Performance of Seven Land Surface Schemes in the WRFv4.3 Model for Simulating Precipitation in the Record‐Breaking Meiyu Season Over the Yangtze–Huaihe River Valley in China

Abstract

In 2020, the Yangtze-Huai river valley (YHRV) experienced the highest record-breaking Meiyu season since 1961, which was mainly characterized by the longest duration of precipitation lasting from early-June to mid-July, with frequent heavy rainstorms that caused severe flooding and deaths in China. Many studies have investigated the causes of this Meiyu season and its evolution, but the accuracy of precipitation simulations has received little attention. It is important to provide more accurate precipitation forecasts to help prevent and reduce flood disasters, thereby facilitating the maintenance of a healthy and sustainable earth ecosystem. In this study, we determined the optimal scheme among seven land surface model (LSMs) schemes in the Weather Research and Forecasting model for simulating the precipitation in the Meiyu season during 2020 over the YHRV region. We also investigated the mechanisms in the different LSMs that might affect precipitation simulations in terms of water and energy cycling. The results showed that the simulated amounts of precipitation were higher under all LSMs than the observations. The main differences occurred in rainstorm areas (>12 mm/day), and the differences in low rainfall areas were not significant (<8 mm/day). Among all of the LSMs, the Simplified Simple Biosphere (SSiB) model obtained the best performance, with the lowest root mean square error and the highest correlation. The SSiB model even outperformed the Bayesian model averaging result. Finally, some factors responsible for the differences modeling results were investigated to understand the related physical mechanism.

Full text

1. Introduction
China experiences some of the heaviest rainstorms among the countries in East Asia. Flood disasters caused by
rainstorms occur frequently every few years, and one of the areas affected most severely by floods is the Yangtze–
Huaihe river valley (YHRV), where about 40% of the total flood disasters in China occur (Huang etal.,2021; Wu
etal.,2021). In particular, flood disasters often occur in the Meiyu season, which usually lasts for two to 3weeks
in June and July each year.
A long Meiyu period can lead to flood disasters. In particular, the Meiyu season during 2020 in YHRV was the
worst since 1961, where it was mainly characterized by the longest duration of precipitation lasting from early-
June to mid-July, with frequent heavy rainstorms that caused severe flooding and deaths in China. According to
statistics from the Department of Emergency Management, the floods during the record-breaking Meiyu season
in 2020 affected 54.81 million people, where 158 people were dead or missing, and the affected area of crops was
Abstract In 2020, the Yangtze–Huai river valley (YHRV) experienced the highest record-breaking Meiyu
season since 1961, which was mainly characterized by the longest duration of precipitation lasting from
early-June to mid-July, with frequent heavy rainstorms that caused severe flooding and deaths in China. Many
studies have investigated the causes of this Meiyu season and its evolution, but the accuracy of precipitation
simulations has received little attention. It is important to provide more accurate precipitation forecasts to
help prevent and reduce flood disasters, thereby facilitating the maintenance of a healthy and sustainable earth
ecosystem. In this study, we determined the optimal scheme among seven land surface model (LSMs) schemes
in the Weather Research and Forecasting model for simulating the precipitation in the Meiyu season during
2020 over the YHRV region. We also investigated the mechanisms in the different LSMs that might affect
precipitation simulations in terms of water and energy cycling. The results showed that the simulated amounts
of precipitation were higher under all LSMs than the observations. The main differences occurred in rainstorm
areas (>12mm/day), and the differences in low rainfall areas were not significant (<8mm/day). Among all
of the LSMs, the Simplified Simple Biosphere (SSiB) model obtained the best performance, with the lowest
root mean square error and the highest correlation. The SSiB model even outperformed the Bayesian model
averaging result. Finally, some factors responsible for the differences modeling results were investigated to
understand the related physical mechanism.
Plain Language Summary The record-breaking precipitation that occurred in the Meiyu season
during 2020 significantly affected humans and ecosystems. Can using different land surface model (LSMs)
schemes affect the accuracy of precipitation simulations in numerical weather prediction models? If this is the
case, how does their performance affect predictions of critical water and energy processes? In this study, we
found that simulations using all of the LSMs overestimated the observations, but Simplified Simple Biosphere
(SSiB) performed relatively better, with the lowest root mean square error and highest correlation. SSiB
performed better than the multi-model ensemble averaging model with all seven LSMs.
DI ETAL.
© 2023 The Authors. GeoHealth
published by Wiley Periodicals LLC on
behalf of American Geophysical Union.
This is an open access article under
the terms of the Creative Commons
Attribution-NonCommercial-NoDerivs
License, which permits use and
distribution in any medium, provided the
original work is properly cited, the use is
non-commercial and no modifications or
adaptations are made.
Performance of Seven Land Surface Schemes in the WRFv4.3
Model for Simulating Precipitation in the Record-Breaking
Meiyu Season Over the Yangtze–Huaihe River Valley in China
Zhenhua Di1 , Shenglei Zhang2 , Jiping Quan3 , Qian Ma4 , Peihua Qin5 , and Jianduo Li6
1State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal
University, Beijing, China, 2State Key Laboratory of Remote Sensing Science, Aerospace Information Research Institute,
Chinese Academy of Sciences, Beijing, China, 3Institute of Urban Meteorology, China Meteorological Administration,
Beijing, China, 4College of Global Change and Earth System Science, Beijing Normal University, Beijing, China, 5State
Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of
Atmospheric Physics, Chinese Academy of Sciences, Beijing, China, 6CMA Earth System Modeling and Prediction Centre,
China Meteorological Administration, Beijing, China
Key Points:
• Seven land surface schemes in the
WRFv4.3 model were evaluated
to obtain the optimal scheme for
precipitation simulation
• Simulated rainfall amounts under all
land surface models were generally
higher than observations, but
Simplified Simple Biosphere (SSiB)
performed best
• Water heat flux, geopotential height,
and moisture flux convergence all
indicated the superiority of SSiB at
precipitation simulation
Correspondence to:
S. Zhang,
zhangsl@aircas.ac.cn
Citation:
Di, Z., Zhang, S., Quan, J., Ma, Q., Qin,
P., & Li, J. (2023). Performance of seven
land surface schemes in the WRFv4.3
model for simulating precipitation in the
record-breaking Meiyu season over the
Yangtze–Huaihe river valley in China.
GeoHealth, 7, e2022GH000757. https://
doi.org/10.1029/2022GH000757
Received 26 NOV 2022
Accepted 12 FEB 2023
Author Contributions:
Conceptualization: Shenglei Zhang
Formal analysis: Zhenhua Di, Jiping
Quan, Peihua Qin, Jianduo Li
Funding acquisition: Shenglei Zhang
Methodology: Zhenhua Di, Peihua Qin
Resources: Zhenhua Di
Software: Peihua Qin
Supervision: Shenglei Zhang
Validation: Qian Ma
Visualization: Qian Ma, Jianduo Li
Writing – original draft: Zhenhua Di
10.1029/2022GH000757
RESEARCH ARTICLE
1 of 17

