1. Introduction
Soil moisture observations are an important component of the hydrological cycle that are increasingly being explored to evaluate hydrological extremes, particularly in the context of climate change. Soil moisture deficits have been shown to be a leading indicator of drought onset and worsening drought conditions (Leeper et al. 2021; Torres et al. 2013), linked with agricultural losses (Krueger et al. 2019; Narasimhan and Srinivasan 2005) and vegetation development (Warter et al. 2021), and increasingly explored as an indicator for wildfire fuel status (Littell et al. 2016; O et al. 2020; Reardon et al. 2007; Krueger et al. 2023). Additionally, antecedent soil moisture conditions have also been utilized in flooding applications (Brocca et al. 2017; Ochsner et al. 2013) as a measure to estimate potential runoff (Aubert et al. 2003; Norbiato et al. 2008). Le Lay and Saulnier (2007) found that initial soil moisture conditions were as important as the spatial distribution of rainfall for a significant flash flooding event over the Cévennes-Vivarais region of France that resulted in 24 deaths and EUR 1.2 billion (USD 1.26 billion) in damages in 2002.
The expanded use of soil moisture conditions to monitor hydrological extremes requires both a temporal record to adequately standardize (Leeper et al. 2019; Ford et al. 2016) or place soil moisture conditions into historical context (i.e., drier or wetter than usual) and sufficient spatial resolution to monitor evolving conditions over a region (i.e., hydrological basin or agricultural zones). Standardizing soil moisture observations generally improves the interpretability of these measures by accounting for local factors such as soil characteristics (i.e., faction of sand, silt, and clay), land-cover type, topography, and local seasonal variations in precipitation (i.e., normal wet/dry periods), which heavily influence these observations (Krueger et al. 2019). Standardization methods for soil moisture can be generally separated into two types: methods that place observations into seasonally historical context that are expressed as standardized soil moisture anomalies (Leeper et al. 2019) or percentiles (Ford et al. 2016) or methods that place observations into the context of soil water holding capacity such as plant available water (Scott et al. 2013). The former approach only requires a sufficient period record from which to evaluate a soil moisture baseline as compared with the later approach that requires measures of wilting point and field capacity that are more challenging to estimate over satellite pixels at national and global scales.
In situ stations monitoring soil moisture conditions continue to rapidly expand across the upper Missouri River basin (U.S. Army Corps of Engineers 2022), Hawaii (Arakaki 2021), and in states across the Southeast United States (National Integrated Drought Information System 2021a) as part of the National Coordinated Soil Moisture Monitoring Network (NCSMMN; National Integrated Drought Information System 2021b). However, the spatial coverage of soil moisture observing stations is often insufficient to effectively monitor conditions across many hydrological basins, agricultural regions (i.e., the Corn Belt), and other relevant zones (coastal, aquifers, etc.) at temporal scales necessary to support climatological assessments.
Remotely sensed observations of soil moisture conditions offer the spatial coverage necessary to support hydrological monitoring across space; however, these data can be temporally limited by the satellite’s life span. For instance, their sensitivity to solar flares, space debris, and malfunctions can result in long (i.e., monthlong or greater) data gaps that can be particularly challenging to overcome in climatological assessments (National Academies of Sciences, Engineering, and Medicine 2015). While most soil-moisture-sensing satellites meet the minimum 5–7 years to place soil moisture conditions into historical context (Leeper et al. 2019; Ford et al. 2016), additional years are generally necessary to characterize thresholds for the detection of extreme soil moisture conditions that occur during droughts and floods. These factors are further constrained by spatial averaging over the satellite pixel and calibration issues that can arise near the observational range (extreme) limits. In addition, more recent single satellite missions may not capture historically relevant soil moisture extremes events. For instance, NASA’s Soil Moisture Active Passive (SMAP) satellite provides one of the most accurate measures of absolute surface soil moisture conditions from space (Beck et al. 2021; Ford and Quiring 2019), but has collected data only since 2015, missing a number of important extreme events in the United States such as the central United States 2012 drought or the 2001 Mississippi River flood, which were prior to its deployment.
The European Space Agency’s (ESA) Climate Change Initiative (CCI) has developed techniques to harmonize remotely sensed soil moisture conditions across differing active and passive satellite platforms (Dorigo et al. 2017; Gruber et al. 2019). This has led to the development of one of the most comprehensive multisatellite-based long-term (since 1978) soil moisture datasets available globally at daily time scales. The ESA has developed and maintains three products that consist of the active satellites (Active), Passive satellites (Passive), and merged Active and Passive satellite (Combined) products. Of these, only the ESA’s Combined and Passive datasets span the temporal range (since the 1980s) with soil moisture conditions provided in volumetric form.
Verification studies of the ESA soil moisture products have largely focused on the Combined dataset, which were found to detect variations in volumetric soil moisture observations (Al-Yaari et al. 2019; Ford and Quiring 2019) despite underperforming the SMAP dataset. However, over North America, Europe, and to some extent Australia the ESA’s Combined measures of bias, root mean squared error, and unbiased root mean squared error were similar to SMAP’s (Al-Yaari et al. 2019). A few studies have evaluated separately the ESA’s Passive dataset, which uses a differing merging algorithm than the Combined dataset, or the recently released version 7.1 that includes the addition of SMAP, MetOp-C, Fengyun (FY) FY-3C, and FY-3D sensors in addition to important changes to the merging algorithm that are documented in both Preimesberger et al. (2021) and Scanlon et al. (2022).
