www.nature.

com/scientificreports

OPEN Interconnected hydrologic

extreme drivers and impacts
depicted by remote sensing data
assimilation
Timothy M. Lahmers 1,2*, Sujay V. Kumar 1, Kim A. Locke 1,2, Shugong Wang 1,3,
Augusto Getirana 1,3, Melissa L. Wrzesien 1,2, Pang‑Wei Liu 1,4 & Shahryar Khalique Ahmad 1,3
Hydrologic extremes often involve a complex interplay of several processes. For example, flood events
can have a cascade of impacts, such as saturated soils and suppressed vegetation growth. Accurate
representation of such interconnected processes while accounting for associated triggering factors
and subsequent impacts of flood events is difficult to achieve with conceptual hydrological models
alone. In this study, we use the 2019 flood in the Northern Mississippi and Missouri Basins, which
caused a series of hydrologic disturbances, as an example of such a flood event. This event began with
above-average precipitation combined with anomalously high snowmelt in spring 2019. This series
of anomalies resulted in above normal soil moisture that prevented crops from being planted over
much of the corn belt region. In the present study, we demonstrate that incorporating remote sensing
information within a hydrologic modeling system adds substantial value in representing the processes
that lead to the 2019 flood event and the resulting agricultural disturbances. This remote sensing
data infusion improves the accuracy of soil moisture and snowmelt estimates by up to 16% and 24%,
respectively, and it also improves the representation of vegetation anomalies relative to the reference
crop fraction anomalies.

In a changing climate, the likelihood of hydrologic extremes has been ­increasing1–3, as climate change can impact
both means and ­extremes4 of hydrologic cycle processes, potentially resulting in an increased frequency of floods
in some ­regions5,6 and decreases in ­others7. In a warming world, the physical processes that affect hydrologic
response, such as rain-on snow runoff events, are also changing, such that the seasonality of streamflow has been
­shifting8,9. The geography of rain-on-snow runoff events is predicted to move from low to high e­ levations10,11.
In addition to floods, there is also potential for an increase in dry extremes in a warming world with increased
drought frequency and occurrences in many parts of the ­world12–14. The increased frequency of drought and
heatwave events is expected to have consequences such as escalating crop failures in future projection ­scenarios1.
Thus, the consensus of literature shows that climate change is increasing the magnitude and frequency of extreme
hydrologic events, and the human influence in many of these events is substantial.
Understanding the drivers and impacts of extreme events in a warming climate requires models and observa-
tions of complex physical processes. Often, conceptual models alone are insufficient to resolve the heterogeneity
and the associated complexity of land surface processes. The significant human footprint and related impacts
further compound the difficulty of accurate characterization of the terrestrial water cycle. When available, obser-
vations, particularly from remote sensing, offer independent and spatially distributed information on such
processes, which can be used to constrain modeled estimates. This infusion of remote sensing data into a model
is typically achieved through data assimilation (DA) efforts that blend observational information with modeled
estimates by accounting for their relative errors and uncertainties.
There have been numerous efforts to incorporate observations of surface datasets of soil ­moisture15,16, ­snow17,
land surface t­ emperature18, and v­ egetation19–21 in land surface models (LSMs). These studies demonstrate that
surface DA can significantly improve the representation of related LSM states and fluxes. Several studies have
also examined the utility of remote sensing DA for hydrologic modeling. For example, bias corrected leaf area

1
Hydrological Sciences Lab, NASA Goddard Space Flight Center (NASA-GSFC), Greenbelt, MD, USA. 2Earth System
Science Interdisciplinary Center (ESSIC), University of Maryland, College Park, MD, USA. 3Science Applications
International Corporation (SAIC), McLean, VA, USA. 4Science Systems and Applications Inc. (SSAI), Lanham, MD,
USA. *email: timothy.lahmers@nasa.gov

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 1

Vol.:(0123456789)
 www.nature.com/scientificreports/

index (LAI) and evapotranspiration (ET) within the variable infiltration capacity (VIC) model improved the
characterization of streamflow estimates for a case study in the Connecticut River ­Basin22. Recent work has
demonstrated added value of Gravity Recovery And Climate Experiment (GRACE) terrestrial water storage
(TWS) for improving streamflow simulations in West ­Africa23. The benefit of assimilating snow products has
been widely demonstrated for both improving modelled snow water e­ quivalent24 (SWE) and for hydrologic
­response17,24–29. Many of these studies also documented the potential utility of DA of remote sensing measure-
ments for characterizing hydrological e­ xtremes30,31.
These efforts on land surface DA suggest that use of satellite remote sensing-based constraints in LSMs adds
substantial value over models a­ lone32,33. Note that much of the literature outlined above focuses on the assimi-
lation of univariate datasets, whereas similar multivariate and multi-sensor DA applications are less common.
There is a benefit of joint assimilation of soil moisture and vegetation data for realizing moisture and evapo-
rative fluxes on the land s­ urface20,34, as well as for drought m
­ onitoring35. The assimilation of L­ AI36 and snow
37
cover ­fraction combined with streamflow can also improve hydrologic model states compared to univariate
DA. Similarly, assimilation of LAI and surface soil moisture improves model estimates over that from a control
­simulation38. The assimilation of both soil moisture and snow datasets resulted in improvements to other model
process variables and for drought ­estimation39. Furthermore, joint assimilation of (Soil Moisture and Ocean
Salinity) SMOS soil moisture and GRACE TWS has been shown to produce more consistent improvements in
water cycle flux e­ stimation16.
These prior studies are focused on quantifying the aggregate impact of DA (with one or two datasets at most),
whereas the examination of the utility of DA for specific extreme events is lacking to the best of our knowledge.
Such quantifications are also important to highlight the utility of remote sensing information, and thus this work
is novel because we consider assimilation of three remote sensing datasets during an extreme event that resulted
in substantial financial losses. Extreme events such as floods and droughts often also have large economic impacts.
Therefore, large scale and timely information can be significantly more valuable in protecting property and the
populace. Here we consider such an event, the 2019 Mississippi River Basin flood, to demonstrate how DA of
multivariate and multi-sensor remote sensing datasets help in capturing the drivers and impacts of the flood.
Extreme events such as the 2019 Mississippi River Basin Flooding follow a complex triggering process, which
often includes a progression of anomalies in meteorological and hydrologic processes such as snowmelt, ante-
cedent soil moisture, and rainfall. This complexity also extends to impacts on soil moisture, evaporation, runoff,
and vegetation, which can be challenging to estimate given the limitations of ­models40. We demonstrate that
both advanced earth observing systems, in addition to LSMs and hydrologic models, are needed to adequately
capture the progression and consequences of such extreme events. Our LSM system is forced with remote sensing
based precipitation from the (Integrated Multi-satellitE Retrievals for GPM) IMERG system. Furthermore, we
use soil moisture retrievals from NASA’s Soil Moisture Active Passive (SMAP) m ­ ission41, LAI from the NASA
Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Terra and Aqua missions, and snow
depth retrievals from the Advanced Microwave Scanning Radiometer 2 (AMSR-2) mission for resolving the 2019
Mississippi basin flood event and its impacts. We consider the effects of the individual DA of each dataset on the
model, as well as multivariate DA by including all three datasets simultaneously. The present study makes use
of the NASA Land Information S­ ystem42,43 (LIS) with the Noah-MP v4.0.1 L ­ SM44 coupled to the Hydrological
Modeling and Analysis Platform (HyMAP) surface routing ­model45. Figure 1 shows a schematic of how various
DA components influence the modification of snow melt, soil moisture, and vegetation states. In this figure, the
model configuration is illustrated, and the impacts of specific DA instances are noted in larger red text. Physical
processes that were significant for this event are also noted in larger blue text.