GeoHealth
DI ETAL.
10.1029/2022GH000757
2 of 17
5.28 million ha, resulting in direct economic losses of 144.43 billion yuan (Xinhua NEWs,2020). Therefore, it
is important to accurately predict the start and end time of the Meiyu in YHRV as well as the rainfall amounts to
facilitate national flood control and disaster reduction.
Many studies have investigated the record-breaking Meiyu in YHRV during 2020 as an example to understand
its causes and evolution (Chen etal.,2022; Ding etal.,2021; Liu & Ding,2020; Qiao etal.,2021). The extreme
Meiyu may have been related to the North Atlantic Oscillation, East Asian monsoon circulation oscillation, and
the combined effect of Indian Ocean warming and Arctic sea ice anomalies. However, few studies have investi-
gated the accuracy of rainfall simulation using different numerical weather prediction (NWP) models, especially
in different land surface models (LSMs).
The land surface is an important and complex underlying surface at the lower boundary of the atmosphere. Phys-
ical exchanges such as those of momentum, energy, and water vapor between the land surface and atmosphere
strongly influence the atmospheric circulation and climate change (Manabe,1969). LSMs are used to describe
the detailed surface characteristics and accurate exchange fluxes in order to provide a reasonable lower boundary
condition for NWP models. Therefore, the accuracy of LSMs is important for the performance of NWP models.
LSMs have developed in three stages from the first-generation “bucket” model to the third-generation biochemi-
cal model, but they are still affected by some uncertainties due to limitations on observations and cognition (van
den Hurk etal.,2016).
Two common methods are used to improve the performance of LSMs. The first method involves developing more
detailed representations of physical land processes. For instance, Friend and Kiang(2005) added descriptions of
the canopy stomatal conductance and photosynthesis to a LSM, which reduced the precipitation bias by about
1mm/day over South America. Yuan etal.(2008) implemented a groundwater dynamic model in the Regional
Climate Model Version 3 (RegCM3) and found that the simulated precipitation bias was reduced by about 40%
over semiarid and humid regions. The second commonly used method involves selecting an optimal LSM from
numerous options. For instance, Constantinidou etal.(2020) assessed the effectiveness of four LSMs at simulat-
ing precipitation at relatively coarse horizontal resolution of 50km by using the Weather Research and Forecast-
ing (WRF) model over the Middle East–North Africa region, and found that Noah obtained the best performance
on the overall domain, followed by its augmented version Noah-MP. Attada etal.(2018) and Reddy etal.(2020)
assessed the effectiveness of three LSMs at simulating the main climatic variables such as precipitation during
the Indian summer monsoon season. In addition, some uncertainty qualification methods such as parameter opti-
mization (Li etal.,2021; Zhang etal.,2020) and ensemble forecasting (Li etal.,2017; Liu etal.,2016) have been
used widely to improve the performance of LSMs.
Previous studies demonstrated that the choice of LSM can improve simulations of precipitation. However, some
issues have not been addressed in previous studies. First, the number of LSMs evaluated was low (usually three
to four) due to limitations on previous versions of the WRF model or a lack of computing resources (Attada
etal.,2018; Constantinidou etal.,2020). At present, seven optional LSMs can be used to support simulations in
the latest WRF (version 4.3). Clearly, including more LSMs in evaluations will be helpful for identifying the best
at accurately simulating precipitation in the NWP model. Second, the Meiyu season is one of the main causes of
flood disasters in East Asia. Many studies have evaluated the effectiveness of different LSMs in various simu-
lations of precipitation (García-García etal.,2022; Yan etal.,2021; Zhong etal.,2021), but few have evaluated
LSMs for simulating precipitation in the record-breaking Meiyu season. In the present study, we assessed seven
LSMs in the latest WRF model (version 4.3) to determine the LSM that obtained the most reasonable simulations
of precipitation during the Meiyu event in 2020 over the YHRV region of East China, which was the worst Meiyu
event since 1961. It is important to recommend the most effective LSM for the NWP model to obtain accurate
simulations of precipitation in Meiyu events in order to prevent and reduce the effects of flood disasters to main-
tain a healthy and sustainable ecosystem.
2. Data and Models
2.1. Precipitation Observations
Precipitation observations were obtained from China's hourly merged precipitation analysis products
(CHMPA-Hourly, version 1.0), which were combined from more than 30,000 automatic weather stations in China
and the Climate Precipitation Center Morphing (CMORPH) satellite precipitation remote sensing product at a

GeoHealth
DI ETAL.
10.1029/2022GH000757
3 of 17
horizontal resolution of 8km covering the area between 60°S and 60°N (Shen etal.,2014). The CHMPA precipi-
tation product has a spatial resolution of 0.1°×0.1° and temporal resolution of 1hr. Daily precipitation data were
used to avoid the influence of time deviations.
2.2. Mesoscale NWP Models and WRF Model
NWP models objectively and quantitatively forecast future weather processes by numerically solving the fluid
mechanics and thermodynamics equations for processes on land, in the atmosphere, and their interactions (Pu
& Kalnay,2018). NWP models are usually implemented with supercomputers. NWP models can be divided
into global and regional climate models according to differences in their spatial scale. In particular, regional
climate models can obtain more accurate predictions due to their high-resolution and spatio-temporal solution of
weather processes. The commonly used mesoscale NWP models mainly include PSU/NCAR mesoscale model
version 5 (MM5; Chen & Dudhia,2001), Global/Regional Assimilation and Prediction System (GRAPES; Zhong
etal., 2019), and WRF (Skamarock etal., 2021). WRF is one of most widely used NWP models in previous
studies.
WRF model version 4.3 (Skamarock etal., 2021; available from http://www2.mmm.ucar.edu./wrf/users) was
applied to the YHRV region of China. The simulated domain comprised double-layered nested regions (Figure1).
Horizontally, the outer layer (D01) had 100×82 grid points with a grid spacing of 27km, and the inner layer
(D02) had 154×85 grid points with a grid spacing of 9km. The inner layer D02 represented the YHRV region
(109.4–123.2°E, 28.3–34.6°N). The terrain in this region is dominated by low hills and plains, and most are
below 80m above sea level. The land types mainly include paddy land, with some grassland and woodland. The
land use/land cover data set required by the WRF simulation was obtained by interpolating the land use categories
from United States Geological Survey 24-category data, which is applicable to all alternative LSMs in WRF. The
spatial resolutions of the outer and inner layers were 10 and 2m, respectively. Forty five uniform vertical layers
were considered from the land surface to the 50hPa level in the atmosphere. The uniform time step was 90s.
The physics options included the WSM 6-class microphysics scheme (Hong & Lim,2006), Kain–Fritsch Eta
cumulus scheme (Kain,2004), Rapid Radiative Transfer Model (RRTM)long-wave radiation scheme (Mlawer
et al., 1997), Dudhia short-wave radiation scheme (Dudhia et al.,1999), Yonsei University (YSU) planetary
boundary layer scheme (Hong,2010), and the land surface scheme (or LSM) being evaluated. This suite of phys-
ics combinations except for LSM followed the operational setup from the China Meteorological Administration.
Figure 1. Weather Research and Forecasting simulation domain with double-layered nested grids, where D01 represents the
outer layer region with a resolution of 27km and the inner layer region (D02) with a resolution of 9km. HaN, HB, and HN
represent Henan, Hubei, and Hunan provinces, respectively, and JS, AH, JX, and ZJ represent Jiangsu, Anhui, Jiangxi, and
Zhejiang provinces, respectively. The topography (unit: m) is also shown on the map.