The purpose of this study is to evaluate the ability of ESA’s, version 7.1, Combined and Passive soil moisture datasets to detect both drought and flooding extremes as monitored by the U.S. Climate Reference Network (USCRN; Diamond et al. 2013; Bell et al. 2013). This feasibility study will focus on how well the separate ESA datasets captured the range in volumetric and standardized soil moisture conditions across the United States with an emphasis on the extreme ends of the distribution. It is anticipated that this analysis will provide additional context for the use of ESA datasets to capture historical soil moisture extremes for climatological purposes.
2. Data
a. Remotely sensed soil moisture
The ESA, through its CCI, has developed techniques to harmonize observed soil moisture conditions across multiple satellite sensing soil moisture missions since the late 1970s (Dorigo et al. 2017; Gruber et al. 2019). These processes continue to be updated and integrate new soil moisture sensing satellites as they become available (Preimesberger et al. 2021; Scanlon et al. 2022). These merging techniques have resulted in three primary datasets that consist of all passive satellite missions (Passive), all active satellite missions (Active) and merged Passive and Active satellite datasets (Combined). In this study, the Passive and Combined datasets, which are the only two products of the three that provide daily global coverage of volumetric soil moisture conditions since 1978 at a 25-km spatial resolution, will be used to evaluate how well these harmonized datasets represent dry and wet extreme soil moisture conditions. More information on the scaling and cumulative distribution function matching methods used to merge these satellite datasets can be found in the Scanlon et al. (2022) technical report.
b. USCRN soil moisture
The USCRN hourly standardized soil moisture dataset (https://www.ncei.noaa.gov/pub/data/uscrn/products/soil/soilanom01/) was used in this study to evaluate the performance of the ESA’s Combined and Passive anomalies. A description of the how the standardized hourly anomalies were evaluated can be found in Leeper et al. (2019), but is briefly described here. Standardized anomalies were evaluated by differencing the hourly soil moisture observations from the respective hourly median climatology and divided through by the interquantile range. The median and interquantile ranges were evaluated for the same clock hour over a centered 31-day window. For instance, the 1800 UTC 15 May climatology and interquantile range were based on all available 1800 UTC soil moisture observations between 30 April and 31 May over the station’s period of record. No hourly climatologies or anomalies were derived when more than 60% of the expected historical soil moisture observations were missing. For comparisons with the satellite datasets, the hourly standardized soil moisture anomalies and volumetric observations were then averaged over (0000–0023 UTC) calendar days to temporally align with the ESA’s daily soil moisture conditions. A sensitivity analysis, not shown here, revealed that the USCRN’s 24-h averages of volumetric soil moisture were better aligned with the ESA’s datasets than the 0012 UTC observations that the ESA products represent.
3. Methods
List of U.S. states (postal abbreviations) by climate region.
The second portion of the analysis was focused on how well these measures captured the evolution of two well-known extreme dry and wet events. The central U.S. 2012 drought, which rapidly expanded and became one of the most widespread droughts on record (Rippey 2015) with over 77% of the United States in abnormally dry or worse (D0 or greater) conditions by 14 August 2012 (National Drought Mitigation Center 2022), was evaluated here as the dry event. The extreme wet event was the flood that occurred across the upper Missouri River basin (UMRB) in 2019. This event was proceeded by higher-than-usual streamflow anomalies and snowpack that rapidly melted during an unusual warm, heavy rain event. As a result of frozen soils that initially prevented the snowmelt from infiltrating the soil and continuing positive precipitation anomalies, flooding conditions developed across the UMRB, saturating soils across the western U.S. Corn Belt and prevented many areas from planting (English et al. 2021). These two extreme events lead to widespread agricultural and economic losses totaling billions of dollars (English et al. 2021; Smith and Katz 2013). To evaluate these two events, impacted stations shown in Fig. 1a for the 2012 drought (those having D0 or greater drought status as of 14 August 2012) and Fig. 1b for the UMRB flood (those within the UMRB region) were averaged to create time series of both satellite and in situ soil moisture conditions, and to evaluate differences in the magnitude and evolution of the extreme events.
USCRN stations used to evaluate (a) the central United States 2012 drought (orange) and (b) the UMRB 2019 flood (green) case studies.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Dry and wet anomalous thresholds used to identify extreme days and evaluate measures of sensitivity and precision.
4. Results
a. Volumetric comparisons
Overall, the ESA Combined and Passive products both had a slight wet bias over the USCRN of 0.03 ± 0.08 m3 m−3 and 0.05 ± 0.15 m3 m−3, respectively. However, distributions of daily soil moisture observations (Fig. 2a) revealed that the Combined and Passive products had fewer soil moisture measures below 0.1 m3 m−3 than did USCRN stations. For the Combined product, this was also the case on the wet end of the distribution with very few observations greater 0.4 m3 m−3. However, this was not the case for the Passive dataset, which had slightly more observations greater than 0.4 m3 m−3 than USCRN. A scatterplot of the volumetric soil moisture conditions revealed that the limited range of the combined satellite dataset resulted in both wet and dry biases when USCRN soil moisture conditions were on the extreme dry or wet ends of its distribution, respectively. Despite having a longer tail on the wet end of the distribution, this was also case for the Passive dataset as its wettest conditions (greater than 0.4 m3 m−3) were generally observed within the middle of USCRN’s (between 0.1 and 0.4 m3 m−3) distribution (Fig. 2b), leading to an underreporting of extreme wet soil moisture conditions. As a result, the Combined dataset had slightly lower measures of MAE (0.07 and 0.12 m3 m−3) and RMSE (0.09 and 0.16 m3 m−3) than the Passive datasets.
