps of the ISM standard deviations in the HRRR and CPC ISM data show similar patterns, with the highest variance occurring along the Pacific Northwest coast (Figure 1d-e). The lowest ISM standard deviations occur along the Intermountain West and High Plains regions. In general, the HRRR ISMs have lower variance than the CPC data across the US except for in the parts of the Rocky Mountains and in the Great Lakes and Northeast regions (Figure 1f). The HRRR produces larger spatial variability within the mountain and valley regions across the western US that cannot be resolved in the CPC data.