Figure 1.  Conceptual illustration of the LSM and Routing Model with the impacts of DA. Physical processes
are labeled in blue, while the impacts of DA are labeled in red. This graphic was generated by the authors using
Adobe Illustrator.

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 2

Vol:.(1234567890)
www.nature.com/scientificreports/

In this article, we demonstrate how an LSM with DA of remote sensing information is able to resolve the driv-
ers and impacts of this event. The use of remote sensing data as both forcing and through data infusion to resolve
an extreme event makes this study novel. This article is organized as follows. Section "Observed progression
of 2019 Mississippi basin flood event" outlines the progression of the 2019 Mississippi Basin flooding through
observation datasets. In Section "Results and analysis", we demonstrate the results of our modeling study, includ-
ing the necessity for remote sensing constraints to resolve the full progression of the 2019 floods and impacts of
the DA on the hydrology. Section "Discussion and conclusion" includes a discussion of the implications of these
results for the hydrologic and remote sensing communities, including possible future work. Section "Methods"
describes the Open-Loop and DA LIS simulations, including the Ensemble Kalman Filter (EnKF) DA settings.

Observed progression of 2019 Mississippi basin flood event

Similar to other recent extreme events, the Mississippi Basin Flood event was preceded by both meteorological
and hydrologic anomalies (Fig. 2). The first major disturbance was an anomalous rainfall event that affected
much of the region in mid-March 2019. Rainfall estimates during the event, derived from the NASA IMERG
dataset, are shown in Fig. 2a (March 14–15). Note that the IMERG realization of this event is consistent with
the NLDAS2 precipitation ­estimates46, shown in Supplementary Fig. S1. NLDAS2 is considered a high quality
product over CONUS given that it also includes radar and gage derived p ­ recipitation46.
This extreme rainfall resulted in a rain-on-snow event with high values of snowmelt, from the abnormally high
snow amounts during the 2018–2019 winter across the basin. Snowmelt, which we represent here as change in
SWE, over the Corn Belt region (denoted by NOAA Climate Regions associated with corn crops; highlighted in
Fig. 3a) from the University of Arizona Snow ­Dataset47 is also shown in Fig. 2b. Prior studies have documented
that the University of Arizona Snow Dataset has high skill compared to several reference ­products48. For example,
it had a 0.98 correlation and 30% mean absolute deviation compared with the ASO d ­ ataset49. The University
of Arizona product was also found to have a 0.78 correlation and 31% mean absolute deviation with airborne
gamma SWE aircraft measurements over the North-Central ­US50. Note that we assume all negative changes in
SWE from the University of Arizona dataset are melt, rather than other processes such as compaction or sublima-
tion, as the University of Arizona dataset does not include a melt product. The spatial distribution of snowmelt
anomalies (also shown as loss of SWE) from March 10 to 29, 2019, compared to the years 2016–2020 (snowmelt
or loss of SWE is positive) from the University of Arizona Snow Dataset is shown in Fig. 2c. In Fig. 2c., the
March snowmelt is primarily concentrated over a domain from Eastern Nebraska through Northwest Indiana,
consistent with the Corn Belt Region (Fig. 3a.) and much of the March rainfall (Fig. 2a). Note that the Corn Belt
domain is based on NOAA Climate Regions in the Central US that are designated by the National Center for
Environmental Information (NCEI) as part of the Corn Belt (available online at: https://​www.​ncdc.​noaa.​gov/​
monit​oring-​refer​ences/​maps/​us-​ag-​belts.​php#​corn).
The snowmelt from the rain-on-snow combined with above average spring precipitation had a significant
contribution to the elevated soil moisture. Monthly soil moisture anomalies (compared to the full period from
2016 to 2020) from four US Climate Reference Network (USCRN) stations with complete data over the affected
region (Fig. 3a.) are shown in Fig. 2d, and they demonstrate the above average soil moisture for the region during
Spring 2019. These observations show wetter than average soil moisture during the typical planting and early
growing seasons, with some drying in late summer and into the fall season. Note that most of these gages are
located further south and west within the Corn Belt domain, where snowmelt (and likely increased soil moisture)
was less substantial.
This above average soil moisture prevented the planting of corn and soy crops throughout the Corn Belt region
and resulted in major crop losses and lower yields compared to prior years (Fig. 2e,f). This contributed to total
losses from the whole flood event of over $20 B ­ illion51. Figure 2e shows the spatial impacts of these losses. Here,
2019 crop fraction of both corn and soy, derived from USDA Cropland Data Layer (CDL) dataset, is plotted as
a standardized anomaly (i.e. [crop fraction – mean] / standard deviation) compared to crop years 2016–2020.
With the exception of much of Iowa, this panel shows that crop losses largely followed the snowmelt and extreme
precipitation over the corn belt region. Figure 2f shows prevented acres of corn and soy for the state of Illinois,
the core of the Corn Belt region affected by the 2019 flooding. This figure shows relative increases of prevented
acres (i.e., areas where crops were not grown), especially for corn crops but also to a lesser extent for soy.
Thus, the impacts and the progression of the 2019 flooding (starting from precipitation and snowmelt anoma-
lies to saturated soils and finally crop losses) are clearly shown in satellite and in-situ observations. In the next
section, we demonstrate how LSM simulations resolve this event with the infusion of information from remote
sensing.

Results and analysis
As demonstrated in Fig. 2 in the previous section, the 2019 Mississippi Basin flooding started with an extreme
precipitation event. The extreme precipitation combined with rapid snowmelt increased soil moisture to levels
that resulted in delays in—and some cases preventing—the planting of crops, causing crop losses throughout
the Corn Belt. Figure 3 shows time series of snowmelt, soil moisture anomalies, and LAI anomalies (com-
pared to 2016–2020 simulations) from all of the LIS simulations (OL, AMSR-DA, SMAP-DA, MODIS-DA,
and multivariate-DA), averaged over the Corn Belt region (Fig. 3a). Figure 3 demonstrates how these processes
and anomalies progressed throughout 2019 in five different LSM simulations, including the OL simulation initial-
ized on 1-January 1980, and with DA starting from the OL simulation on 1-April 2015. DA combinations include
AMSR-2 snow depth, SMAP soil moisture, MODIS LAI, and all three products as a multivariate DA simulation.
Consistent with observed UA snowmelt (Fig. 2b), Fig. 3b shows that nearly all the simulations (regardless of
DA) depict the spring snowmelt in March 2019. Table 1 shows skill scores for the time series of spatially averaged

