oEPA
United States
Environmental Protection
Agency
tPA/600/R-17/441 |September 2017| www.epa.gov/research
A Survey of Precipitation Data
for Environmental Modeling
-------
1
-------
EPA/600/R-14/152
September 2017
A Survey of Precipitation Data for
Environmental Modeling
Jan Sitterson1, Chris Knightes2, Rajbir Parmar2, Kurt Wolfe2, Brian
Avant3, Amber Ignatius4, Deron Smith1
1Oak Ridge Associated Universities
2U.S. Environmental Protection Agency, Office of Research and Development
3Oak Ridge Institute for Science and Education
4Oak Ridge Institute for Science and Education now at University of North Georgia
Chris Knightes - Project Officer
Office of Research and Development
National Exposure Research Laboratory
Athens, Georgia, 30605
2
-------
Notice
The U.S. Environmental Protection Agency (EPA) through its Office of Research and Development
funded and managed the research described here. The research described herein was conducted at the
Computational Exposure Division of the U.S. Environmental Protection Agency National Exposure
Research Laboratory in Athens, GA. Any mention of trade names, products, or services does not imply
an endorsement by the U.S. Government or the U.S. Environmental Protection Agency. The EPA does
not endorse any commercial products, services, or enterprises. This document has been reviewed by the
U.S. Environmental Protection Agency, Office of Research and Development, and approved for
publication.
3
-------
Abstract
This report explores the types of precipitation data available for environmental modeling.
Precipitation is the main driver in the hydrological cycle and modelers use this information to
understand water quality and water availability. Models use observed precipitation information for
modeling past or current conditions, while simulated data are used to predict future conditions as
well as re-create historic conditions. Rain gauge-, radar-, and satellite-based measurements are
categorized in the observed precipitation dataset. Calculated precipitation data from numerical
weather predictors, stochastic models, and nonparametric models are part of the simulated data
available for modeling. Temporal resolution, data availability, spatial resolution, and method of
measuring precipitation are described for each dataset; global datasets and datasets of the contiguous
United States are explained in this report. Our goal is to inform modelers of the various types,
resolutions, and sources of precipitation data available for environmental modeling. We discuss only
a few frequently cited datasets in detail due to the vast amounts of precipitation data available for
modeling purposes.
4
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the Nation's
land, air, and water resources. Under a mandate of national environmental laws, the Agency strives to
formulate and implement actions leading to a compatible balance between human activities and the
ability of natural systems to support and nurture life. To meet this mandate, EPA's research program is
providing data and technical support for solving environmental problems today and building a science
knowledge base necessary to manage our ecological resources wisely, understand how pollutants affect
our health, and prevent or reduce environmental risks in the future.
The National Exposure Research Laboratory (NERL) Computational Exposure Division (CED)
develops and evaluates data, decision-support tools, and models to be applied to media-specific or
receptor-specific problem areas. CED uses modeling-based approaches to characterize exposures,
evaluate fate and transport, and support environmental diagnostics/forensics with input from multiple
data sources. It also develops media- and receptor-specific models, process models, and decision support
tools for use both within and outside of EPA.
The goal of the Hydrologic Micro Services (HMS) project is to develop a collection of inter-operable
water quantity and quality modeling components. Components can be integrated to rapidly compose
work flows to address water quantity and quality related questions. Each component may have multiple
implementations ranging from macro (coarse) to micro (detailed) levels of modeling the physical
processes. The components leverage existing internet-based data sources and sensors. They can be
integrated into a work flow in two ways: calling a web service or downloading component libraries. It is
generally more efficient to call a web service for less computational intensive components, yet, local
copies of components are needed if the component requires large amounts of input/output data.