GeoHealth
DI ETAL.
10.1029/2022GH000757
4 of 17
The final analysis data (FNL) of the global forecasting system from the National Center for Environmental
Prediction available at a horizontal resolution of 1°× 1° and 6 hr intervals were used to generate the initial
and lateral boundary conditions. The simulation period lasted 33days from 18:00 on June 10 to 00:00 on July
13 in 2020. The first six hourly simulations were discarded as the spin-up time. Seven numerical experiments
were performed by using the WRF model with seven different LSMs comprising the 5-layer thermal diffusion
scheme (Slab; Dudhia,1996), Noah scheme (Noah; Chen & Dudhia,2001), LSM in the NOAA Rapid Update
Cycle (RUC) weather prediction model (referred to as RUC in the LSM options for the WRF model; Smirnova
etal.,2016), Noah multi-parameterization (MP) scheme (Noah-MP; Niu etal.,2011), Community Land Model
version 4 scheme (CLM4; Oleson etal.,2010), Pleim-Xiu scheme (Pleim-Xiu; Pleim & Xiu,1995), and Simpli-
fied Simple Biosphere scheme (SSiB; Xue etal.,1991).
2.3. LSMs
In this study, we evaluated seven LSMs available in the WRFv4.3 model comprising Slab, Noah, RUC,
Noah-MP, CLM4, Pleim-Xiu, and SSiB, where their soil and vegetation features are listed in Table1. Some
of the most significant differences among these models are summarized as follows. (a) Model complexity:
Slab is the simplest due to the excessive simplification of physical process descriptions, whereas CLM4 is the
most complicated because of the complex parameterization formula and multiple parameters. The complexity
of Noah-MP following CLM4 is due to the use of multiple optional parameterization models for each of main
physical processes. The other LSMs have moderate complexity. (b) Solutions for soil moisture and tempera-
ture: the soil moisture is a simple function in the Slab model and it is related to the land use type, while the soil
temperature is derived by using the thermal diffusion equation with the upper boundary of diurnal variation.
The soil moisture and temperature solutions for the LSMs comprising Noah, Noah-MP, RUC, and CLM4 are
based on Richard moisture transferring equation and thermal diffusion equation, respectively. In Pleim-Xiu
and SSiB, the force-restore method is used to solve the soil temperature and soil moisture. The boundary
conditions are also different. For instance, CLM4 takes the ground water table or gravity discharge as the
lower boundary, whereas the others use a fixed value or zero flux as the lower boundary. (c) Descriptions of
vegetation and snow: Slab does not consider the vegetation effect, whereas the other LSMs calculate the effects
of one or more vegetation types in each calculation cell. The numbers of dominant vegetation types for each
LSM are shown in Table1. Snow cover models are considered in all of the LSMs, but their complexities are
different, where Slab uses the simplest snow cover model with a fixed coverage rate, and the highly complex
Community Land Model (CLM) considers a five-layer snow cover model to accurately calculate surface water
and energy fluxes.
2.4. Bayesian Model Averaging (BMA)
Bayesian Model Averaging (BMA) was originally developed as a statistical post-processing method by combin-
ing multiple statistical models for joint inference and prediction. Raftery etal.(2005) extended BMA to several
dynamic models. Let
=1,
⋯
,

represent the results of K different numerical models, y represents the
Table 1
Main Characteristics of Slab, Noah, Rapid Update Cycle, Noah-MP, Community Land Model Version 4 Scheme, Pleim-
Xiu, and Simplified Simple Biosphere Land Surface Models in the WRF
Scheme Soil level Soil depth (m)
Water/heat transfer
between soil layers Vegetation type in each cell
Canopy
processes
Slab 5 0.31 No Explicit vegetation effect not considered No
Noah 4 2 Ye s One dominant vegetation type Fixed
RUC 9 3 Yes Specified from USGS data Fixed
Noah-MP 4 2 Yes One dominant vegetation type Dynamic
CLM4 10 4.32 Ye s Four dominant vegetation types Dynamic
Pleim-Xiu 2 1 No Specified from USGS data Fixed
SSiB 3 1.5 No One dominant vegetation type Fixed

GeoHealth
DI ETAL.
10.1029/2022GH000757
5 of 17
predicted quantity, and


represents the training data set. The BMA ensemble model can be expressed by the
weighted averages of multi-model probability distributions, as follows:
[
|
(1,⋯,
,
)] =

∑
=1
[
|
(,
)]
,
where

[
|(
,

)
is the conditional probability density function (PDF) associated with the result of single
ensemble member fk and


is the posterior probability when fk is in its optimal estimation.

(=1,
⋯
,)
is
non-negative and constrained by


∑
=1

=1
, which reflects the relative contribution of each model member to the
forecasting skill for the predicted quantity y in the model training stage.
Precipitation is a non-uniform quantity, and thus its PDF is described as follows (Ji etal.,2019):


[

|(
,
)]
=
[
=0
|(
,
)]
[y = 0] + 
[
𝑦0
|(
,
)]

[

|(
,
)]
[y 𝑦0]
,
where

[
=0
|(
,
)]
represents the PDF at zero precipitation, which is parameterized as a logistic regression
model associated with the cube root of


, and
[
⋅
]
is a general indicative function, with a value of 1 when the
condition in [ ] is true; otherwise, the value is 0.
p[
𝑦0
|(
,
)]
is the PDF when precipitation occurs, and


[

|(
,
)]
is a gamma PDF.
3. Results
3.1. Spatial Comparison of Rainfall Simulations
As shown in Figure2, the daily mean precipitation values for the Meiyu period (from June 11 to July 13 in
2020) simulated by the WRF model with different LSMs had similar spatial distributions. Overall, the amounts
of precipitation simulated by all LSM models were higher than the observed values, although the patterns were
similar to the observation. The four LSMs comprising Noah, RUC, CLM4, and Pleim-Xiu predicted a continuous
strong rain band (Figures2d, 2e, 2g, and2h), whereas three other LSMs (Slab, Noah-MP, and SSiB) produced
two stronger centers (Figures2c–2f and2i), with one in Hunan (HN) province and the other in the central and
southern parts of Anhui (AH) province. Clearly, these patterns were more similar to the observations.
The biases of the spatial distributions are shown in Figure3. No significant differences were found among the
LSMs in the low rainfall areas (<8mm/day), but significant differences were detected in the rainstorm areas
(>12mm/day). Positive bias was dominant, with one in HN province and the other mainly in the central part of
AH province and the border of Jiangsu (JS) and Zhejiang (ZJ) provinces. The most significant weak bias occurred
in the WRF simulation with the SSiB LSM, thereby indicating that the SSiB LSM was more advantageous for
WRF simulations of precipitation during the Meiyu season in YHRV compared with the other LSMs.
Figure4 compares the observed and simulated daily mean precipitation values for the seven LSMs in the WRF.
According to the errors, no significant differences were found among Noah, RUC, Noah-MP, and CLM4 because
their root mean square error (RMSE) values were about 6.8mm/day and their correlation coefficients (CCs) were
about 0.67. However, Pleim-Xiu obtained the worse performance (RMSE=7.71mm/day). By contrast, the Slab
and SSiB LSMs performed the best among the seven LSMs, with RMSEs lower than 6mm/day. In particular,
SSiB had the lowest RMSE (5.59mm/day) and highest CC (0.692mm/day). Overall, the correlations among the
seven LSMs were basically consistent and the main differences occurred in the error size. SSiB had the lowest
RMSE and highest CC, and thus it was considered the optimal LSM in the WRF model for simulating the Meiyu
season in YHRV during 2020.
3.2. Temporal Comparison of Rainfall Simulations
Figure5 shows time series of the area-averaged daily precipitation from June 11 to July 13 in 2020 for the differ-
ent LSMs. The variations in the simulated daily precipitation amounts were basically consistent for all seven
LSMs. However, SSiB performed better than the other LSMs on the days with heavy precipitation compared with
the observations (June 11, June 17, June 22, July 5, and July 11). On days with light precipitation, the simulation
results were basically not significantly different. In general, the errors in the daily mean precipitation amounts
simulated by using RUC, Noah-MP, and CLM4 were close to 5.76 mm/day, and that with Noah was slightly