(a) Daily national USCRN 5-cm (black), ESA Combined surface (orange), and ESA Passive surface (green) volumetric soil moisture histogram, and heat maps of ESA (b) Passive and (c) Combined surface soil moisture against USCRN 5-cm soil moisture.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
The fewer measures of remotely sensed extreme soil moisture conditions were more notable for the Combined dataset in regions where precipitation had pronounced wet and dry seasons (Fig. 3). For instance, the USCRN had more frequent drier observations than either satellite dataset at stations in the Southwest followed by the West and South (that includes west Texas). Likewise, the USCRN tended to have more frequent wetter observations at stations in the Ohio Valley and to some extent in the Southeast (both with longer distribution tails). However, this was not true of the Passive dataset, which in most regions equaled or exceeded the frequency of USCRN for 0.4 m3 m−3 or greater observations. The northern Rockies and plains was the region where the distributions of both remotely sensed datasets aligned best with USCRN. Not surprisingly this region had some of the lowest measures of error and highest measures of D index for the Combined and Passive datasets (Table 3).
Daily USCRN (black) 5-cm and ESA’s surface Combined (orange) and Passive (green) soil moisture histograms for the (a) Northwest, (b) northern Rockies and plains, (c) Northeast, (d) West, (e) Upper Midwest, (f) Ohio Valley, (g) Southwest, (h) South, and (i) Southeast climate regions.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Volumetric soil moisture regional medians of bias (volumetric), MAE (volumetric), RMSE (volumetric), and D index of agreement for the ESA’s Passive and Combined datasets.
b. Anomaly comparisons
The distribution of remotely sensed standardized soil moisture anomalies had modes (around zero) that were more similar to USCRN despite the in situ network having more near-zero anomalies, resulting in higher measures of kurtosis (9.3) relative to the Combined (5.4) and Passive (6.2) datasets (Fig. 4a). The scatterplots revealed that both the Passive (Fig. 4b) and Combined (Fig. 4c) datasets had the bulk of their observations around the 1-to-1 line when the anomalies were within +1 to −1 standardized units (stu). However, there was considerable spread at the extremes, with a low-density cloud of pixels in the heat maps far removed from the 1-to-1 line, yielding regression slopes of less than 0.5 for both. Overall, the Combined dataset had slightly lower measures mean bias (0.03 stu), MAE (0.55 stu), and RMSE (0.74 stu) than the Passive product, which were −0.03, 0.60, and 0.84 stu, respectively. The higher MAE and RMSE values were likely caused by the spread or variance in the daily standardized soil moisture datasets, which were 0.59, 0.57, and 0.64 for the USCRN 5-cm, Combined, and Passive datasets, respectively. Similar to the volumetric comparisons, the northern Rockies and plains had some of the lowest measures of regional errors and higher D index of agreements in both the Passive and Combined datasets (Table 4).
(a) Daily national USCRN 5-cm (black), ESA Combined surface (orange), and ESA Passive surface (green) standardized soil moisture anomaly histogram, and heat maps of ESA (b) Passive and (c) Combined surface soil moisture anomalies against USCRN 5-cm anomalies.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Standardized soil moisture anomalies by regional medians of bias (standardized), MAE (standardized), RMSE (standardized), and D index of agreement for the ESA’s Passive and Combined datasets.
The ESA satellite products were generally in good agreement with USCRN standardized anomalies for the eastern two-thirds of the United States with a slight dry bias for western stations (Fig. 5). However, for wet conditions, when USCRN standardized anomalies exceeded +0.3 standardized units, the Passive and Combined ESA datasets reported a dry bias for a majority (greater than 85%) of stations with stronger biases over the western United States. For dry conditions (USCRN less than −0.3 interquartile deviations), both the Passive and Combined datasets had a strong wet bias for roughly 84% of stations relative to USCRN (Fig. 6), suggesting that in situ observations have a larger range during more extreme conditions. Despite these differences in sign of the mean bias, mean absolute deviations for all wet and dry conditions were much more similar for both the passive and combined datasets (Fig. 6). Both ESA datasets had slightly lower measures of MAE in the central United States and higher MAE in the western United States. Interestingly, there were more stations at the highest range of the MAE (>0.7 stu) for the extreme wet conditions (40 for Passive and 30 for Combined) than for the extreme dry condition (19 for Passive and 11 for Combined). That said, the Combined dataset overall had fewer stations in the higher MAE ranges for both wet and dry conditions as compared with the Passive dataset.