Elaine Hubal, Acting Division Director for CED
5
-------
Table of Contents
Notice 3
Abstract 4
Foreword 5
List of Tables 7
List of Figures 7
Acronyms and Abbreviations 8
1. Introduction 9
2. History of Precipitation Data 10
3. Differences in Observed and Simulated Precipitation Data 11
3.1 Spatial Resolution 13
3.2 Precipitation Data Limitations 13
4. Observed Data 14
4.1 Rain Gauge 14
4.1.1 NCDC/NCEI 15
4.1.2 GPCC 16
4.1.3 Daymet 16
4.2 RADAR 16
4.2.1 NEXRAD 17
4.3 Satellite 18
4.3.1 TRYIYI 19
4.3.2 GPM 20
4.3.3 CMORPH 20
4.3.4 PERSIANN 20
4.4 Combining Datasets 20
4.4.1 LDAS 21
4.4.2 PRISM 21
5. Simulated Data 21
5.1 Numerical Weather Prediction (NWP) 22
5.1.1 WRF Model 22
5.1.2 ECHAM 23
5.2 Stochastic 23
5.2.1 Weather Generators (WGEN) 24
6
-------
5.3 Nonparametric Models 24
5.3.1 GCM 24
6. Discussion 25
References 27
List of Tables
Table 1. Description of Precipitation Datasets 12
Table 2. Comparison of the major types of precipitation datasets 11
Table 3. Summary of observed precipitation dataset characteristics 14
Table 4. Description of precipitation datasets that use satellite technology 18
Table 5. Summary of simulated precipitation dataset characteristics 22
List of Figures
Figure 1. A simplified diagram of the water cycle within a watershed. (Brewster, 2017; ESRI, 2015) .... 9
Figure 2. Map of rain gauge stations in the United States showing precipitation detected on April, 12,
2016. Each dot represents one rain gauge at a specific location, colored by the amount of precipitation
measured. Image from NOAA's website 15
Figure 3. A map showing the NEXRAD site locations in the contiguous United States. Image from
(NOAA, 2017) 17
Figure 4. TRMM's orbit path captures the spatial resolution over the tropics on April 12th, 2012. Yellow
areas depict probable rainfall. Image from NASA 19
7
-------
Acronyms and Abbreviations
CAM Community Atmosphere Model
CESM Community Earth Systems Model
CHIRPS Climate Hazards Group Infrared Precipitation with Station
CMAP Climate Prediction Center Merged Analysis of Precipitation
CMIP Coupled Model Intercomparison Project
CMORPH Climate Prediction Center Morphing
ECHAM European Centre Hamburg Model
GCM Global Circulation Model
GEOS-16 Geostationary Operational Environmental Satellite System
GLDAS Global Land Data Assimilation System
GPCC Global Precipitation Climatology Centre
GPM Global Precipitation Mission
NASA National Aeronautics and Space Administration
NCDC/NCEI National Climatic Data Center / National Center for Environmental Information
NEXRAD Next Generation Weather Radar
NLDAS North American Land Data Assimilation System
NWP Numerical Weather Prediction
ORD Office of Research and Development
PERSIANN Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks
PRISM Parameter-elevation Relationship on Independent Slopes Model
RADAR Radio Detection and Ranging
RCP Representative Concentration Pathways
SWAT Soil and Water Assessment Tool
TDWR Terminal Doppler Weather Radar
TRMM Tropical Rainfall Measuring Mission
US EPA United States Environmental Protection Agency
WGEN Weather Generator
WRF Weather Research and Forecasting
8
-------
1. Introduction
This report provides a survey of precipitation data sources and generation methods for
environmental modeling. As a main component of the hydrological cycle and contaminant fate and
transport, precipitation data are needed for hydrological modeling, erosion modeling, and water
quality research. We compile descriptions of several precipitation datasets and data generation
methods available globally and for the contiguous United States, specifically. The datasets and data
generation methods included are publicly available online and often cited by the modeling
community. Some of the datasets are purely observed and derived from rain gauges, radar, and
satellites, while others are simulated datasets generated by mathematical equations to predict future
weather conditions and recreate past events. Additionally, some precipitation data products are
derived from a combination of observed data and model equations to generate weather estimates.