GeoHealth
DI ETAL.
10.1029/2022GH000757
6 of 17
higher at 5.95mm/day. Pleim-Xiu had the largest error (6.10mm/day), whereas Slab and SSiB had relatively
small errors of 5.59mm/day and 5.31mm/day, respectively. Therefore, SSiB performed the best at simulating the
daily precipitation time series.
In addition to the continuous statistical metrics (e.g., RMSE and CC), the threat score (TS) probability statistic
was used to evaluate the performance of the seven LSMs. Table2 shows the TS values for the daily precipita-
tion simulations with each of the LSMs, including light rain (daily rainfall greater than 0.1mm and less than
10mm), moderate rain (daily rainfall greater than 10mm and less than 25 mm), and heavy rain (daily rainfall
greater than 25mm and less than 50mm). It should be noted that the TS values for storm rain (daily rainfall
greater than 50mm) were equal to zero on each day during the Meiyu in 2020, and thus, they are not listed in
Table2. Table2 shows that Slab, RUC, and SSiB obtained higher TS values than the other models for light rain,
and their TS values were basically equal. SSiB obtained significantly higher TS values than the other models for
moderate rain. Thus, SSiB again performed better at precipitation simulation for the Meiyu event during 2020.
It should also be noted that the TS value for SSiB was equal to zero for heavy rain, but not equal to zero for the
other models. However, the TS values were basically close to zero, and only 2–4 of the 33days had heavy rain TS
Figure 2. Spatial distributions of observed and simulated precipitation with different LSMs for the Meiyu period from June 11 to July 13 in 2020: (a) observations, (b)
Slab, (c) Noah, (d) Rapid Update Cycle, (e) Noah-MP, (f) Community Land Model version 4 scheme, (g) Pleim-Xiu, and (h) Simplified Simple Biosphere.

GeoHealth
DI ETAL.
10.1029/2022GH000757
7 of 17
values. Therefore, the comparison of the TS values for heavy rain was not representative. In addition, the spatial
distribution of the precipitation was weak with SSiB, which was closer to the observations.
3.3. Comparison of Optimal Simulation and Ensemble Average Simulation
Ensemble averaging is another method for reducing model uncertainty. The seven LSMs were treated as ensemble
members to conduct comprehensive uncertainty qualification analysis for the simulation results using the BMA method.
Figure 6 compares the precipitation simulations using BMA and the observations. The BMA results were highly
spatially consistent with the observations, but the optimal LSM comprising SSiB had stronger correlations than BMA.
Further comparisons were also conducted to represent the differences in the LSMs more accurately. Figure7 shows
scatter plots of the precipitation simulation with BMA versus the observations and precipitation simulation with
the optimal LSM versus the observations. BMA obtained relatively good performance compared with the obser-
vations, but SSiB performed better than BMA in terms of RMSE and CC. In particular, the CC for precipitation
simulation with SSiB was 0.73, which was higher than that for the BMA simulation (0.69). Similarly, the RMSE
Figure 3. Spatial distributions of observed precipitation and biases of simulated precipitation (simulation–observation) with different LSMs averaged over the Meiyu
period from June 11 to July 13 in 2020: (a) observations, (b) Slab, (c) Noah, (d) Rapid Update Cycle, (e) Noah-MP, (f) Community Land Model version 4 scheme, (g)
Pleim-Xiu, and (h) Simplified Simple Biosphere.

GeoHealth
DI ETAL.
10.1029/2022GH000757
8 of 17
Figure 4. Simulated daily mean precipitation values (y-axis, mm/day) versus observed values (x-axis, mm/day): (a) Slab,
(b) Noah, (c) Rapid Update Cycle, (d) Noah-MP, (e) Community Land Model version 4 scheme, (f) Pleim-Xiu, and (g)
Simplified Simple Biosphere.
Figure 5. Time series of area-averaged daily accumulated precipitation amounts in simulations using seven LSMs and
observations.

GeoHealth
DI ETAL.
10.1029/2022GH000757
9 of 17
for the precipitation simulation with SSiB was 5.59mm/day, which was signif-
icantly lower than that for the BMA simulation (6.0mm/day). In addition, the
improvement in the precipitation simulation results with BMA for each LSM
ranged from 10.3% for RUC to 22.2% for Pleim-Xiu. Clearly, BMA integrated
the advantages of each ensemble member to help obtain greater performance
than each separate LSM to effectively reduce the uncertainties associated with
each model and between the models.
4. Discussion
4.1. Comparisons With Similar Studies
Many studies have evaluated the differences in the performance of LSMs in
terms of precipitation, and they mainly used Noah, Noah-MP, CLM4, and
Pleim-Xiu. For instance, Zhong etal.(2021) found that Noah-MP obtained
the optimal simulation of precipitation compared with CLM4 and Pleim-Xiu
in the Tibetan Plateau region. After analyzing the mechanism responsible
Table 2
Threat Score (TS) Statistics for Daily Precipitation Simulations Using
Different LSMs Over the Meiyu Event in Yangtze–Huai River Valley During
2020
Light rain (mm/
day) [0.1, 10)
Moderate rain(mm/
day) [10, 25)
Heavy rain (mm/
day) [25, 50)
Slab 0.1156 0.0901 0.0101
Noah 0.0966 0.0970 0.0115
RUC 0.1188 0.0883 0.0120
Noah-MP 0.1019 0.0810 0.0208
CLM4 0.1019 0.0941 0.0115
Pleim-Xiu 0.0759 0.0919 0.0208
SSiB 0.1150 0.1015 0
Figure 6. Comparison of precipitation simulations using Bayesian Model Averaging with the seven LSMs and the optimal
LSM comprising Simplified Simple Biosphere.

GeoHealth
DI ETAL.
10.1029/2022GH000757
10 of 17
for the differences, they found that the drier soil moisture conditions and
lower latent heat flux produced by Noah-MP constrained the wet bias in
the precipitation simulation, so the results were closer to the observations
compared with the other two LSMs. Similar conclusions were obtained
based on precipitation simulations over North America by Chen etal.(2014)
and García-García etal. (2022). Zhuo etal.(2019) found that three LSMs
comprising Noah, Noah-MP, and CLM4 basically obtained consistent perfor-
mance in Northern Italy, and we also produced similar results in our study.
In addition, Slab performed better than other LSMs comprising Noah, RUC,
and Pleim-Xiu in simulations of storms in the YHRV region over China
(Wang etal., 2016; Zeng et al., 2012). These differences in performance
may be related to the different surface energy flux distributions produced
by different LSMs. Thus, the decreased latent heat flux with Slab could
have resulted in a low contribution of evapotranspiration (ET) to precipita-
tion, which reduced the occurrence of high precipitation situations in LSM
simulation. We found that Noah-MP and Slab performed well at simulating
precipitation, but SSiB performed the best in the YHRV region of China. In
previous studies, SSiB also performed better in precipitation simulations over
East Asia (Li etal.,2016) and South Asia (Lu & Zuo,2021), although these
studies focused on low resolution climate simulations (∼50km) rather than
Meiyu seasons, and only four LSMs were considered (Noah, CLM, Pleim-
Xiu, and SSiB).
4.2. Surface Water Vapor Flux
Land surface processes affect precipitation via two routes in terms of the
water balance (Zeng etal.,2011), that is, by directly affecting the surface ET
flux to produce precipitation and influencing precipitation by changing the
low-level wind circulation and moisture flux convergence. Table3 summa-
rizes the daily area-averaged amounts of ET and precipitation simulated by the seven LSMs for the YHRV region.
ET accounted for 16%–23% of the rainfall, thereby demonstrating that moisture flux convergence was still the
dominant factor that affected precipitation. ET increased as the rainfall increased. In addition, the proportion of
ET relative to rainfall differed among the LSMs. For instance, the ET was smallest with Slab, while that under
Noah was close to those under Noah-MP and CLM4, and it was largest with Pleim-Xiu. These different propor-
tions demonstrate that precipitation simulations in WRF are sensitive to the choice of LSM.
Table3 shows that Slab obtained the smallest simulated ET and precipitation amounts among the seven LSMs.
Compared with Slab, the increases in ET were more than half of the increases in rainfall with five other models
(Noah, RUC, Noah-MP, CLM4, and Pleim-Xiu). Thus, the differences in rainfall caused by these LSMs were
mainly due to differences in ET. By contrast, the rainfall simulated by SSiB decreased as the ET increased. Thus,
SSiB contributed less to water vapor convergence than Slab. Table3 shows that the percentage of ET relative to
rainfall was higher under SSiB than Slab, and thus, less moisture flux convergence occurred with SSiB accord-
ing to the water balance principle, and the SSiB results were closer to the observations. Clearly, the ET/rainfall
of no more than 23% demonstrated that moisture flux convergence was still the dominant factor that affected
precipitation.
Figure 7. Scatter plots for precipitation simulation using Bayesian Model
Averaging versus observations (top) and precipitation simulation with the
optimal scheme versus observations (bottom).
Table 3
Daily Area-Averaged Evapotranspiration (ET) and Precipitation Amounts Simulated by the Seven LSMs for the Yangtze–
Huai River Valley Region During the Meiyu Period
Slab Noah RUC Noah-MP CLM4 Pleim-Xiu SSiB
ET (mm/day) 2.44 2.89 3.26 2.98 2.98 3.66 2.83
Rainfall (mm/day) 14.93 15.81 15.31 15.78 15.75 16.10 14.10
ET/Rainfall 16.31% 18.27% 21.31% 18.90% 18.95% 22.74% 20.09%