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 3

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 2.  Observed Impacts of the 2019 Mississippi Basin Flood Event. IMERG Precipitation for March 14–15
(top left, a), University of Arizona SWE melt over the corn belt domain (top right, b), University of Arizona
(Dawson et al. 2016) SWE snowmelt anomalies (compared to 2016–2020) for 10 March to 29 March 2019
(middle left, c), USCRN soil moisture anomalies for four sites in the corn belt domain (middle right, d), 2019
Crop Area Fraction for Soy and Corn Standardized Anomalies based on the CropScape dataset (bottom left, e),
and USDA Farm Service Agency (FSA) prevented acres (bottom right, f) from 2015 through 2020 for the State
of Illinois are shown. Map panels were generated by Matlab version r2017b.

daily simulated SWE-melt compared to computed UA SWE melt, and there are statistically significantly improved
correlation coefficients (from 0.726 to 0.810) and reduced root mean square error (RMSE) by 24% in the DA

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 4

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 3.  Timeseries of modeled anomalies over the Corn Belt Region. The Corn Belt domain with USCRN
gages (black w/bold text) and plotted USGS gages (blue w/italic text) (a), snow melt (b), soil moisture
anomalies (c), and LAI anomalies (d) timeseries are shown. Panels (b) through (d) include the open-loop
(OL), AMSR-DA, SMAP-DA, MODIS-DA, and multivariate DA timeseries. Sample hydrographs for 2019 with
USGS observations, OL, AMSR-DA, SMAP-DA, MODIS-DA, and multivariate DA are shown for USGS gages
04185000 in Northern Ohio (e) and 05585000 in Illinois (f). The GIS panel was generated using ArcMap v.8.1.

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 5

Vol.:(0123456789)
 www.nature.com/scientificreports/

Skill metric LIS OL LIS AMSR-DA LIS SMAP-DA LIS MODIS-DA LIS MV-DA
2019 SWE melt
 Bias (mm/day) 0.019 (− 0.086 to 0.125) 0.023 (− 0.056 to 0.102) 0.018 (− 0.086 to 0.122) 0.017 (− 0.087 to 0.120) 0.024 (− 0.056 to 0.103)
 Correlation 0.726 (0.674 to 0.771) 0.811 (0.773–0.844)* 0.723 (0.670–0.769) 0.724 (0.671–0.770) 0.810 (0.772–0.843)*
 RMSE 1.021 (0.952–1.101) 0.767 (0.715–0.827)* 1.010 (0.942–1.089) 1.005 (0.937–1.084) 0.772 (0.720–0.832)*
USCRN 54902
 Bias (soil volumetric water
− 0.019 (− 0.021 to − 0.017) − 0.019 (− 0.021 to − 0.017) − 0.010 (− 0.012 to − 0.008)* − 0.006 (− 0.008 to − 0.004)* − 0.003 (− 0.005 to − 0.001)*
content)
 Correlation 0.814 (0.797–0.830) 0.815 (0.798–0.831) 0.812 (0.795–0.828) 0.795 (0.776–0.812) 0.797 (0.779–0.815)
 RMSE 0.043 (0.041–0.044) 0.043 (0.041–0.044) 0.041 (0.040–0.042)* 0.041 (0.040–0.043)* 0.041 (0.040–0.042)*
USCRN 54808
 Bias (soil volumetric water
− 0.019 (− 0.023 to − 0.016) − 0.020 (− 0.023 to − 0.016) − 0.010 (− 0.013 to − 0.007)* − 0.022 (− 0.025 to − 0.019) − 0.019 (− 0.022 to − 0.016)
content)
 Correlation 0.788 (0.769–0.805) 0.789 (0.770–0.806) 0.782 (0.763–0.800) 0.706 (0.681–0.729)* 0.705 (0.680–0.728)*
 RMSE 0.067 (0.065–0.070) 0.067 (0.065–0.070) 0.067 (0.065–0.070) 0.071 (0.069–0.074) 0.071 (0.068–0.073)
USCRN 63898
 Bias (soil volumetric water
− 0.028 (− 0.030 to− 0.025) − 0.028 (− 0.030 to − 0.025) − 0.024 (− 0.026 to − 0.022) − 0.026 (− 0.028 to − 0.023) − 0.025 (− 0.027 to − 0.023)
content)
 Correlation 0.824 (0.808–0.838) 0.823 (0.807–0.837) 0.832 (0.818–0.846) 0.815 (0.799–0.830) 0.816 (0.800–0.831)
 RMSE 0.060 (0.058–0.062) 0.060 (0.058–0.062) 0.058 (0.056–0.060) 0.056 (0.054–0.058) 0.056 (0.054–0.058)
USCRN 94995
 Bias (soil volumetric water
− 0.067 (− 0.070 to − 0.063) − 0.066 (− 0.070 to − 0.063) − 0.052 (− 0.056 to − 0.049)* − 0.054 (− 0.057 to − 0.051)* − 0.052 (− 0.055 to − 0.049)*
content)
 Correlation 0.585 (0.550–0.618) 0.587 (0.552–0.620) 0.605 (0.571–0.637) 0.716 (0.689–0.740)* 0.717 (0.691–0.741)*
 RMSE 0.093 (0.090–0.097) 0.093 (0.090–0.097) 0.083 (0.080–0.086)* 0.079 (0.076–0.082)* 0.078 (0.075–0.081)*

Table 1.  Model Skill Compared to Observations. Skill score metrics (bias, correlation coefficient, and RMSE)
for modeled 2019 LIS OL, AMSR-DA, SMAP-DA, MODIS-DA, and Multivariate-DA simulated SWE melt over
the Corn Belt domain compared to UA SWE melt are plotted in the top rows. July 2015–July 2020 0–10 cm
soil moisture from the same LIS simulations compared to 5-cm soil moisture at USCRN sites are listed in the
bottom rows. 95% confidence intervals for all metrics are shown. Confidence intervals for bias and RMSE use a
student’s t-test and a Chi-Square Distribution, respectively. Confidence intervals for correlation are computed
using a Fisher transform.