Precipitation has great importance because it influences drinking water availability, supports
agriculture, and maintains freshwater resources. It is a vital component in the global hydrological
cycle due to its direct effect on the circulation of Earth's latent heat (Ebert, 2007). Chahine (1992)
states that "the hydrological cycle is the largest movement of any substance on Earth's surface".
Most water movement occurs through precipitation and evaporation. Controlled by the sun's
radiation, water evaporates from the ocean and the land's surface where it moves with winds in the
atmosphere. It then condenses into clouds to fall back to Earth's surface as precipitation flowing
toward the oceans to complete the global hydrological cycle (Chahine, 1992). With the exception of
arid climates, precipitation often exceeds evaporation over land and the excess drains to a reservoir
or recharges groundwater (Fig. 1). Precipitation is highly variable and influences vegetation,
droughts, floods, and the movement of minerals and chemicals. In agriculture and urban areas,
precipitation drives contaminant and nutrient transport in water systems through runoff.
The Water Cycle
Figure 1. A simplified diagram of the water cycle within a watershed. (Brewster, 2017; ESR1, 2015)
9
-------
Precipitation data are integral inputs for many watershed, air, erosion and agricultural models
as well as climate-predicting projects. It determines flood/drought conditions, hydrologic
transportation of contaminants, best management practices, and regulations. Precipitation data are
generated through direct observation and model simulation. Observed data are captured directly
from rain gauge stations, or technologically observed from radar and satellites. Simulated
precipitation data are mathematically generated through parameterizations, statistical probability, or
historical trends. There are many types of precipitation datasets with different spatial and temporal
resolutions available for project needs (Table 1). Each has strengths and weaknesses depending on
its intended use. This report presents different types of available precipitation datasets with details
including temporal and spatial resolution, potential errors in the dataset, and optimal performance
scenarios. It also discusses the benefits and deficiencies of certain precipitation datasets. Focusing on
a project's purpose and understanding its questions, goals, and needs are vital for selecting input data
since exploratory, planning, and regulatory purposes have different input criteria and uncertainty
thresholds (Harmel et al., 2014).
2. History of Precipitation Data
Rain gauge records have been available for hundreds of years. The first scientific report on
differences in measured precipitation using height from rain gauges was by William Heberden in
1770 (Tapiador et al., 2012). As technology has advanced, rain gauge data has become more
accurate in determining the amount of rainfall at a particular location. During World War II, radar
operators searching for enemy ships and aircraft found that precipitation caused 'false' echoes on
their screens (NOAA, 2017); thus began the development of precipitation detection radar. Radar-
based precipitation information overcame the lack of spatial resolution in rain gauge data (Hu et al.,
2014). As research in meteorology continued, there was a need to study macro-scale rainfall which
then led to the use of satellites to monitor cloud cover and precipitation events around the world.
Scientific advancements in spacecraft satellites and high-resolution sensors provide information to
calculate precipitation amounts. Observed data gives a realistic view of precipitation in the past or in
real time, but cannot predict future conditions. Using mathematical equations, simulated
precipitation was therefore created to fill data gaps and predict future scenarios.
With so many precipitation datasets being available, rain gauge data are universally
considered the best source of reference data for precipitation observations (Tapiador et al., 2012).
Some weaknesses of rain gauges include being able to observe precipitation at only one site in space
and often underestimating rainfall amounts. Despite these limitations, many researchers use rain
gauge data because it is assumed to represent the most accurate source of information at the exact
location, and installing a rain gauge is easy and fairly inexpensive (Kim, 2014; Price et al., 2014).
Rain gauge networks have been around for at least a century and, therefore, provide the longest
precipitation record. Because rain gauge data are widely used and free of assumptions, research
methodology involving rain gauge networks is well accepted. Research on climatology requires
long-term datasets of precipitation, and only rain gauge data have enough historical information for
this type of research. Other methods of observing precipitation do not have the necessary decadal
time series although they do have better spatial resolution.