GeoHealth
DI ETAL.
10.1029/2022GH000757
11 of 17
4.3. Surface Energy Flux
The choice of LSM directly affects the simulated surface energy flux. Figure 8 shows the simulated
area-averaged daily heat fluxes comprising the sensible heat and latent heat fluxes for the WRF model
with different LSMs over the Meiyu period from June 11 to July 13 in 2020. Both the sensible and latent
heat fluxes varied under the different LSMs. In general, the sensible heat flux decreased in the order of:
Slab>CLM4>SSiB>Noah-MP>Noah>RUC>Pleim-Xiu, whereas the latent heat flux increased accord-
ing to the same sequence. Among the LSMs, the heat fluxes for Slab, Pleim-Xiu, and RUC were significantly
different from those with the other LSMs. For example, the sensible heat flux was the smallest with Pleim-Xiu
and the latent heat flux was the largest, with a difference of up to 92W/m
2, and the difference with RUC was
up to 70W/m
2. Slab obtained the lowest difference of 40W/m
2. The sensible heat fluxes with Noah, Noah-MP,
CLM4, and SSiB were 23–25W/m
2, and latent sensible heat fluxes were 80–84W/m
2, and thus the difference
in the two fluxes were about 60W/m
2. The approximate flux simulation results may have been responsible for
the consistent differences in the precipitation simulation results (Figure4). In addition, with all of the LSMs,
the simulated sensible heat fluxes ranged from 11.37W/m
2 to 30.74W/m
2 and the simulated latent heat fluxes
ranged from 60.09W/m
2 to 103.86W/m
2. These findings also demonstrate that selecting different LSMs will
lead to large differences in the simulated surface energy flux allocations.
The latent heat flux is significantly affected by ET and soil moisture. Zhong etal.(2021) found that higher soil
moisture increased the latent heat flux, and thus the atmospheric water content and moist static energy were
greater in the planetary boundary layer to ultimately increase the frequency and intensity of precipitation. We also
found that the latent heat flux was basically consistent with the average simulated rainfall. For instance, the latent
heat fluxes were relatively small with Slab and SSiB (Figure8b), thereby resulting in relatively small simulated
precipitation and ET amounts with all seven LSMs (Figure2 and Table3), and thus, the simulated results were
closer to the observations.
Figure9 shows the spatial distributions of the simulated latent heat flux averaged over the Meiyu period with
the different LSMs. Except for Slab, the latent heat fluxes simulated by the other LSMs tended to increase from
north to south, which was consistent with the observed precipitation pattern. Overall, Slab obtained the smallest
heat flux. Noah, Noah-MP, and CLM4 produced similar latent heat flux distributions, and their precipitation
errors were also similar (Figure4), thereby indicating that Noah, Noah-MP, and CLM4 simulate precipitation
Figure 8. Time series of area-averaged daily heat fluxes obtained from simulations using the seven LSMs: (a) sensible heat
flux and (b) latent heat flux.

GeoHealth
DI ETAL.
10.1029/2022GH000757
12 of 17
with a similar mechanism. In addition, the simulated latent heat flux distributions were stronger with RUC and
Pleim-Xiu than SSiB, which corresponded to their different precipitation distributions. Therefore, the difference
in the latent heat flux (or ET) was one of the major causes of the different precipitation spatial distributions in
the Meiyu event.
4.4. 850hPa Geopotential Height Fields, Wind, and Moisture Flux Convergence
The 850hPa geopotential height field can reflect the water vapor convergence and divergence, and thus, it was
analyzed as an indicator of precipitation forecasts (Shirvani etal.,2019). Figure10 shows the simulations of the
850hPa geopotential heights for the seven LSMs. The simulated low-pressure center in the west of the 850hPa
geopotential height field differed significantly among the LSMs, which corresponded well with their simu-
lated heavy precipitation centers (Figure2). For instance, the low-pressure centers with Noah, RUC, Noah-MP,
and Pleim-Xiu stretched from the southwest to northeast and formed a continuous rain band. However, the
low-pressure (1,440gpm) center zones with Slab and SSiB were in the southwest, forming a southwest precipi-
tation center, which corresponded well with the observed precipitation center.
Figure 9. Spatial distributions of the simulated latent heat fluxes averaged over the Meiyu period from June 11 to July 13 in 2020: (a) Slab, (b) Noah, (c) Rapid Update
Cycle, (d) Noah-MP, (e) Community Land Model version 4 scheme, (f) Pleim-Xiu, and (g) Simplified Simple Biosphere.

GeoHealth
DI ETAL.
10.1029/2022GH000757
13 of 17
Under the influence of the summer monsoon, the prevailing southwest wind brings abundant warm and moist
water vapor to the area simulated in this study. The water vapor convergence area mainly formed in the south-
west corner of the simulated area and the central and southern part of Anhui province. Figure10 illustrates the
integrated moisture flux convergence (Tsuji etal.,2021). Compared with the spatial rainfall diagram (Figure2),
the area with stronger moisture flux convergence (>2mm/day) corresponded to the stronger rainfall area, thereby
indicating that the precipitation in this area was contributed mainly by moisture flux convergence. In addition,
the moisture flux convergence was smallest with SSiB, which corresponded to the low rainfall distribution
(Figure2h), and Table3 also shows that the increased ET with SSiB led to decreased precipitation.
4.5. Critical LSMs Parameters That Affected Precipitation Simulations
The parameter settings are the main factors that affect the simulation results obtained by different LSMs. In
order to ensure objectivity and fairness, the default parameters were used for each of the LSMs evaluated in this
study without artificial parameter adjustment. Similarly, for the Noah-MP model with alternative multi-schemes,
the default parameterization option was adopted for each physics. The critical LSM parameters that affected
Figure 10. Fields of daily mean geopotential height (unit: gpm, blue contour lines), wind (unit: m/s, gray vectors), and temperature (unit: °C, pink contour lines) at
850hPa, and integrated moisture flux convergence (unit: mm/day, contour shade) over the Meiyu period from June 11 to July 13 in 2020: (a) Slab, (b) Noah, (c) Rapid
Update Cycle, (d) Noah-MP, (e) Community Land Model version 4 scheme, (f) Pleim-Xiu, and (g) Simplified Simple Biosphere.