Station biases for ESA’s (a)–(c) Passive and (d)–(f) Combined for (left) all standardized soil moisture anomaly observations, (center) extreme wet conditions, and (right) extreme dry conditions.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Station MAE for ESA’s (a)–(c) Passive and (d)–(f) Combined for (left) all standardized soil moisture anomaly observations, (center) extreme wet conditions, and (right) extreme dry conditions.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
c. Extreme events
To explore these results a bit further, time series of mean standardized anomalies over stations impacted by the central 2012 drought and 2019 UMRB floods were analyzed. For the 2012 drought, the USCRN had lower volumetric soil moisture conditions than either satellite dataset for all three years with all three datasets showing a distinct wet to dry pattern from March to September (Fig. 7a). That said, the standardized anomalies from all three products exhibit a decline in anomalous conditions around May 2012 with prolonged conditions of −0.2 or less standardized anomalies through July. While the satellite dataset observed some of the driest daily anomalies particularly the combined dataset dipping below −0.7, both the ESA Combined and Passive datasets had greater variability than in situ datasets and a larger range (from −0.7 to 0.3 standardized units) relative to USCRN (from −0.4 to 0.2 standardized units). However, the satellite datasets were able to capture the evolution of the 2012 drought from the dry down in 2012 to recovery (standardized anomalies approach zero) in 2013. For the flooding scenario, the remotely sensed datasets had a persistent dry volumetric offset in all three years (2018–20). This was also evident in the standardized anomalies that were particularly notable during the spring of 2019 when USCRN reached its peak slightly above the satellite datasets between March and April. The ESA satellite datasets also peaked again in early summer around June not detected by USCRN. Despite important offsets in the magnitude and timing of the flooding conditions, the satellite datasets were able to capture the wet-up trends in late 2018 followed by wetter than usual conditions (standardized anomalies greater than +0.3) in 2019 before conditions normalized (dried down) going into 2020.
Station averaged volumetric time series of USCRN (black), ESA Combined (orange), and Passive (green) datasets for stations impacted by the (a) 2012 drought and (b) the 2019 flood over the UMRB. Also shown are standardized anomaly averages for (c) the 2012 drought and (d) the 2019 flood.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
d. Detection of extremes
Evaluations of sensitivity and precision revealed that, while the Combined dataset had only a slight advantage over the Passive dataset, the performance of both ESA datasets diminished as the magnitude of the standardized anomaly threshold increased for both wet and dry conditions (Fig. 8).
ESA Passive (green) and Combined (orange) measures of (left) sensitivity and (right) precision for various dry (negative) and wet (positive) anomalous thresholds by 1-, 3-, and 7-day-averaged soil moisture standardized anomalies.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
For example, the sensitivity of the Combined dataset showed that it could detect conditions wetter than 0.3 standardized units above normal roughly 60% of the time USCRN met these conditions as compared with 58% for the Passive dataset; however, the Combined dataset diminished to 30% at the 1.3 standardized unit threshold. This was also case for the dry thresholds albeit the drop was slightly greater, dropping from 58% at the −0.3 threshold to 18% at the −1.3 threshold. Measures of precision, which reports the fraction of times the satellite matched USCRN with respect to the number of times the satellite met the conditions, were very similar to sensitivity suggesting that the ESA datasets did not overreport extreme conditions. Apart from slight adjustments for the more extreme thresholds, these results were little changed when averaging the daily standardized anomalies over 3- and 7-day moving windows, which suggests that noise in the daily dataset had little influence on ESA performance.
The annualized counts of extreme wet and dry days were similar between ESA’s remotely sensed (Passive and Combined) datasets and USCRN (Fig. 9). At the national level, the USCRN as well as the ESA’s Passive and Combined datasets showed 2012 as having the greatest number of extreme dry days followed by 2011. Similarly, 2010 and 2019 were the years with the greatest number of wet days (0.5 standardized unit anomalies or greater) for all three products. However, there were some offsets with both ESA’s Combined and Passive datasets reporting slightly more dry days from 2013 to 2020, especially during 2017 and 2018, and slightly fewer wet days from 2015 onward than USCRN stations.
Annual counts of extreme dry days for soil moisture standardized anomalies (a) −0.5 or below and (b) 0.5 or greater from 2010 to 2021.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Regional evaluations revealed larger offsets in extreme counts of both wet and dry days (Figs. 10 and 11). In the West, for instance, the satellite datasets (Combined and Passive) initially underreported the number of dry days relative to USCRN particularly in 2010 and 2011, which extended into 2012 for the Combined dataset. However, this reversed in 2015 with both satellite datasets overreporting the number of dry days detected at USCRN stations from 2015 to 2021 with the exception of 2017. The northern Rockies and plains, South, and Southeast regions also had higher numbers of satellite based dry days than did USCRN beginning in 2015. The satellite datasets fell more in line with USCRN dry day counts by 2018 over the northern Rockies and plains and South and 2020 for the Southeast. Counts of wet days were generally more similar to USCRN than dry day counts, but there were notable differences with the satellite dataset greatly underreporting the number of wet days relative to USCRN over the West (2019 to 2021). The satellite datasets also underreported USCRN wet days persistently, albeit not by a large amount, in the Midwest from 2015 onward (2016 being an exception) and in the Southeast from 2017 through 2021 (excluding 2020).
Annualized counts of the number of extreme dry days where the soil moisture standardized anomalies were −0.5 or less for the (a) Northwest, (b) northern Rockies and plains, (c) Northeast, (d) West, (e) Upper Midwest, (f) Ohio Valley, (g) Southwest, (h) South, and (i) Southeast climate regions.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Annualized counts of the number of extreme wet days where the soil moisture standardized anomalies were 0.5 or greater for the (a) Northwest, (b) northern Rockies and plains, (c) Northeast, (d) West, (e) Upper Midwest, (f) Ohio Valley, (g) Southwest, (h) South, and (i) Southeast climate regions.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Despite these offsets, the satellite datasets were able to detect higher counts of dry days during significant drought episodes for the differing regions such as the 2011 drought in the South, the 2012 drought over the northern Rockies and plains and Ohio Valley, and the recent (2020/21) droughts over the Northwest and northern Rockies and plains regions. This was also case for extreme wet days, with both ESA satellite datasets and USCRN capturing the important flood conditions over the northern Rockies and plains in 2011 and 2019, as well as extreme wet conditions across the Southwest and South regions in 2015. In fact, the overall interannual pattern in the number of extreme wet and dry days were similar among the three datasets for most regions with slightly tighter agreement for extreme wet days as compared with extreme dry days.