10
-------
3. Differences in Observed and Simulated Precipitation Data
Both observational and simulated precipitation datasets have strengths and weaknesses.
Observed datasets give information about past or current rainfall events, but often have gaps in the
time series due to lack of measurement. Observed data are also more localized spatially due to
providing information at specific sites or over an area. Rain gauge data are observed at a single
location in space and interpolation methods must be used to estimate precipitation across a broader
spatial extent, or assumed to represent constant precipitation over a region. Precipitation simulations
can provide past or future precipitation quantities in a seamless time series over a global extent, at
different spatial discretizations. Generally, simulated data are best used for non-extreme weather
patterns, mountainous regions, and colder weather; observed data performs best in warm weather
and documents extreme events very well (Table 2) (Harmel et al., 2002). Satellite-derived data
perform better than numerical models in warm seasons and over the tropics (Ebert, 2007; Hu et al.,
2014). Studies have shown that input precipitation data from observed and simulated datasets impact
watershed model outputs (Golden et al., 2010; Tuo et al., 2016). Modeling projects that use more
than one type of dataset may be more accurate in reproducing precipitation patterns than a single
dataset, but the spatial and temporal resolutions of different datasets must be considered in
calibration (Tuo et al., 2016). Observed and simulated data can generate outputs in different spatial
scales and can be provided as accumulated precipitation per hour, day, or month. Simple algorithms
fix time or measurement differences and find a common spatial resolution.
Table 1. Comparison of the major types ofprecipitation datasets.
Type
Method
Spatial
Extent
Time Series
Time Step
Best
Performance
Observed
Direct
Specific or
Includes gaps
Half-
Warm
measurement or
range of
of past or
hourly,
weather,
technologically
area
current data
hourly,
extreme
observed
daily,
monthly,
yearly
events
Simulated
Numerical
Global,
Seamless,
Daily,
Cold weather,
calculations
downscaled
future
monthly,
mountainous
based on
to regional
predictions,
yearly
regions, non-
historical
past data
extreme
events
events
11
-------
Table 2. Description of Precipitation Datasets
Type/Name
Precip
Output
Temporal
Resolution
Resolution
(Degree Grid)
Time
Period
Coverage
Time lag
Method
Source
TRMM
mm/hr
3 hourly
0.25x0.25
1998-2015
35N 35S to
50NS
n/a
Microwave, Infrared
TRMM
GPM
mm/hr
0.5 hourly
0.1x0.1
2014-
60N 60S
4-6 hours
Microwave, Infrared, Satellite Precip
Radar
GPM
"3
"ea
CMORPH
mm/hr
0.5, 3 hourly
0.07277x0.07277,
0.25x0.25
2002-
60N 60S
18 hours
Morphing of Microwave and Infrared
CMORPH
cn
PERSIANN
CCS
mm/hr
Hourly
0.04x0.04
2003-
60N 60S
1-2 days
Infrared, Cloud segmentation algorithm
PERSIAN
N-CCS
PERSIANN
CDR
mm/day
Daily
0.25x0.25
1983-2015
60N 60S
n/a
Infrared, Artificial Neural Network
PERSIAN
\i !>s
¦-
CS
¦a
NEXRAD
mm/hr
1, 3 hour
lxl N. America
1994-
160 sites in the
US
2-4 days
Radar, Precipitation Processing System
M \V\x!>
CS
C£
TDWR
mm/hr
Hourly
lxl N. America
2001-
45 sites in US
4days
Radar, Precipitation Processing System
RADAR
a>
ex
s
GPCC Full
Data
mm/mo
Monthly
0.5x0.5
1901-2013
7000 US, 65000
Worldwide
n/a
Weighted Method for grid
GPCC
a
O
a
NCDC
inch
Hourly
By Station
1951-
72N -15S, -60E
130W
6 months
Gathering of multiple stations GHCN,
COOP, QCLCD
NCDC
cS
C£
Daymet
mm/day
Daily
0.0089x0.0089
1980-2015
N. America
1 year
Spatial truncation of Gaussian weighting
filters of ground station locations
Davmet
NLDAS
kg/m2/hr
hourly
0.125x0.125
1979-
N. America
4 days
Integration of CMORPH and RADAR
LDAS
¦a
a>
s
GLDAS
kg/m2/hr
3 hourly
0.25x0.25
2000-Dec
2016
90N 60S
2 months
Incorporation of satellites and ground-
based observations
1 KO-*
a
S
o
U
PRISM
mm/mo
Monthly,
Yearly
0.04x0.04
1981-
CONUS
1 month