solutions (AMSR-2 and Multivariate) that include snow. This is despite slightly higher percent bias of SWE melt,
thus showing some added value of AMSR-2 snow DA.
The magnitude of the soil moisture anomalies is greater in the SMAP soil moisture DA and the multivariate
DA cases by late spring (Fig. 3c). The DA simulation also begins to capture drying conditions later in the summer,
with the progression of summer soil moisture anomalies consistent with station observations shown in Fig. 2d.
This shows that the OL simulation (and other simulations without soil moisture DA) tend to underestimate
the soil moisture anomalies from the 2019 flood event. The SMAP soil moisture DA solution shows the most
pronounced positive soil moisture anomalies, while the positive soil moisture anomalies in the multivariate solu-
tion are greater than the OL simulation but lower than the SMAP DA simulation. Table 1 shows skill scores for
Noah-MP surface soil moisture (i.e., 0–10 cm soil depth) compared to 5-cm soil moisture measurements from
the same four USCRN sites in Fig. 2d for 2016 through 2020. These results show reduced bias and RMSE at site
numbers 54902, 63898 and 94995 (statistically significant at sites 54902 and 94995). These reductions in bias and
RMSE are as high as 21% and 16%, respectively, at site number 94995. Statistically significant improvements to
correlation also occur at site 94995. DA slightly degrades model performance at site number 54808 (correlation
is significantly reduced).
As saturated soils resulted in crop losses, LAI anomalies, a proxy for changes in green vegetation, are shown
in Fig. 3d. In this panel, it is clear that the simulations without LAI DA fail to show decreased LAI over the Corn
Belt as LAI anomalies are positive. For the LAI and multivariate DA simulations, negative anomalies associated
with crop losses are present over the Corn Belt domain. The negative LAI anomaly for 2019 (compared to the
years 2016–2020) in the LAI DA and multivariate DA simulation shows that there is reduced vegetation in the
simulations with LAI DA, following the 2019 floods. This suggests that the model with LAI DA is able to capture
the crop losses, which were a result of human activity (not represented by Noah-MP without data assimilation),
as were shown previously from USDA data (Fig. 2). Note that other DA instances (i.e. SWE and soil moisture)
have little impact on the LAI timeseries, likely due to the role of dynamic vegetation in Noah-MP, which is not
as sensitive to the other modes of DA.
The effects of DA are also realized in the spatial anomalies (compared to 5-year averages from 2016 to 2020
when all datasets were available) of surface variables (Fig. 4). For SWE melt and soil moisture, anomalies are
compared to the average and not normalized (i.e., WY2019—Average), while LAI anomalies are normalized by
standard deviation (i.e., standardized anomalies). Absolute model values are also shown in supplementary Mate-
rial (Fig. S2). In Fig. 4a,b, snowmelt anomalies from March 10 to 29 are shown for the OL (left) and multivariate

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 6

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 4.  Spatial Extent of Modeled 2019 Mississippi Basin anomalies compared to averages from 2016 to 2020.
Snow melt anomalies (10 March–29 March) from the LIS-OL simulation (a), snow melt anomalies (10 March–
29 March) from the LIS-DA simulation (b), soil moisture anomalies (1 May–1 July) from the LIS-OL simulation
(c), soil moisture anomalies (1 May–1 July) from the LIS-DA simulation (d), LAI standardized anomalies (1
June–1 July) from the LIS-OL simulation (e), and LAI standardized anomalies (1 June–1 July) from the LIS-DA
simulation (f) are plotted. The color bars for the OL simulations (left) are the same as for the DA simulations
(right). Map panels were generated by Matlab version r2017b.

DA (right) simulations. Both simulations show a significant area of anomalously high snowmelt near the Corn
Belt area; however, the OL solution has more snowmelt than is depicted in the University of Arizona observa-
tions (shown in Fig. 2c). The multivariate DA simulation has reduced high bias, and the snowmelt extends
further west towards Nebraska, as in the observations. We also compared spatial similarity between modeled
and observed SWE melt using the structural similarity index (SSIM)52. SSIM between UA observations and the
LIS-OL simulation is 0.42, which increases to 0.44 with multivariate DA. Note that a gaussian smoother with
a sigma value of 2 was applied to the data to remove discontinuities in the data fields before computing SSIM.
Soil moisture anomalies for May and June 2019 are shown in Fig. 4c,d, and this Figure shows (as in Fig. 3) that
the multivariate DA simulation has more pronounced soil moisture anomalies over the Corn Belt compared to
the OL solution. The added value of DA is best realized by comparing the standardized LAI anomalies in Fig. 4e,f
to the change in crop fraction in Fig. 2e, as the losses of corn and soil crops shown in Fig. 2e mostly follow the
negative LAI anomalies in the multivariate solution in Fig. 4f. The SSIM between the LAI standardized anomalies

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 7

Vol.:(0123456789)
 www.nature.com/scientificreports/

compared to observed crop fraction standardized anomalies (Fig. 2e) is 0.06 in the OL simulation and is increased
to 0.07 with multivariate DA. Rather than depicting reduced LAI from crop losses, the OL solution shows posi-
tive LAI anomalies (Fig. 4e), thus demonstrating the added value of LAI DA. This inconsistency between the
OL and DA simulations reflects the lack of human processes (i.e., delayed or prevented planting) in Noah-MP.
This progression of extremes over the Corn Belt, examined both temporally (Fig. 3) and spatially (Fig. 4)
demonstrates how the cascade of hydrologic anomalies associated with the 2019 floods, which were realized by
observations in Fig. 2, are depicted in the LSM with multivariate DA. These results demonstrate that the model is
unable to resolve these processes without DA of relevant surface variables. Therefore, these results show the util-
ity of remote sensing products combined with the LSM model for depicting the cascade of hydrologic extremes
during this extreme flood event.
The impacts of the different DA variables and the full multivariate DA on streamflow are considered in
Supplementary Fig. S3. Figure S3 shows the Kling-Gupta-Efficiency53 (KGE) and percent bias skill scores of
the multivariate DA LIS-HyMAP streamflow compared to available USGS gages-II observations within the
Corn Belt domain. The analysis period for these metrics is 2015–2020. Note that only gages with limited human
modifications (i.e., reference basins) are considered. These results show that LIS-HyMAP has high skill scores
across the south and east parts of the Corn Belt; however, KGE drops below zero (indicating lower skill) further
west. In several areas, this is associated with both increased high and low biases. The middle and bottom pan-
els show changes in skill score (also KGE and bias) of the multivariate DA and soil moisture DA, respectively,
compared to the OL simulation. The changes in model skill are similar in both cases, and they demonstrate that
multivariate DA improves model skill, and most of this change is from soil moisture DA (see bottom panels of
Supplementary Fig. S1). Multivariate DA also increases model streamflow bias (in many cases this is negative
bias going to near zero), such that DA increases model streamflow (i.e., the OL simulations have negative bias
that is reduced from DA). Sample hydrographs (gages shown in Fig. 3a) from 2019 in two basins are plotted in
Fig. 3e,f, and an additional two basins are plotted in Supplementary Fig. S4. USGS gage 04185000 (Tiffin River
at Stryker, OH; Fig. 3e) and USGS gage 05585000 (La Moine River at Ripley, IL; Fig. 3f) have higher drainage
areas and associated with improved KGE following DA. For both of these basins, KGE is improved with multi-
variate DA, and negative bias is reduced. Therefore, the impacts of flooding in 2019 are more substantial in the
DA simulation. Note that some positive bias is added with multiple instance of DA. USGS gage 03346000 (North
Fork Embarras River near Oblong, IL; Supplementary Fig. S4) and USGS gage 05408000 (Kickapoo River at La
Farge, WI; Supplementary Fig. S4) have lower drainage areas (and less skill overall), but show similar patterns
of improved KGE following DA.
All of these solutions show a systematic tendency for LIS-HyMAP to produce too much baseflow, while
limiting surface runoff events. The choice of the TOPMODEL runoff scheme in Noah-MP (see details of LSM
settings in section "Methods") likely limits the skill of the streamflow ­simulations54,55. This tendency is par-
ticularly noticeable in higher order basins. These results are also consistent with the mixed KGE skill shown in
Supplementary Fig. S3.