5. Discussion and conclusions
Volumetric soil moisture comparisons between USCRN and ESA’s Combined and Passive satellite datasets revealed important differences. The Combined dataset, for instance, had a much tighter range in volumetric conditions with few measures outside of 0.05 and 0.45 m3 m−3; some regions had even tighter ranges (Northwest, Northeast, and Ohio Valley) when compared with either USCRN or the Passive data. Conversely, the Passive dataset had a range in volumetric soil moisture that was more similar to the USCRN range, but with slightly higher frequency of soil moisture observations exceeding 0.4 m3 m−3 than USCRN (Figs. 2 and 3) that was fairly persistent across most regions.
Differences in the distribution of possible soil moisture conditions lead to contrasting biases that were dependent upon USCRN soil moisture conditions. For instance, the ESA’s Combined and Passive datasets had a wet bias for drier USCRN observations (i.e., soil moisture less than 0.15 m3 m−3) and a dry bias for wetter measures (i.e., soil moisture conditions exceeding 0.4 m3 m−3). To explore if these biases were related to land-cover type and or heterogeneity of the land cover (25 km) around the station (satellite footprint), vegetation cover data around each USCRN station was extracted from the 500-m resolution Modis Terra + Aqua Combined Land Cover product (Friedl and Sulla-Menashe 2019). With an exception for barren land cover that consisted of only two stations (Yuma, Arizona, and Stovepipe Well, California), these biases were found to be persistent regardless of USCRN’s land-cover type and land-cover heterogeneity with biases only slightly diminishing as land cover within a 25-km area of a USCRN site became more uniform (Fig. 12). These results suggest that land cover heterogeneity or land-cover type may not fully explain the offsets in volumetric biases between USCRN and the ESA Combined and Passive datasets.
ESA’s standardized anomaly MAE by (top) percent of land cover matching USCRN’s and by (bottom) USCRN land-cover type for (left) Combined and (right) Passive datasets.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
In addition, the volumetric biases and MAEs between the separate satellite products (Combined and Passive) and USCRN measurements were not similar despite the Combined product including the Passive dataset over much of this study period. This suggests that differences in the range of possible volumetric observations for the Combined dataset may be related to the use of Global Land Data Assimilation System (GLDAS), which is a land surface model providing historical soil moisture conditions used to merge the Passive and Active datasets (Dorigo et al. 2017). For instance, the limited range in volumetric soil moisture from the Combined dataset was similar to other comparison studies between USCRN and modeled soil moisture conditions from the North American Regional Reanalysis (Leeper et al. 2017) and the High-Resolution Rapid Refresh (Lee et al. 2022) models. Regardless, since the magnitude of volumetric soil moisture biases increase toward the extreme ends (i.e., wet and dry sides) of the distribution, there will be limitations on the use of remotely sensed volumetric soil moisture conditions to evaluate the magnitude of extreme conditions. That said, the constrained range of the ESA’s Combined dataset may be useful in other applications such as assimilating soil moisture conditions into numerical models that operate over a similar volumetric soil moisture range (de Goncalves et al. 2006; Koster et al. 2004).
Distributions of the standardized soil moisture anomalies between USCRN and ESA’s Combined and Passive datasets were more aligned with the USCRN standardized soil moisture anomalies than the volumetric data. However, the USCRN had a higher frequency of near-zero anomalies than either satellite dataset. This was also the case regionally, which suggests that the remotely sensed biases in volumetric conditions were persistent over time and in general could be mitigated when differencing daily conditions from a seasonally adjusted baseline. These results are in line with Entekhabi et al. (2010) who noted that remotely sensed soil moisture observations can still be useful if the variability in these measures are persistent in time. In fact, this is fairly evident in the extreme drought and flood case studies of the 2012 drought in the central United States and the upper Missouri River basin 2019 flood. Despite offsets in the timing and magnitude of extreme conditions, the standardized anomalies from both remotely sensed datasets captured the evolution of USCRN’s conditions from onset to recovery with the Combined dataset slightly outperforming the Passive product.
This was also found to extend to annualized counts of extreme days where both satellite datasets captured the interannual variability in extreme wet and dry soil moisture days from 2010 to 2021. For instance, both the USCRN and ESA satellite products reported 2011, 2012, and 2021 as the top three years with the greatest number of dry days. This was also for counts of wet days with all three datasets highlighting 2010, 2015, and 2019 as the top three years with a slight divergence in the rank of those years. Nationally, the satellite datasets tended to overreport the number dry days after 2012 and underreport wet days after 2014 relative to USCRN, albeit differences were generally within a window of ±15 days.