GeoHealth
DI ETAL.
10.1029/2022GH000757
14 of 17
the precipitation simulation performance are summarized as follows. (a) The soil hydraulic conductivity
affects the conduction velocity of soil water between the soil upper and lower layers. When the soil hydraulic
conductivity increases, soil water will be transferred faster in the soil layer. The rainfall amount is affected when
more soil water is transferred to the ground surface to participate in surface evapotranspiration. (b) The soil
thermal conductivity determines the soil heat conduction velocity between the upper and lower layers in a simi-
lar manner to the soil hydraulic conductivity. Finally, the surface soil temperature is affected by the soil thermal
conductivity to influence the surface evapotranspiration and further affect the rainfall amount. (c) The scale factor
between actual and potential ET determines how well the simulated ET agrees with the observed value. The scale
factor is a power function for the soil moisture in CLM4 and SSiB and a linear function for the soil moisture in
Noah and Noah-MP. Xiao etal.(2017) found that the power parameterization in CLM and SSiB performed better
than the linear parameterization in Noah. (d) Other parameters. For example, soil parameters such as the number
of soil layers and soil depth, snow cover parameters such as the number of snow layers, snow albedo, and snow
duration, and vegetation parameters such as the vegetation fraction, minimum stomatal resistance, and roughness
length can also affect the performance of different LSMs to some extent.
4.6. Effects of Different LSM Rainfall Accuracies on Public Health Evaluations
The relationships between the improved LSM configuration and public health evaluations were analyzed in addi-
tion to quantifying the optimal LSM and explaining the possible associated mechanisms. Using the China Water
Statistical Yearbook for 2008−2017, data sets for the annual rainfall amounts (unit: mm), direct gross domestic
product (GDP) loss caused by flood disasters, including floods (Tarabochia-Gast etal.,2022), landslides (Pollock
& Wartman,2020), and debris flows (unit: billion yuan), and flood-affected areas caused by flood disasters,
including floods, landslides, and debris flows (unit: 10
3ha) were collected for the YHRV region. These data sets
were then used to determine the relationship between the annual rainfall and flood-affected GDP loss, and the
relationship between the annual rainfall and flood-affected area. Finally, the effects of the precipitation simula-
tion accuracies of different LSMs on the estimated GDP losses and flood-affected areas in the YHRV during the
Meiyu in 2020 were evaluated by establishing appropriate relationships.
Figure11a shows a scatter plot of the annual rainfall versus the GDP loss for the YHRV region from 2008 to
2017 as well as a fitted straight line. The fitted equation indicates that the GDP loss caused by flood disasters
increased by about 51.33 billion yuan when the annual rainfall increased by 100mm. Similarly, Figure11b shows
a scatter plot of the annual rainfall versus the flood-affected area for the YHRV region from 2008 to 2017 as well
as a fitted straight line. When the annual rainfall increased by 100mm, the flood-affected area increased by about
107.47 thousand hectares. Based on these proportional relationships, the errors in the GDP loss (f lood-affected
area) estimates among different LSMs for the Meiyu in YHRV during 2020 were solved and the results are shown
in Figures11c and11d. The GDP loss overestimation ranged from 19.1 billion yuan with SSiB to 48.38 billion
yuan with Pleim-Xiu, and the flood-affected area overestimation ranged from 19.10 thousand hectares with SSiB
to 101.38 thousand hectares with Pleim-Xiu. Clearly, the flood disaster loss estimations differed greatly when
different LSMs were used. Therefore, employing an accurate LSM can obtain more accurate precipitation fore-
casts and also reduce the uncertainties in flood disaster loss estimations.
5. Conclusions and Prospective
This study assessed the effectiveness of seven LSMs in WRFv4.3 precipitation simulations of the record-breaking
Meiyu event in the YHRV region during 2020. The simulated period lasted 33days from June 11 to July 13. A
two-grid horizontally nested domain was used to simulate the Meiyu event in the YHRV region. The outer layer
was the East China area with a spatial resolution of 27km, and the inner layer was the YHRV region with a spatial
resolution of 9km. The uniform time step was 60s.
Compared with the observations, the SSiB model obtained the best simulated daily precipitation results among
the seven LSMs, followed by the Slab model and Pleim-Xiu. Evenly, it outperformed BMA result with seven
LSMs.The proportion of ET relative to rainfall, surface heat flux, 850hPa fields for the geopotential height and
wind, and the moisture flux convergence were possible causes of the differences in the results simulated with the
different LSMs. In addition, BMA was compared with the optimal scheme to demonstrate the superiority of the
optimal LSM selected. Finally, the effects of the precipitation simulation accuracies of different LSMs on flood
disaster loss estimation were also discussed to further demonstrate the superiority of the optimal LSM selected.

GeoHealth
DI ETAL.
10.1029/2022GH000757
15 of 17
Many studies have evaluated the performance of different LSMs in WRF weather or climate simulations, but
we considered some new LSMs and weather or climate events in a more comprehensive evaluation of the
performance of LSMs in the WRF model. (a) We evaluated more LSMs by running the latest version of the
WRF model. For example, all seven LSMs in WRF v4.3 were evaluated in this study and the results showed
that the SSiB model performed best at precipitation simulation over the YHRV region in China. Most previous
studies focused on evaluating the three LSMs comprising Noah, Noah-MP, and CLM4 in different regions
throughout the world, but the SSiB model has not been considered widely despite its excellent performance. In
addition, Pleim-Xiu and RUC have rarely been evaluated, and thus our findings may promote the use of these
LSMs in future studies of land–atmosphere interactions. (b) We considered a representative precipitation event
comprising the Meiyu season, which is a unique weather event in the East Asian monsoon region characterized
by continuous precipitation. Many studies have investigated the formation and development of Meiyu events,
and its relationship with tropical ocean signal, but few have assessed the effectiveness of different LSMs at
simulating Meiyu events, especially the record-breaking Meiyu that occurred in June and July during 2020 in
China.
However, it should be noted that selecting an optimal LSM can significantly reduce the simulation bias but model
uncertainty still exists, and thus the constructed model may be defective and the parameters could be inaccu-
rate in the current state. Therefore, future studies should focus on model development by refining the physical
process descriptions (Zhou etal.,2018) as well as selecting and optimizing the critical physical parameters (Di
etal.,2018,2019; Duan etal.,2017) for the optimal LSM.
Conflict of Interest
The authors declare no conflicts of interest relevant to this study.
Figure 11. Relationships between rainfall amounts and flood disaster losses in the Yangtze–Huai river valley region. (a) Scatter plot of annual rainfall versus the GDP
loss from 2008 to 2017 as well as a fitted straight line. (b) Scatter plot of annual rainfall versus the flood-affected area from 2008 to 2017 as well as a fitted straight line.
(c) Overestimated GDP losses corresponding to overestimated rainfall amounts among different LSMs for the Meiyu in 2020. (d) Overestimated flood-affected areas
corresponding to overestimated rainfall amounts among different LSMs for the Meiyu in 2020.