Climatologically Aided Interpolation
(CAI) of gauge stations with RADAR
PRISM
CMAP
Pentad RT
mm/day
Daily
2.5x2.5
1979-Dec
2016
88N 88S
1 month
Filling in gaps from gauge data with
satellite (CMORPH)
CMAP
WRF
mm/hr
Daily
0.03x0.03
User
Specified
Global
n/a
NWP Microphysics/Cumulus Schemes
WRF
a>
¦H
ECHAM
mm/day
Daily
0.703125x0.7031
25
User
Specified
Global
n/a
Numeric Weather Prediction and
Parameterization
ECHAM
s
CESM-
CAM
mm/day
Daily
0.35x0.35
User
Specified
Global
n/a
NWP and Non-parametric, CMIP5
CAM
WGEN
mm/day
Daily
HRU
1960-2100
Site Specific
n/a
Stochastic
WGEN
12
-------
3.1 Spatial Resolution
The spatial resolution in purely observed datasets are not uniform due to random
station locations or radar blocking. Gridded precipitation data, i.e., those that provide
precipitation information at each point across an entire domain at a specified grid resolution,
are useful in environmental modeling. To produce a gridded precipitation dataset, values at
locations without observations are manipulated using many methods including, nearest
neighbor, weighted average, geostatistics, mechanistic methods, etc. Such methods assume
that points close to each other are better correlated (Tuo et al., 2016). Since rainfall is not
distributed evenly, rainfall estimates are often misleading; interpolation of observed data is a
significant limitation in accurately modeling responses to rainfall because data cannot be
validated at every position. Globally-simulated datasets must be downscaled to reflect a
study area. This can cause error because global generalizations may not represent local
processes. The statistical relationships between large climatic parameters and local variables
affecting precipitation (e.g., temperature and its effects on evaporation) are needed to
downscale global outputs accurately (Wilks & Wilby, 1999). A major problem in scaling to a
grid for observed and simulated data is that precipitation is not evenly distributed and values
may differ within a grid. With all precipitation datasets, coarser spatial resolutions lead to
more approximations about rainfall distribution, and interpolation introduces known biases to
the results (Tapiador et al., 2012).
3.2 Precipitation Data Limitations
There is no way to determine the exact weather condition at every point in space,
which means that all datasets have limitations. Observational data often have missing values
due to station maintenance or equipment malfunction; error sources can be due to sampling
errors, calibration uncertainty, or random errors. Instrument and calibration uncertainty also
pose potential sources of bias. Due to the inability to accurately measure frozen precipitation
in all observational techniques, observed datasets are most accurate during warm weather
conditions. The length of recorded data also differs between datasets. Radar and satellite data
do not yet have records old enough for climatology research, which requires at least 30 years
of historical data. Simulated outputs from mathematical equations do not depict precipitation
events with as much detail as observed datasets. Correctly simulating patterns, seasonal
variations, and characteristics of precipitation with mathematical models is an area of active
research (Eyring et al., 2016). Harmel et al. (2002) revealed that weather variability,
especially in extreme events, is difficult to predict since the event does not fit common
mathematical distributions. Model drift is a common problem among simulated precipitation
datasets due to the probability distribution being skewed away from observed data; modelers
often can compensate for this using a correction factor after simulation runs. The algorithm
or parameterization scheme selected in a numerical weather prediction model influences
model uncertainty since some schemes work better in certain locations. Combining different
output datasets improves regional and global precipitation data results (Ebert, 2007; Huffman
et al., 1995; NOAA, 2017).