Discussion and conclusion
This analysis demonstrates the added value of multivariate DA from remote sensing datasets on the simulation
of a hydrological extreme event; the 2019 Midwest floods. The improvements to model performance are dem-
onstrated by the reduction of soil moisture RMSE of up to 16% and snowmelt RMSE of up to 24% (Table 1) and
by the increases to structural similarity for model snow melt and LAI, compared to observed snow melt and
crop fraction anomalies.
The impacts on physical processes enabled by DA (illustrated in Fig. 1) demonstrate that the DA solution
can capture increased soil moisture anomalies and resultant flooding and crop losses. Comparison of the OL
simulations with the DA simulations suggests the OL simulation fails to capture the precise location of snow-
melt anomalies over the Midwest (Fig. 3), as well as the translation of snow and precipitation anomalies to soil
moisture and crop losses. The latter two biases (i.e., SWE and soil moisture) likely reflect systematic biases of
the Noah-MP LSM snow and soil moisture schemes; however, the lack of crop losses in the OL simulation are
due to agriculture impacts not being parameterized by Noah-MP. More broadly, this suggests that the modelled
relationships from extreme precipitation and snowmelt to soil moisture and the effect of saturated soil on crop
growth are not sufficiently resolved by the Noah-MP LSM. Given these differences between the OL and DA
results, this analysis demonstrates the benefits of remote sensing DA for use within modeling environments that
are used to inform management decisions around extreme events.
Given the limitations of the OL compared to the multivariate DA solution, these results demonstrate potential
benefits of future high resolution remote sensing measurements. The added value of different surface datasets
used in the DA framework corroborate prior work showing improvements to streamflow from snow and LAI
­DA22,27,29,56. Previous work has also demonstrated added value of LAI and soil moisture DA for resolving ET
anomalies over the continent of ­Australia34,38. Thus, the current study adds to the body of literature demonstrating
the benefits of remote sensing infrastructure for resolving the processes leading up to natural disasters, includ-
ing floods and droughts. This is potentially beneficial for improving physical process understanding through
models and for monitoring extreme events for operational stakeholders. Future work could extend the elements
of the present study by incorporating surface DA with LSMs to better understand physical processes associated
with hydrologic extremes such as ­floods30,31, ­fires57, and ­droughts13. Streamflow DA could also help to resolve
the impacts of diversions and other human modifications to hydrologic systems. Assimilation of data from the
recently launched NASA Surface Water and Ocean Topography (SWOT) mission could help to address this issue
in the future. Future work could also include varying LSMs to explore the limits of specific models and adjusting
LSM parameters and parameterizations to better capture hydrologic response during extreme events.

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 8

Vol:.(1234567890)
www.nature.com/scientificreports/

Methods
The NASA LIS system was used in the present study, as it supports ­DA58,59, multiple LSMs, and streamflow rout-
ing through the HyMAP routing m ­ odel45. The LIS LSM and DA system is well suited to resolve this and other
extreme events, as it has been successfully demonstrated in several DA scenarios, specifically in the assimilation
of surface soil m ­ oisture39,58,59, ­snow24–26,39, leaf area ­index20,57, and other surface v­ ariables18,57,60.
Version 4.0.1 of the Noah-MP ­LSM44 is used with dynamic vegetation, and the runoff scheme is TOPMODEL
with ­groundwater44. Noah-MP is run at 10-km grid resolution for the full Mississippi River basin domain
(28.55°–49.95° N, 77.95°–113.95° W). Noah-MP uses four soil levels with depths (from top to bottom) of 0.1, 0.3,
0.6 and 1.0 m. The OL simulation was spun up from 1 January 1980 and executed through 1 January 2021 with
MERRA2 atmospheric forcing and IMERG precipitation (when available starting in 2002). HyMAP streamflow
routing is added to the OL simulation starting 1 January 2000 and continued through 1 January 2021. As the
SMAP dataset does not begin until early 2015, the AMSR-2 DA, SMAP DA, and full multi-variate DA simulations
were all started on 1 April 2015 with initial conditions from the OL simulation. The MODIS-DA simulation was
initialized from the OL simulation on 1 July 2002, at the start of the MODIS LAI dataset.
LIS uses a 1-dimensional Ensemble Kalman F ­ ilter61 (EnKF) for DA. This DA method is used for all three data-
sets and the multivariate DA. EnKF has been widely used in prior s­ tudies61–66, and it enables the DA to account
for both model errors and non-linear tendencies within the model processes. EnKF additionally allows for the
characterization of model errors with an ensemble, and it is capable of resolving non-linear model dynamics and
discontinuities in temporal ­observations61. EnKF DA has also been demonstrated to add value to model variables
in a theoretical case study, where forcing and model biases were introduced, with snow and soil m ­ oisture65. The
EnKF process alternates between the model forecast timestep and the analysis update step. EnKF is executed
starting with the forecast timestep (i.e., running the LSM forward). In Eq. 1, let the forecast timestep be (k) and
ƒsu the prior analysis timestep be (k−1). Subscripts (i) and (j) are the x and y grid points in the model domain,
respectively. When the LSM is run, the model state (for any given variable) at a grid point is projected forward
from (  +
xk−1,i,j ) to the LSM state at k (  −
xk,i,j ). The analysis step is then computed as follows:
 