However, there were important regional differences that varied by year. For instance, in the Southwest, both satellite datasets overreported the number of extreme dry days relative to USCRN every year from 2010 to 2021. This was largely the case for the West, South, and Southeast regions after 2014 with an exception for a few years. Interestingly, these were the same regions where USCRN stations tended to have a higher frequency of near-zero (0–0.15 m3 m−3) volumetric soil moisture observations (Fig. 3), which suggests USCRN stations in these regions may be located in well drained or sandier soils. To explore this further, the dominate 5-cm soil type (i.e., sand, silt, or clay) were identified for each USCRN station using a combination of Pedon results derived from USCRN soil samples by the U.S. Department of Agriculture’s (USDA) National Soils Survey Center, and when not available extracted from USDA’s Soil Survey Geographic Database (SURRGO). Station means of the ESA Combined and Passive soil moisture standardized anomaly MAE revealed that stations with higher concentration of sandy soils tended to have larger measures of error (Fig. 13). These results suggest that soil type was more important to satellite measures of error than land cover, and that in sandier locations the detection of extremes may be more challenging.
Remotely sensed soil moisture standardized anomaly MAE for (left) Combined and (right) Passive datasets by (a),(b) percent sand; (c),(d) percent clay; and (e),(f) percent silt.
Citation: Journal of Applied Meteorology and Climatology 62, 11; 10.1175/JAMC-D-23-0059.1
Another interesting feature of both the dry and wet day counts was how the differences between USCRN and satellite counts varied over differing years. For instance, the West region initially underreported the number of USCRN dry days from 2010 to 2014 but reversed to overreporting USCRN from 2015 to 2021. This was also the case over the northern Rockies and plains for most years with the exception to the latter years (2020 and after). Some regions had similarly timed patterns for the wet day counts (i.e., northern Rockies and plains, and Upper Midwest), albeit the inverse with slightly smaller shifts. Overall, it is not clear if the temporal shifts in extreme day counts, which were not detected in all regions, are related to the mix of available satellites in the ESA products, the availability of station and satellite observations within regions, and or the nature and seasonal timing of the extreme events.
The regional offsets in the number of dry or wet extremes may not be that surprising since both the ESA’s Combined and Passive standardized anomalies were not able to capture the magnitude of extremes well on individual days. Measures of sensitivity and precision dropped as the anomalous threshold used to define extreme events increased; the Combined and Passive datasets sensitivity and precision dropped below 50% at 0.5 and −0.3 thresholds for the extreme wet and dry events, respectively. While 3- and 7-day moving averages did little to change the overall measures of sensitivity and precision, there were slight improvements to correlation and error metrics (not shown here), which likely resulted from smoothing out some of the noise in the daily estimates of soil moisture conditions. These results suggest that in cases where the magnitude of soil moisture extremes are important (i.e., vegetation health, wildfire), in situ records of soil moisture observations are particularly relevant despite their limited spatial coverage.
In this study, two long-term remotely sensed soil moisture datasets from the ESA CCI, version 7.1, were compared with USCRN soil moisture conditions from 83 stations across the United States. Evaluations revealed important volumetric soil moisture biases particularly on the extreme ends of the distribution that may be more related to soil characteristics than vegetation cover or its heterogeneity. Despite these offsets, standardized soil moisture anomalies from ESA’s Combined and Passive products were better aligned with USCRN and able to detect trends in worsening or improving soil moisture conditions, but not necessarily the magnitude of its severity. These results indicate that standardized metrics (i.e., anomalies and percentiles) are preferred over volumetric conditions when using ESA’s Combined or Passive remotely sensed products to detect soil moisture extremes. In addition, the longevity of the ESA’s dataset back to 1978, while not fully analyzed here, can reasonably be expected to provide important insights on soil moisture trends and variability over climatological scales. While these results may need to be verified in other regions, this could be particularly useful in global assessments of soil moisture change. Furthermore, the ESA also provides semi-near-real-time Combined and Passive datasets that are available at a ∼10-day latency (https://cds.climate.copernicus.eu/cdsapp#!/dataset/satellite-soil-moisture?tab=overview), which could support drought related monitoring efforts over sparsely instrumented regions. The harmonization of historical and current satellite measures across time makes the ESA’s Combined and Passive datasets very useful for detecting climatological variations in soil moisture conditions. When combined with in situ measurements, a convergence of evidence approach can be used to detect the presence and magnitude of hydrological extremes.
Acknowledgments.
This work was supported by NOAA through the Cooperative Institute for Satellite Earth System Studies (CISESS) under Cooperative Agreement NA19NES4320002 and by the National Integrated Drought Information System (NIDIS). Special thanks are given to the USCRN programmatic team for their vital support.
Data availability statement.
All data created or used for this study are openly available either from NOAA’s National Centers for Environmental Information (NCEI) for USCRN data (https://www.ncei.noaa.gov/pub/data/uscrn/products/soil/soilanom01/) or through the ESA CCI for the remotely sensed Passive and Combined soil moisture datasets used here (https://esa-soilmoisture-cci.org/).
REFERENCES
Al-Yaari, A., and Coauthors, 2019: Assessment and inter-comparison of recently developed/reprocessed microwave satellite soil moisture products using ISMN ground-based measurements. Remote Sens. Environ., 224, 289–303, https://doi.org/10.1016/j.rse.2019.02.008.
Arakaki, M., 2021: $1.33M grant to better understand, forecast Hawaii‘s complex weather and climate. University of Hawai‘i News, accessed 30 December 2022, https://www.hawaii.edu/news/2021/10/10/hawaii-mesonet-project/.
Aubert, D., C. Loumagne, and L. Oudin, 2003: Sequential assimilation of soil moisture and streamflow data in a conceptual rainfall–runoff model. J. Hydrol., 280, 145–161, https://doi.org/10.1016/S0022-1694(03)00229-4.