13
-------
4. Observed Data
Observational data provide a historical record of past precipitation events using direct
rainwater catchment in rain gauges or with technical instruments from a distance (e.g.,
sensors). A historical record of precipitation is helpful in looking at changes and trends in
day-to-day climate. Observed precipitation data are specific to the sampling location and
range of detection (Table 3), which often leaves gaps in the spatial and temporal resolution of
the dataset. Observational systems give an accurate depiction of the amount of rain produced
by an extreme event like hurricanes or monsoons. Rain gauge data can provide estimates of
rain accumulation at exact locations; when using rain gauge data, an observed value
represents uniform precipitation in the area around the gauge. If two or more rain gauges are
used for a study area, a method for interpolating data across the region, such as the Thiessen
Polygon Method, can be applied. Radar and satellite data have larger ranges of detection and
can serve as a warning devices for current and near future precipitation events. Many studies
have compared observational datasets and their ability to estimate precipitation amounts and
their effect on model output. Gao et al. (2017) studied the impacts of three different
precipitation sources (rain gauge, radar, and a combined reanalysis dataset) in SWAT
streamflow simulations.
Table 3. Summary of observed precipitation dataset characteristics.
Observed
Method
Spatial
Extent
Spatial
Resolution
Temporal
Resolution
Years
of
data
Precipitation
Output
Error
Rain
Gauge
Physically
collected on
the ground
Specific
locations
Lat.-Lon. of
station,
0.009x0.009
or 0.5x0.5-
degree grid
interpolation
Hourly,
Daily,
Monthly,
Yearly
100
years
Underestimates
heavy rainfall
events
Random
error,
mechanical
issues,
location
Radar
Technological!
y collected on
the ground
Radial
area
around
station
(radius
230km)
lxl degree
grid, lat. and
long of
station
Hourly, 3-
hourly,
30-40
years
Overestimates
heavy rainfall
events,
underestimates
light rainfall
Signal
blockage, hail
misreading
Satellite
Technologicall
y collected
from space
Latitude
range
(60°N,
60°S)
0.04x0.04-
degree grid
0.1x0.1-
degree grid
0.25x0.25-
degree grid
Half-
Hourly,
Hourly,
Daily
20
years
or
less
Underestimates
rain from warm
top clouds
Frozen
precipitation,
multilayer
clouds
4.1 Rain Gauge
Rain gauge precipitation data represent the direct capture and measurement of
rainwater at a specific location. Inexpensive and easy to install rain gauges are found all over
the world. There are many different methods for capturing direct rainfall varying in
14
-------
complexity from measured cylinders to sophisticated weighing gauges. One of the most
common methods for determining the amount of precipitation is the tipping bucket method,
which records the time and a count of the number of times a premeasured bucket tips from
overfilling with rainwater (Tapiador et al,, 2012). The tipping bucket method becomes more
inaccurate with increasing rainfall intensities due to catching and counting errors (Shedekar
et al., 2016). Gauge data are the most accurate representation of precipitation at a precise
location (Kim, 2014; Price et al., 2014) but limitations stem from mechanical issues or
operational errors. Random errors in datasets can occur due to damage to the gauge from
wildlife or humans, and gauge data often underestimates precipitation amount due to wind
effects, frozen precipitation, and rain particles that evaporate before contact with the gauge
(Kidd & Huffman, 2011; Tapiador et al.s 2012). Despite these limitations, many studies use
gauge stations in modeling total maximum daily loads for management purposes. Liu et al.