GeoHealth
DI ETAL.
10.1029/2022GH000757
16 of 17
Data Availability Statement
The precipitation data sets used in our study can be freely accessed at https://www.ckcest.cn/default/es3/
detail/4004/dw_dataset/C92AF495DE300001E327C1BD56401982. The data sets for the direct gross domes-
tic product (GDP) losses and flood-affected areas can be obtained from the China Water Statistical Yearbook
(https://sltjnj.digiwater.cn/index.htm?locale=zh_CN).
References
Attada, R., Kumar, P., & Dasari, H. P. (2018). Assessment of land surface models in a high-resolution atmospheric model during Indian summer
monsoon. Pure and Applied Geophysics, 175(10), 3671–3696. https://doi.org/10.1007/s00024-018-1868-z
Chen, F., & Dudhia, J. (2001). Coupling an advanced land surface-hydrology model with the Penn State-NCAR MM5 mode-
ling system. Part I. Model implementation and sensitivity. Monthly Weather Review, 129(4), 569–585. https://doi.
org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2
Chen, F., Liu, C., Dudhia, J., & Chen, M. (2014). A sensitivity study of high-resolution regional climate simulations to three land surface models
over the western United States. Journal of Geophysical Research: Atmospheres, 119(12), 7271–7291. https://doi.org/10.1002/2014JD021827
Chen, X., Wen, Z., Song, Y., & Guo, Y. (2022). Causes of extreme 2020 Meiyu-Baiu rainfall: A study of combined effect of Indian Ocean and
Arctic. Climate Dynamics, 59(11–12), 3485–3501. https://doi.org/10.1007/s00382-022-06279-0
Constantinidou, K., Hadjinicolaou, P., Zittis, G., & Lelieveld, J. (2020). Performance of land surface schemes in the WRF model for climate
simulations over the MENA-CORDEX domain. Earth Systems and Environment, 4, 647–665. https://doi.org/10.1007/s41748-020-00187-1
Di, Z., Ao, J., Duan, Q., Wang, J., Gong, W., Shen, C., et al. (2019). Improving WRF model turbine-height wind-speed forecasting using
a surrogate-based automatic optimization method. Atmospheric Research, 226, 1–16. https://doi.org/10.1016/j.atmosres.2019.0
4.011
Di, Z., Duan, Q., Wang, C., Ye, A., Miao, C., & Gong, W. (2018). Assessing the applicability of WRF optimal parameters under the different
precipitation simulations in the Greater Beijing Area. Climate Dynamics, 50(5–6), 1927–1948. https://doi.org/10.1007/s00382-017-3729-3
Ding, Y., Liu, Y., & Hu, Z. (2021). The record-breaking Meiyu in 2020 and associated atmospheric circulation and tropical SST anomalies.
Advances in Atmospheric Sciences, 38(12), 1980–1993. https://doi.org/10.1007/s00376-021-0361-2
Duan, Q., Di, Z., Quan, J., Wang, C., Gong, W., Gan, Y., etal. (2017). Automatic model calibration-a new way to improve numerical weather
forecasting. Bulletin of the American Meteorological Society, 98(5), 959–970. https://doi.org/10.1175/BAMS-D-15-00104.1
Dudhia, J. (1996). A multi-layer soil temperature model for MM5. In The sixth annual PSU/NCAR mesoscale model users' workshop. PSU/
NCAR.
Dudhia, J., Gill, D., Manning, K., Wang, W., & Br uyere, C. (1999). PSU/NCAR mesoscale modeling system tutorial class notes and user's guide:
MM5 modeling system version 3. NCAR. https://doi.org/10.1016/S0924-0136(97)00397-X
Friend, A. D., & Kiang, N. Y. (2005). Land surface model development for the GISS GCM: Effects of improved canopy physiology on simulated
climate. Journal of Climate, 18(15), 2883–2902. https://doi.org/10.1175/JCLI3425.1
García-García, A., Cuesta-Valero, F. J., Beltrami, H., González-Rouco, J. F., & García-Bustamante, F. (2022). WRF v.3.9 sensitivity to land
surface model and horizontal resolution changes over North America. Geoscientific Model Development, 15(2), 413–428. https://doi.org/
10.5194/gmd-15-413-2022
Hong, S. Y. (2010). A new stable boundary layer mixing scheme and its impact on the simulated East Asian summer monsoon. Quarterly Journal
of the Royal Meteorological Society, 136(651), 1481–1496. https://doi.org/10.1002/qj.665
Hong, S. Y., & Lim, J. O. J. (2006). The WRF single-moment 6-class microphysics scheme (WSM6). Asia-Pacific Journal of Atmospheric
Sciences, 42(2), 129–151.
Huang, X., Zhou, H., Yang, X., Zhou, W., Huang, J., & Yuan, Y. (2021). Spatial characteristics of coronavirus disease 2019 and their possi-
ble relationship with environmental and meteorological factors in Hubei province, China. GeoHealth, 5(6), e2020GH000358. https://doi.
org/10.1029/2020GH000358
Ji, L., Zhi, X., Zhu, S., & Fraedrich, K. (2019). Probabilistic precipitation forecasting over East Asia using Bayesian model averaging. Monthly
and Forecasting, 34(2), 377–392. https://doi.org/10.1175/WAF-D-18-0093.1
Kain, J. S. (2004). The Kain-Fritsch convective parametrization: An update. Journal of Applied Meteorology and Climatology, 43(1), 170–181.
https://doi.org/10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2
Li, M., Di, Z., & Duan, Q. (2021). Effect of sensitivity analysis on parameter optimization: Case study based on streamflow simulations using the
SWAT model in China. Journal of Hydrology, 603, 126896. https://doi.org/10.1016/j.jhydrol.2021.126896
Li, M., Ma, Z., Gu, H., Yang, Q., & Zheng, Z. (2017). Production of a combined land surface data set and its use to assess land-atmosphere
coupling in China. Journal of Geophysical Research: Atmospheres, 122(2), 948–965. https://doi.org/10.1002/2016JD025511
Li, W., Guo, W., Xue, Y., Fu, C., & Qiu, B. (2016). Sensitivity of a regional climate model to land surface parameterization schemes for East Asian
summer monsoon simulation. Climate Dynamics, 47(7–8), 2293–2308. https://doi.org/10.1007/s00382-015-2964-8
Liu, J., Jia, B., Xie, Z., & Shi, C. (2016). Ensemble simulation of land evapotranspiration in China based on a multi-forcing and multi-model
approach. Advances in Atmospheric Sciences, 33(6), 673–684. https://doi.org/10.1007/s00376-016-5213-0
Liu, Y., & Ding, Y. (2020). Characteristics and possible causes for the extreme Meiyu in 2020. Meteorological Monthly, 46(11), 1393–1404.
https://doi.org/10.7519/j.issn.1000-0526.2020.11.001
Lu, S., & Zuo, H. (2021). Sensitivity of South Asian summer monsoon simulation to land surface schemes in weather research and forecasting
model. International Journal of Climatology, 41(15), 6805–6824. https://doi.org/10.1002/joc.7278
Manabe, S. (1969). Climate and the ocean circulation: I the atmospheric circulation and the hydrology of the Earth's surface. Monthly Weather
Review, 97(11), 739–774. https://doi.org/10.1175/1520-0493(1969)097<0739:CATOC>2.3.CO;2
Mlawer, E., Taubman, S. J., Brown, P.D., Iacono, M. J., & Clough, S. A. (1997). Radiative transfer for inhomogeneous atmospheres: RRTM, a
validated correlated-k model for the longwave. Journal of Geophysical Research, 102(D14), 16663–16682. https://doi.org/10.1029/97JD00237
Niu, G. Y., Yang, Z., Mitchell, K. E., Chen, F., Ek, M. B., Barlage, M., etal. (2011). The community Noah land surface model with multipa-
rameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. Journal of Geophysical Research,
116(D12), D12109. https://doi.org/10.1029/2010JD015139
Oleson, K. W., Lawrence, D. M., Bonan, G. B., Flanner, M. G., Kluzek, E., Lawrence, P., etal. (2010). Technical description of version 4 of the
community land model (CLM). NCAR tech. Note NCAR/TN-478+STR.
Acknowledgments
This research was funded by the National
Natural Science Foundation of China
(Grants 41930970, 41571368 and
42275021), the State Key Laboratory of
Earth Surface Processes and Resource
Ecology (2022-TS-01), and the National
Key Scientific and Technological
Infrastructure project “Earth System
Science Numerical Simulator Facility”
(EarthLab).