Beck, H. E., and Coauthors, 2021: Evaluation of 18 satellite- and model-based soil moisture products using in situ measurements from 826 sensors. Hydrol. Earth Syst. Sci., 25, 17–40, https://doi.org/10.5194/hess-25-17-2021.
Bell, J. E., and Coauthors, 2013: U.S. Climate Reference Network soil moisture and temperature observations. J. Hydrometeor., 14, 977–988, https://doi.org/10.1175/JHM-D-12-0146.1.
Brocca, L., L. Ciabatta, C. Massari, S. Camici, and A. Tarpanelli, 2017: Soil moisture for hydrological applications: Open questions and new opportunities. Water, 9, 140, https://doi.org/10.3390/w9020140.
de Goncalves, L. G. G., W. J. Shuttleworth, S. C. Chou, Y. Xue, P. R. Houser, D. L. Toll, J. Marengo, and M. Rodell, 2006: Impact of different initial soil moisture fields on Eta model weather forecasts for South America. J. Geophys. Res., 111, D17102, https://doi.org/10.1029/2005JD006309.
Diamond, H. J., and Coauthors, 2013: U.S. Climate Reference Network after one decade of operations: Status and assessment. Bull. Amer. Meteor. Soc., 94, 485–498, https://doi.org/10.1175/BAMS-D-12-00170.1.
Dorigo, W., and Coauthors, 2017: ESA CCI soil moisture for improved Earth system understanding: State-of-the art and future directions. Remote Sens. Environ., 203, 185–215, https://doi.org/10.1016/j.rse.2017.07.001.
English, B. C., S. A. Smith, R. J. Menard, D. W. Hughes, and M. Gunderson, 2021: Estimated economic impacts of the 2019 Midwest floods. Econ. Disasters Climate Change, 5, 431–448, https://doi.org/10.1007/s41885-021-00095-2.
Entekhabi, D., R. H. Reichle, R. D. Koster, and W. T. Crow, 2010: Performance metrices for soil moisture retrievals and application requirements. J. Hydrometeor., 11, 832–840, https://doi.org/10.1175/2010JHM1223.1.
Ford, T. W., and S. M. Quiring, 2019: Comparison of contemporary in situ, model, and satellite remote sensing soil moisture with a focus on drought monitoring. Water Resour. Res., 55, 1565–1582, https://doi.org/10.1029/2018WR024039.
Ford, T. W., Q. Wang, and S. M. Quiring, 2016: The observational record length necessary to generate robust soil moisture percentiles. J. Appl. Meteor. Climatol., 55, 2131–2149, https://doi.org/10.1175/JAMC-D-16-0143.1.
Friedl, M., and D. Sulla-Menashe, 2019: MCD12Q1 MODIS/Terra+Aqua land cover type yearly L3 global 500m SIN grid V006. NASA EOSDIS Land Processes Distributed Active Archive Center, accessed 20 July 2023, https://doi.org/10.5067/MODIS/MCD12Q1.006.
Gruber, A., T. Scanlon, R. van der Schalie, W. Wagner, and W. Dorigo, 2019: Evolution of the ESA CCI soil moisture climate data records and their underlying merging methodology. Earth Syst. Sci. Data, 11, 717–739, https://doi.org/10.5194/essd-11-717-2019.
Karl, T. R., and W. J. Koss, 1984: Regional and national monthly, seasonal, and annual temperature weighted by area, 1895–1983. Historical Climatology Series 4-3, National Climatic Data Center, 38 pp.
Koster, R. D., and Coauthors, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305, 1138–1140, https://doi.org/10.1126/science.1100217.
Krueger, E. S., T. E. Ochsner, and S. M. Quiring, 2019: Development and evaluation of soil moisture-based indices for agricultural drought monitoring. Agron. J., 111, 1392–1406, https://doi.org/10.2134/agronj2018.09.0558.
Krueger, E. S., and Coauthors, 2023: Using soil moisture information to better understand and predict wildfire danger: A review of recent developments and outstanding questions. Int. J. Wildland Fire, 32, 111–132, https://doi.org/10.1071/WF22056.
Lee, T. R., R. D. Leeper, T. Wilson, H. J. Diamond, T. P. Meyers, and D. D. Turner, 2022: Using the U.S. Climate Reference Network to identify biases in near- and sub-surface meteorological fields in the High-Resolution Rapid Refresh (HRRR) weather prediction model. Wea. Forecasting, 38, 879–900, https://doi.org/10.1175/WAF-D-22-0213.1.
Leeper, R. D., J. E. Bell, C. Vines, and M. Palecki, 2017: An evaluation of the North American regional reanalysis simulated soil moisture conditions during the 2011 to 2013 drought period. J. Hydrometeor., 18, 515–527, https://doi.org/10.1175/JHM-D-16-0132.1.
Leeper, R. D., J. E. Bell, and M. A. Palecki, 2019: A description and evaluation of U.S. Climate Reference Network standardized soil moisture dataset. J. Appl. Meteor. Climatol., 58, 1417–1428, https://doi.org/10.1175/JAMC-D-18-0269.1.
Leeper, R. D., B. Petersen, M. A. Palecki, and H. Diamond, 2021: Exploring the use of standardized soil moisture as a drought indicator. J. Appl. Meteor. Climatol., 60, 1021–1033, https://doi.org/10.1175/JAMC-D-20-0275.1.