x+
k,i,j =  x−
k,i,j + K k,i,j  x−
y k,i,j − H k,i,j k

Here, the posterior state (  +

xk,i,j  ) combines the observation state (  y k,i,j ) and the a priori state (  −
xk,i,j ). In this
equation, H k,i,j represents the observation operator that relates  the model states to the observation. K k,i,j is the
Kalman gain, which weights the impact of forecast innovations  y k,i,j − H k,i,j  x−
k,i,j in the analysis update. Kalman
gain is based on the model and observation error covariances.
The DA methods and settings used here follow previous w ­ ork57, and the settings for forcing perturbations
and model variable perturbations are shown in Supplementary Table S1. The forcing perturbations are the same
for the snow depth, soil moisture, LAI, and multivariate DA simulations, and these settings follow prior applica-
tions of EnKF D ­ A59,67. Note that the state vectors for snow depth, soil moisture, and LAI, are all independent and
executed as separate instances of EnKF DA. All DA cases have 20 ensemble members. The 20-member ensemble
configuration is sufficient to avoid sampling errors for snow DA, and has been demonstrated in prior w ­ ork26.
The forcing state perturbation is 1-h, and the model state perturbation is 3-h. Observation state perturbations
depend on the data temporal resolution and are 3-h for soil moisture and snow depth, and 24-h for LAI. Per-
turbations to atmospheric forcing account for cross-correlation in space and time between variables, which can
better simulate the effects of model ­uncertainty25.
For snow depth DA, AMSR-2 microwave snow depth retrievals are used, and this is constrained by MODIS
snow cover. The Snow Depth and SWE variables of Noah-MP were updated for the AMSR-2 assimilation step.
Both of these variables are also perturbed in the DA step (see Table S1 in Supplementary Material). Previous
work demonstrated that this method could improve simulated snow states across North A ­ merica24. Furthermore,
the Noah-MP snow scheme adds value over the earlier Noah LSM, as it uses a 3-layer snow model that accounts
for the full energy balance of the snow ­pack44. This generally improves the performance of the Noah-MP LSM
for resolving ­snowpack68–72.
SMAP soil ­moisture20,41,57 is used for soil moisture assimilation. SMAP soil moisture is downscaled to 1-km
using the Thermal Hydraulic of Soil Moisture Disaggregation (THySM) a­ lgorithm73, which is also available online
through the USDA Crop-CASMA system (https://​nassg​eo.​csiss.​gmu.​edu/​CropC​ASMA/). When available, both
ascending (6 PM) and descending (6 AM) paths of data are used for assimilation. As THySM SMAP product is
based on surface soil moisture, the shallowest layer (0–10 cm) of Noah-MP is updated. The LIS “anomaly cor-
rection” algorithm is used to correct any biases in the SMAP ­product74.
The MCD15A2H Version 6 MODIS ­product75 is used for LAI assimilation. While the MODIS MCD15A2H
Version 6 product has an 8-day temporal resolution, a smoothing algorithm within the LIS system is used, so
that assimilation may be performed daily.
For all the model experiments, means and standard deviations for the model variables (including soil mois-
ture and LAI) are computed from 2016 through 2020. While we acknowledge this is a short analysis period, it
is necessary because of the limits of the DA datasets (SMAP is only available from April 2015 to present). For
analysis, SWE melt, soil moisture, and LAI are considered in 2019. SWE melt and soil moisture analysis use
anomalies (analysis year—mean), and LAI analysis uses standardized anomalies (analysis year–mean)/standard
deviation, respectively.
SSIM for SWE melt and LAI is computed using the SSIM function of the python scikit package (available
online at: https://s​ cikit-i​ mage.o
​ rg/d​ ocs/s​ table/a​ uto_e​ xampl​ es/t​ ransf​ orm/p ​ lot_s​ sim.h ​ tml), and gaussian smooth-
ing was performed using the ndimage.filters.gaussian_filter function of the python SciPy package (available
online at: https://d
​ ocs.s​ cipy.o
​ rg/d​ oc/s​ cipy-0.1​ 4.0/r​ efere​ nce/g​ enera​ ted/s​ cipy.n ​ dimag​ e.fi ​ lter​ s.g​ aussi​ an_fi
​ lter.h ​ tml).

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 9

Vol.:(0123456789)
 www.nature.com/scientificreports/

Note that this additional step was necessary since EnKF does not include smoothing when DA is performed,
and smoothing of model and observation data eliminates the influence local-scale discontinuities and errors in
the DA and observation based datasets.

Data availability
The NASA LIS Model is available online at: https://​github.​com/​NASA-​LIS/​LISF. The MERRA and IMERG
atmospheric forcing and precipitation datasets are located online at https://​disc.​gsfc.​nasa.​gov/​datas​ets?​proje​ct=​
MERRA-2 and https://​gpm.​nasa.​gov/​data/​direc​tory respectively. USGS streamflow data are stored at: https://​
water​servi​ces.​usgs.​gov/ and University of Arizona SWE data are available at: https://​nsidc.​org/​data/​nsidc-​0719/​
versio​ ns/1. USCRN soil moisture data are available online at https://w ​ ww.d
​ rough
​ t.g​ ov/d
​ ata-m
​ aps-t​ ools/u​ s-c​ lima​
te-​refer​ence-​netwo​rk-​uscrn-​soil-​tempe​rature-​and-​soil-​moist​ure, and USDA CropScape data are available at:
https://n​ assge​ odata.g​ mu.e​ du/C
​ ropSc​ ape/. The MODIS LAI product is available at: https://l​ pdaac.u
​ sgs.g​ ov/p​ rodu​
cts/​mcd15​a2hv0​06/. AMSR-2 snow depth data are available at: https://​nsidc.​org/​data/​au_​si12/​versi​ons/1, and
SMAP soil moisture are available at: https://​nsidc.​org/​data/​smap. The datasets used and/or analyzed during the
current study available from the corresponding author on reasonable request.

Received: 21 September 2022; Accepted: 23 February 2023