GeoHealth
DI ETAL.
10.1029/2022GH000757
17 of 17
Pleim, J. E., & Xiu, A. (1995). Development and testing of a surface flux and planetary boundary layer model for application in mesoscale models.
Journal of Applied Meteorology and Climatology, 34(1), 16–32. https://doi.org/10.1175/1520-0450-34.1.16
Pollock, W., & Wartman, J. (2020). Human vulnerability to landslides. GeoHealth, 4(10), e2020GH000287. https://doi.org/10.1029/2020GH000287
Pu, Z., & Kalnay, E. (2018). Numerical weather prediction basics: Models, numerical methods, and data assimilation. In Q. Duan, F. Pappen-
berger, J. Thielen, A. Wood, H. Cloke, & J. Schaake (Eds.), Handbook of hydrometeorological ensemble forecasting. Springer.
Qiao, S., Chen, D., Wang, B., Cheung, H., Liu, F., Cheng, J., etal. (2021). The longest 2020 Meiyu season over the past 60 years: Subseasonal
perspective and its predictions. Geophysical Research Letters, 48(9), 1–11. https://doi.org/10.1029/2021GL093596
Raftery, A. E., Gneiting, T., Balabdaoui, F., & Polakowski, M. (2005). Using Baysian model averaging to calibrate forecast ensembles. Monthly
Weather Review, 133(5), 1155–1174. https://doi.org/10.1175/MWR2906.1
Reddy, B. R., Srinivas, C. V., Shekhar, S. S. R., & Venkatraman, B. (2020). Impact of land surface physics in WRF on the simulation of sea breeze
circulation over southeast coast of India. Meteorology and Atmospheric Physics, 132(6), 925–943. https://doi.org/10.1007/s00703-020-00726-5
Shen, Y., Zhao, P., Pan, Y., & Yu, J. (2014). A high spatiotemporal gauge-satellite merged precipitation analysis over China. Journal of Geophys-
ical Research: Atmosphere, 119(6), 3063–3075. https://doi.org/10.1002/2013JD020686
Shirvani, A., Fadaei, A. S., & Landman, W. A. (2019). The linkage between geopotential height and monthly precipitation in Iran. Theoretical
and Applied Climatology, 136(1–2), 221–236. https://doi.org/10.1007/s00704-018-2479-4
Skamarock, W. C., Klemp, J. B., Dudhia, J., Gill, D. O., Liu, Z., Berner, J., etal. (2021). A description of the advanced research WRF model
version 4.3. No. NCAR/TN-556+STR. https://doi.org/10.5065/1dfh-6p97
Smirnova, T. G., Brown, J. M., Benjamin, S. G., & Kenyon, J. S. (2016). Modifications to the rapid update cycle land surface model (RUC LSM)
available in the weather research and forecasting (WRF) model. Monthly Weather Review, 144(5), 1851–1865. https://doi.org/10.1175/MW
R-D-15-0198.1
Tarabochia-Gast, A. T., Michanowicz, D. R., & Bernstein, A. S. (2022). Flood risk to hospitals on the United States Atlantic and Gulf Coasts from
hurricanes and sea level rise. GeoHealth, 6(10), e2022GH000651. https://doi.org/10.1029/2022GH000651
Tsuji, H., Takayabu, Y. N., Shibuya, R., Kamahori, H., & Yokoyama, C. (2021). The role of free-tropospheric moisture convergence for summer-
time heavy rainfall in Western Japan. Geophysical Research Letters, 48(18), e2021GL095030. https://doi.org/10.1029/2021gl095030
van den Hurk, B., Kim, H., Krinner, G., Seneviratne, S. I., Derksen, C., Oki, T., etal. (2016). LS3MIP (v1.0) contribution to CMIP6: The land
surface, snow and soil moisture model intercomparison project—Aims, setup and expected outcome. Geoscientific Model Development, 9(8),
2809–2832. https://doi.org/10.5194/gmd-9-2809-2016
Wang, M., Zeng, X., Wang, N., Yi, X., Zhou, X., & Wang, B. (2016). Sensitivity of simulated summer temperature and precipitation to land surface
schemes by WRF. Meteorology and Disaster Reduction Research, 39(1), 32–40. (in Chinese). https://doi.org/10.12013/qxyjzyj2016-004
Wu, Y., Miao, C., Sun, Y., AghaKouchak, A., Shen, C., & Fan, X. (2021). Global observations and CMIP6 simulations of compound extremes of
monthly temperature and precipitation. GeoHealth, 5(5), e2021GH000390. https://doi.org/10.1029/2021GH00039
Xiao, Y., Ma, Z., & Li, M. (2017). Evaluation of the parameterizations of soil moisture influence on evapotranspiration in land surface models.
Chinese Journal of Atmospheric Sciences, 41(1), 132–146. (in Chinese). https://doi.org/10.3878/j.issn.1006-9895.1606.1529
Xinhua NEWs. (2020). Over 54.8 MLN people affected by floods in China. Archived from Retrieved from http://www.xinhuanet.com/english/
2020-07/29/c_139247084.htm
Xue, Y., Sellers, P.J., Kinter, J. L., & Shukla, J. (1991). A simplified biosphere model for global climate studies. Journal of Climate, 4(3),
345–364. https://doi.org/10.1175/1520-0442(1991)004<0345:ASBMFG>2.0.CO;2
Yan, Y., Tang, J., Wang, S., Niu, X., & Wang, L. (2021). Uncertainty of land surface model and land use data on WRF model simulations over
China. Climate Dynamics, 57(7–8), 1833–1851. https://doi.org/10.1007/s00382-021-05778-w
Yuan, X., Xie, Z., Zheng, J., Tian, X., & Yang, Z. (2008). Effects of water table dynamics on regional climate: A case study over East Asian
monsoon area. Journal of Geophysical Research, 113(D21), D21112. https://doi.org/10.1029/2008JD010180
Zeng, X., Wu, Z., & Song, S. (2012). Effects of land surface schemes on the simulation of a heavy rainfall event by WRF. Chinese Journal of
Geophysics, 55, 16–28. (in Chinese). https://doi.org/10.6038/j.issn.0001-5733.2012.01.002
Zeng, X., Wu, Z., Xiong, S., Song, S., Zheng, Y., & Liu, H. (2011). Sensitivity of simulated short-range high-temperature weather to land surface
schemes by WRF. Science China Earth Sciences, 54(4), 581–590. https://doi.org/10.1007/s11430-011-4181-6
Zhang, C., Di, Z., Duan, Q., Xie, Z., & Gong, W. (2020). Improved land evapotranspiration simulation of the community land model using a
surrogate-based automatic parameter optimization method. Water, 12(4), 943. https://doi.org/10.3390/w12040943
Zhong, L., Huang, Z., Ma, Y., Fu, Y., Chen, M., Ma, M., & Zheng, J. (2021). Assessments of weather research and forecasting land surface models
in precipitation simulation over the Tibetan Plateau. Earth and Space Science, 8(3), e2020EA00156. https://doi.org/10.1029/2020EA001565
Zhong, S., Yang, S., & Chen, Z. (2019). Evaluation of the parameterization schemes and nudging techniques in GRAPES for warm sector torren-
tial rains using surface observations. Journal of Tropical Meteorology, 25(3), 353–364. https://doi.org/10.16555/j.1006-8775.2019.03.007
Zhou, X., Yang, K., & Wang, Y. (2018). Implementation of a turbulent orographic form drag scheme in WRF and its application to the Tibetan
Plateau. Climate Dynamics, 50(7–8), 2443–2455. https://doi.org/10.1007/s00382-017-3677-y
Zhuo, L., Dai, Q., Han, D., Chen, N., & Zhao, B. (2019). Assessment of simulated soil moisture from WRF Noah, Noah-MP, and CLM land
surface schemes for landslide hazard application. Hydrology and Earth System Sciences, 23(10), 4199–4218. https://doi.org/10.5194/hess-
23-4199-2019