Le Lay, M., and G. M. Saulnier, 2007: Exploring the signature of climate and landscape spatial variabilities in flash flood events: Case of the 8–9 September 2002 Cévennes-Vivarais catastrophic event. Geophys. Res. Lett., 34, L13401, https://doi.org/10.1029/2007GL029746.
Littell, J. S., D. L. Peterson, K. L. Riley, Y. Liu, and C. H. Luce, 2016: A review of the relationships between drought and forest fire in the United States. Global Change Biol., 22, 2353–2369, https://doi.org/10.1111/gcb.13275.
Narasimhan, B., and R. Srinivasan, 2005: Development and evaluation of soil moisture deficit index (SMDI) and evapotranspiration deficit index (ETDI) for agricultural drought monitoring. Agric. For. Meteor., 133, 69–88, https://doi.org/10.1016/j.agrformet.2005.07.012.
National Academies of Sciences, Engineering, and Medicine, 2015: Continuity of NASA Earth Observations from Space: A Value Framework. National Academies Press, 118 pp., https://doi.org/10.17226/21789.
National Drought Mitigation Center, 2022: U.S. Drought Monitor GIS data. NOAA/USDA, accessed 20 November 2022, https://droughtmonitor.unl.edu/DmData/GISData.aspx.
National Integrated Drought Information System, 2021a: Southeastern U.S. soil moisture network build-out and applications. NOAA/NIDIS Tech. Rep., 2 pp., https://www.drought.gov/documents/southeastern-us-soil-moisture-network-build-out-and-research.
National Integrated Drought Information System, 2021b: A strategy for the National Coordinated Soil Moisture Monitoring Network. NOAA/NIDIS Tech. Rep., 76 pp., https://www.drought.gov/documents/strategy-national-coordinated-soil-moisture-monitoring-network.
Norbiato, D., M. Borga, S. D. Esposti, E. Gaume, and S. Anquetin, 2008: Flash flood warning based on rainfall thresholds and soil moisture conditions: An assessment for gauged and ungauged basins. J. Hydrol., 362, 274–290, https://doi.org/10.1016/j.jhydrol.2008.08.023.
O, S., X. Hou, and R. Orth, 2020: Observational evidence of wildfire-promoting soil moisture anomalies. Sci. Rep., 10, 11008, https://doi.org/10.1038/s41598-020-67530-4.
Ochsner, T. E., and Coauthors, 2013: State of the art in large-scale soil moisture monitoring. Soil Sci. Soc. Amer. J., 77, 1888–1919, https://doi.org/10.2136/sssaj2013.03.0093.
Preimesberger, W., T. Scanlon, C.-H. Su, A. Gruber, and W. Dorigo, 2021: Homogenization of structural breaks in the global ESA CCI soil moisture multisatellite climate data record. IEEE Trans. Geosci. Remote Sens., 59, 2845–2862, https://doi.org/10.1109/TGRS.2020.3012896.
Reardon, J., R. Hungerford, and K. Ryan, 2007: Factors affecting sustained smouldering in organic soils from pocosin and pond pine woodland wetlands. Int. J. Wildland Fire, 16, 107–118, https://doi.org/10.1071/WF06005.
Rippey, B. R., 2015: The U.S. drought of 2012. Wea. Climate Extremes, 10, 57–64, https://doi.org/10.1016/j.wace.2015.10.004.
Scanlon, T., and Coauthors, 2022: ESA climate change initiative plus—Soil moisture. Algorithm Theoretical Baseline Document supporting product version 07.1 (Rep. D2.1 Version 3). Earth Observation Data Centre for Water Resources Monitoring, 68 pp., https://www.esa-soilmoisture-cci.org/sites/default/files/documents/public/CCI%20SM%20v07.1%20documentation/ESA_CCI_SM_RD_D2.1_v3_ATBD_v07.1_issue_1.0.pdf.
Scott, B. L., T. E. Ochsner, B. G. Illston, C. A. Fiebrich, J. B. Basara, and A. J. Sutherland, 2013: New soil property database improves Oklahoma Mesonet soil moisture estimates. J. Atmos. Oceanic Technol., 30, 2585–2595, https://doi.org/10.1175/JTECH-D-13-00084.1.
Smith, A. B., and R. W. Katz, 2013: US billion-dollar weather and climate disasters: Data sources, trends, accuracy and biases. Nat. Hazards, 67, 387–410, https://doi.org/10.1007/s11069-013-0566-5.
Torres, G. M., R. P. Lollato, and T. E. Ochsner, 2013: Comparison of drought probability assessments based on atmospheric water deficit and soil water deficit. Agron. J., 105, 428–436, https://doi.org/10.2134/agronj2012.0295.
U.S. Army Corps of Engineers, 2022: Upper Missouri River basin—Plains snow and soil moisture monitoring network. USACE, 4 pp., https://www.nwd-mr.usace.army.mil/rcc/reports/pdfs/umb_mon_network_factsheet.pdf.
Warter, M. M., M. B. Singer, M. O. Cuthbert, D. Roberts, K. K. Caylor, R. Sabathier, and J. Stella, 2021: Drought onset and propagation into soil moisture and grassland vegetation responses during the 2012–2019 major drought in Southern California. Hydrol. Earth Syst. Sci., 25, 3713–3729, https://doi.org/10.5194/hess-25-3713-2021.
Yun, H., 2021: Prediction model of algal blooms using logistic regression and confusion matrix. Int. J. Electr. Comput. Eng., 11, 2407–2413, https://doi.org/10.11591/ijece.v11i3.pp2407-2413.