References
1. Raymond, C. et al. Increasing spatiotemporal proximity of heat and precipitation extremes in a warming world quantified by a
large model ensemble. Environ. Res. Lett. 17, 035005 (2022).
2. Chen, H. & Sun, J. Significant increase of the global population exposure to increased precipitation extremes in the future. Earths
Future 9, 1–16 (2021).
3. Prein, A. F. et al. The future intensification of hourly precipitation extremes. Nat. Clim. Chang 7, 48–52 (2017).
4. Prein, A. F. & Pendergrass, A. G. Can we constrain uncertainty in hydrologic cycle projections?. Geophys. Res. Lett. 46, 3911–3916
(2019).
5. Taylor, R. G. et al. Ground water and climate change. Nat. Clim. Chang 3, 322–329 (2013).
6. Salathé, E. P. et al. Estimates of twenty-first-century flood risk in the Pacific Northwest based on regional climate model simula-
tions. J. Hydrometeorol. 15, 1881–1899 (2014).
7. Hirabayashi, Y. et al. Global flood risk under climate change. Nat. Clim. Chang 3, 816–821 (2013).
8. Knighton, J. O., Degaetano, A. & Walter, M. T. Hydrologic state influence on riverine flood discharge for a small temperate water-
shed (Fall creek, united states): Negative feedbacks on the effects of climate change. J. Hydrometeorol. 18, 431–449 (2017).
9. Brunner, M. I., Melsen, L. A., Newman, A. J., Wood, A. W. & Clark, M. P. Future streamflow regime changes in the United States:
Assessment using functional classification. Hydrol. Earth Syst. Sci. 24, 3951–3966 (2020).
10. Musselman, K. N. et al. Projected increases and shifts in rain-on-snow flood risk over western North America. Nat. Clim. Chang
8, 808–812 (2018).
11. Li, D., Lettenmaier, D. P., Margulis, S. A. & Andreadis, K. The role of rain-on-snow in flooding over the conterminous United
States. Water Resour. Res. 55, 8492–8513 (2019).
12. Setegn, S. G., Rayner, D., Melesse, A. M., Dargahi, B. & Srinivasan, R. Impact of climate change on the hydroclimatology of Lake
Tana Basin, Ethiopia. Water Resour. Res. 47, 1–13 (2011).
13. Samaniego, L. et al. Anthropogenic warming exacerbates European soil moisture droughts. Nat. Clim. Chang 8, 421–426 (2018).
14. Zhang, P., Chen, G., Ma, W., Ming, Y. & Wu, Z. Robust atmospheric river response to global warming in idealized and compre-
hensive climate models. J. Clim. 34, 7717–7734 (2021).
15. Kolassa, J., Reichle, R. H. & Draper, C. S. Merging active and passive microwave observations in soil moisture data assimilation.
Remote Sens. Environ. 191, 117–130 (2017).
16. Girotto, M. et al. Multi-sensor assimilation of SMOS brightness temperature and GRACE terrestrial water storage observations
for soil moisture and shallow groundwater estimation. Remote Sens. Environ. 227, 12–27 (2019).
17. Margulis, S. A., Fang, Y., Li, D., Lettenmaier, D. P. & Andreadis, K. The utility of infrequent snow depth images for deriving con-
tinuous space-time estimates of seasonal snow water equivalent. Geophys. Res. Lett. 46, 5331–5340 (2019).
18. Reichle, R. H., Kumar, S. V., Mahanama, S. P. P., Koster, R. D. & Liu, Q. Assimilation of satellite-derived skin temperature observa-
tions into land surface models. J. Hydrometeorol. 11, 1103–1122 (2010).
19. Fairbairn, D. et al. The effect of satellite-derived surface soil moisture and leaf area index land data assimilation on streamflow
simulations over France. Hydrol. Earth Syst. Sci. 21, 2015–2033 (2017).
20. Kumar, S. V., Mocko, D. M., Wang, S., Peters-Lidard, C. D. & Borak, J. Assimilation of remotely sensed leaf area index into the
noah-mp land surface model: Impacts on water and carbon fluxes and states over the continental United States. J. Hydrometeorol.
20, 1359–1377 (2019).
21. Albergel, C. et al. Data assimilation for continuous global assessment of severe conditions over terrestrial surfaces. Hydrol. Earth
Syst. Sci. 24, 4291–4316 (2020).
22. Parr, D., Wang, G. & Bjerklie, D. Integrating remote sensing data on evapotranspiration and leaf area index with hydrological
modeling: Impacts on model performance and future predictions. J. Hydrometeorol. 16, 2086–2100 (2015).
23. Getirana, A. et al. Satellite gravimetry improves seasonal streamflow forecast initialization in Africa. Water Resour. Res. 56,
e2019WR026259 (2020).
24. Kumar, S. V. et al. Quantifying the added value of snow cover area observations in passive microwave snow depth data assimilation.
J. Hydrometeorol. 16, 1736–1741 (2015).
25. Liu, Y. et al. Assimilating satellite-based snow depth and snow cover products for improving snow predictions in Alaska. Adv.
Water Resour. 54, 208–227 (2013).
26. Liu, Y., Peters-Lidard, C. D., Kumar, S. V., Arsenault, K. R. & Mocko, D. M. Blending satellite-based snow depth products with
in situ observations for streamflow predictions in the Upper Colorado River Basin. Water Resour. Res. 51, 1182–1202 (2015).
27. Huang, C., Newman, A. J., Clark, M. P., Wood, A. W. & Zheng, X. Evaluation of snow data assimilation using the ensemble Kalman
filter for seasonal streamflow prediction in the western United States. Hydrol. Earth Syst. Sci. 21, 635–650 (2017).
28. Oaida, C. M. et al. A high-resolution data assimilation framework for snow water equivalent estimation across the Western United
States and validation with the Airborne Snow Observatory. J. Hydrometeorol. 20, 357–378 (2019).
29. Lahmers, T. M. et al. Assimilation of NASA’s airborne snow observatory snow measurements for improved hydrological modeling:
A case study enabled by the coupled LIS/WRF-Hydro system. Water Resour. Res. https://​doi.​org/​10.​1029/​2021W​R0298​67 (2022).
30. Margulis, S. A., Cortés, G., Girotto, M. & Durand, M. A Landsat-era Sierra Nevada snow reanalysis (1985–2015). J. Hydrometeorol.
17, 1203–1221 (2016).

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 10

Vol:.(1234567890)
www.nature.com/scientificreports/

31. el Gharamti, M. et al. Ensemble streamflow data assimilation using WRF-Hydro and DART: Novel localization and inflation
techniques applied to Hurricane Florence flooding. Hydrol. Earth Syst. Sci. 25, 5315–5336 (2021).
32. Lettenmaier, D. P. et al. Inroads of remote sensing into hydrologic science during the WRR era. Water Resour. Res. 51, 7309–7342
(2015).
33. Durand, M. et al. Achieving breakthroughs in global hydrologic science by unlocking the power of multisensory, multidisciplinary
earth observations. AGU Adv. 2, 1–13 (2021).
34. Kumar, S. V., Holmes, T. R., Bindlish, R., de Jeu, R. & Peters-Lidard, C. Assimilation of vegetation optical depth retrievals from
passive microwave radiometry. Hydrol. Earth Syst. Sci. 24, 3431–3450 (2020).
35. Gavahi, K., Abbaszadeh, P., Moradkhani, H., Zhan, X. & Hain, C. Multivariate assimilation of remotely sensed soil moisture and
evapotranspiration for drought monitoring. J. Hydrometeorol. 21, 2293–2308 (2020).
36. Mohammadi Igder, O., Alizadeh, H., Mojaradi, B. & Bayat, M. Multivariate assimilation of satellite-based leaf area index and
ground-based river streamflow for hydrological modelling of irrigated watersheds using SWAT+. J Hydrol Amst 610, 128012 (2022).
37. Bayat, M., Alizadeh, H. & Mojaradi, B. SWAT_DA: Sequential multivariate data assimilation-oriented modification of SWAT.
Water Resour. Res. 58, e2022WR032397 (2022).
38. Barbu, A. L., Calvet, J. C., Mahfouf, J. F. & Lafont, S. Integrating ASCAT surface soil moisture and GEOV1 leaf area index into the
SURFEX modelling platform: A land data assimilation application over France. Hydrol. Earth Syst. Sci. 18, 173–192 (2014).
39. Kumar, S. V. et al. Assimilation of remotely sensed soil moisture and snow depth retrievals for drought estimation. J. Hydrometeorol.
15, 2446–2469 (2014).
40. Lafferty, D. C. et al. Statistically bias-corrected and downscaled climate models underestimate the adverse effects of extreme heat
on U.S. maize yields. Commun. Earth Environ. 2, 1–10 (2021).
41. Entekhabi, D. et al. The soil moisture active passive (SMAP) mission. Proc. IEEE 98, 704–716 (2010).
42. Kumar, S. et al. Land information system: An interoperable framework for high resolution land surface modeling. Environ. Model.
Softw. 21, 1402–1415 (2006).
43. Peters-Lidard, C. D. et al. High-performance earth system modeling with NASA/GSFC’s land information system. Innov. Syst.
Softw. Eng. 3, 157–165 (2007).
44. Niu, G.-Y. et al. The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description
and evaluation with local-scale measurements. J Geophys Res 116, D12109 (2011).
45. Getirana, A., Peters-Lidard, C., Rodell, M. & Bates, P. D. Trade-off between cost and accuracy in large-scale surface water dynamic
modeling. Water Resour. Res. 53, 4942–4955 (2017).
46. Xia, Y. et al. Continental-scale water and energy flux analysis and validation for the North American Land Data Assimilation
System project phase 2 (NLDAS-2): 1. Intercomparison and application of model products. J Geophys Res. Atmos. 117, n/a–n/a
(2012).
47. Dawson, N. et al. An evaluation of snow initializations in NCEP global and regional forecasting models. J. Hydrometeorol. 17,
1885–1901 (2016).
48. Broxton, P. D., Zeng, X. & Dawson, N. Why do global reanalyses and land data assimilation products underestimate snow water
equivalent. J. Hydrometeorol. 17, 2743–2761 (2016).
49. Dawson, N., Broxton, P. & Zeng, X. Evaluation of remotely sensed snow water equivalent and snow cover extent over the contigu-
ous United States. J. Hydrometeorol. 19, 1777–1791 (2018).
50. Cho, E., Jacobs, J. M. & Vuyovich, C. M. The value of long-term (40 years) airborne gamma radiation SWE record for evalu-
ating three observation-based gridded SWE data sets by seasonal snow and land cover classifications. Water Resour Res 56,
e2019WR025813 (2020).
51. Shirzaei, M. et al. Persistent impact of spring floods on crop loss in U.S. Midwest. Weather Clim Extrem 34, 100392 (2021).
52. Wang, Z., Bovik, A. C., Sheikh, H. R. & Simoncelli, E. P. Image quality assessment: From error visibility to structural similarity.
IEEE Trans. Image Process. 13, 600–612 (2004).
53. Gupta, H. V., Kling, H., Yilmaz, K. K. & Martinez, G. F. Decomposition of the mean squared error and NSE performance criteria:
Implications for improving hydrological modelling. J. Hydrol. (Amst) 377, 80–91 (2009).
54. Schaake, J. C., Koren, V. I., Duan, Q. Y., Mitchell, K. & Chen, F. Simple water balance model for estimating runoff at different spatial
and temporal scales. J. Geophys. Res. Atmos. 101, 7461–7475 (1996).
55. Lahmers, T. M. et al. Evaluation of NOAA national water model parameter calibration in semi-arid environments prone to channel
infiltration. J. Hydrometeorol. https://​doi.​org/​10.​1175/​JHM-D-​20-​0198.1 (2021).
56. Dechant, C. & Moradkhani, H. Radiance data assimilation for operational snow and streamflow forecasting. Adv. Water Resour.
34, 351–364 (2011).
57. Kumar, S. V. et al. The 2019–2020 Australian drought and bushfires altered the partitioning of hydrological fluxes. Geophys. Res.
Lett. 48, e2020GL091411 (2021).
58. Kumar, S. V., Reichle, R. H., Koster, R. D., Crow, W. T. & Peters-Lidard, C. D. Role of subsurface physics in the assimilation of
surface soil moisture observations. J. Hydrometeorol. 10, 1534–1547 (2009).
59. Peters-Lidard, C. D., Kumar, S. V., Mocko, D. M. & Tian, Y. Estimating evapotranspiration with land data assimilation systems.
Hydrol. Process. 25, 3979–3992 (2011).
60. Kumar, S. V. et al. Assimilation of gridded GRACE terrestrial water storage estimates in the North American land data assimilation
system. J. Hydrometeorol. 17, 1951–1972 (2016).
61. Reichle, R. H., McLaughlin, D. B. & Entekhabi, D. Hydrologic data assimilation with the ensemble Kalman filter. Mon. Weather
Rev. 130, 103–114 (2002).
62. Houtekamer, P. L. & Mitchell, H. L. Data assimilation using an ensemble Kalman filter technique. Mon. Weather Rev. 126, 796–811
(1998).
63. Zhou, Y., McLaughlin, D. & Entekhabi, D. Assessing the performance of the ensemble Kalman filter for land surface data assimila-
tion. Mon. Weather Rev. 134, 2128–2142 (2006).
64. Pan, M. & Wood, E. F. Data assimilation for estimating the terrestrial water budget using a constrained ensemble Kalman filter. J.
Hydrometeorol. 7, 534–547 (2006).
65. Kumar, S. V. et al. A land surface data assimilation framework using the land information system: Description and applications.
Adv. Water Resour. 31, 1419–1432 (2008).
66. Hain, C. R., Crow, W. T., Anderson, M. C. & Mecikalski, J. R. An ensemble Kalman filter dual assimilation of thermal infrared and
microwave satellite observations of soil moisture into the Noah land surface model. Water Resour Res 48 (2012).
67. Su, H., Yang, Z.-L., Dickinson, R. E., Wilson, C. R. & Niu, G.-Y. Multisensor snow data assimilation at the continental scale: The
value of gravity recovery and climate experiment terrestrial water storage information. J. Geophys. Res. 115, D10104 (2010).
68. Chen, F. et al. Modeling seasonal snowpack evolution in the complex terrain and forested Colorado Headwaters region: A model
intercomparison study. J. Geophys. Res. Atmos. 119, 13795–13819 (2014).
69. Chen, F., Liu, C., Dudhia, J. & Chen, M. A sensitivity study of high-resolution regional climate simulations to three land surface
models over the western United States. J. Geophys. Res. Atmos. 119, 7271–7291 (2014).
70. Wrzesien, M. L., Pavelsky, T. M., Kapnick, S. B., Durand, M. T. & Painter, T. H. Evaluation of snow cover fraction for regional
climate simulations in the Sierra Nevada. Int. J. Climatol. 35, 2472–2484 (2015).

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 11

Vol.:(0123456789)
 www.nature.com/scientificreports/

71. Minder, J. R., Letcher, T. W. & Skiles, S. M. An evaluation of high-resolution regional climate model simulations of snow cover
and albedo over the Rocky Mountains, with implications for the simulated snow-albedo feedback. J. Geophys. Res. Atmos. 121,
9069–9088 (2016).
72. Kim, R. S. et al. Snow Ensemble Uncertainty Project (SEUP): quantification of snow water equivalent uncertainty across North
America via ensemble land surface modeling. Cryosphere 15, 771–791 (2021).
73. Liu, P.-W. et al. Thermal hydraulic disaggregation of SMAP soil moisture over the continental United States. IEEE J. Sel. Top. Appl.
Earth Obs. Remote Sens. 15, 4072–4093 (2022).
74. Kwon, Y. et al. Irrigation characterization improved by the direct use of SMAP soil moisture anomalies within a data assimilation
system. Environ. Res. Lett. 17, 084006 (2022).
75. Myneni, R., Knyazikhin, Y. & Park, T. MCD15A2H MODIS/Terra+Aqua Leaf Area Index/FPAR 8-day L4 Global 500m SIN Grid
V006. (2015).

Acknowledgements
This research was supported by funding from NASA Headquarters for the Earth Information System (EIS) pilot
study. Computing was supported by the resources at the NASA Center for Climate Simulation (NCCS) at NASA
Goddard Space Flight Center.

Author contributions
T.L. designed the experiment questions and executed the model evaluation and data analysis, with guidance
from S.K. The model configuration for this experiment was set up by S.K., S.W., and A.G. The data assimilation
configurations were developed by S.K., M.W., P.-W.L. and T.L. The streamflow post-processing code was written
by S.A., with assistance from A.G. P.-W.L. post-processed the USCRN soil moisture data. K.L. provided guidance
to the research team on the impacts of the Mississippi basin flood event, particularly for agriculture. All authors
contributed to the preparation of the manuscript.

Competing interests 
The authors declare no competing interests.

Additional information
Supplementary Information The online version contains supplementary material available at https://​doi.​org/​
10.​1038/​s41598-​023-​30484-4.
Correspondence and requests for materials should be addressed to T.M.L.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access  This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.

© The Author(s) 2023

Scientific Reports | (2023) 13:3411 | https://doi.org/10.1038/s41598-023-30484-4 12

Vol:.(1234567890)