(2008) used precipitation data from five stations to model nitrogen transportation in three
models (WASP1, EFDC2, and HSPF3). Below is a short description of a few often cited rain
gauge data sources and datasets that primarily use rain gauge data.
Figure 2. Map of rain gauge stations in the United States showing precipitation detected on April, 12, 2016. Each dot represents
one rain gauge at a specific location, colored by the amount ofprecipitation measured. Image from NOAA's website.
4.1.1 NCDC/NCEI
The National Climatic Data Center (NCDC), now named the National Center for
Environmental Information (NCEI), provides precipitation data recorded at rain gauge
1 Water Quality Analysis Simulation Program epawasp.lwool.com
2 Environmental Fluid Dynamics Code https://www.epa.gov/exposure-assessment-models/efdc
3 Hydrological Simulation Program Fortran ittps://www.epa.gov/exposure-assessment-models/hspf
15
-------
stations around the world. NCEI uses a network of volunteer centers to collect daily weather
observations from local rain gauges. NCEI has access to about 53,000 stations worldwide
some with data going as far back as 1901 (NOAA, 2017), these data can be accessed online
and queried by location (Latitude, Longitude) of a gauge station. The spatial coverage of
NCEI gauge stations in the United States is shown in Figure 2; the network of gauge stations
and precipitation data can be accessed at NOAA's website.
4.1.2 GPCC
The Global Precipitation Climatology Centre (GPCC), another source for global
precipitation data from rain gauge stations on the ground, is comprised of about 67,000 rain
gauge stations worldwide. This dataset is available from 1901 to 2013 on a monthly time step
and provides interpolated data on three grid sizes, with the finest resolution on a 0.5-degree
grid (NOAA, 2017). A spherical adaptation of Shepard's empirical weighting scheme is used
to transpose gauge stations to a grid point (Becker, 2013). GPCC data are accessible at
UCAll's website.
4.1.3 Day met
Daymet is a daily dataset of rain gauge data interpolated and extrapolated by the
Daymet algorithm. It uses ground station data from NCEI with its model algorithm to
produce gridded estimates of daily weather parameters. Interpolation for the gridded
resolution uses the spatial convolution of a truncated Gaussian filter from the local station
density (Thornton, 2017). Daily rainfall is rounded to the nearest whole number. The
interpolated spatial resolution is about a 0.009-degree grid, or approximately 1km resolution,
over North America. Data are accessible since 1980, to the latest full year, due to
interpolation at Davmet's website.
4.2 RADAR
Radio Detection and Ranging (RADAR), detects precipitation in the troposphere.
Radar weather systems are mostly found on the ground, with a few on satellites. They send
radio waves into the atmosphere in pulses and radio waves are sent back when the wave
makes contact with a raindrop. The system calculates the distance and direction of the rain
and uses the Doppler Effect to provide precipitation characteristics like reflectivity and
droplet size (NOAA, 2017). A reflectivity-to-rainfall equation - the Precipitation Processing
Subsystem (PPS) — estimates rainfall amounts (Nelson et al., 2010). The relationship
between reflectivity and rainfall amount varies for different forms of precipitation, resulting
in uncertainty. An example of modified return values for different precipitation types is that
interpretation of hail from radar can send a signal that resembles heavy rainfall. Radar data
can make short-term forecasts and show intensity of a storm event as a warning to the public
(NOAA, 2017). Because radar stations are very expensive, some countries cannot afford
radar equipment. With its higher range of detection, radar data covers more area than rain
gauge data. Radar technology can locate precipitation within a range of 230 km from the
16
-------
station and reports data close to real time (NOAA, 2017). Although radar detects rainfall on a
larger scale than rain gauges, ground-based radar cannot reach high altitude clouds and
signals can be blocked by topographic effects. Bias in the radar dataset comes from signal
blockage, bright band contaminations, range dependency, and radar calibration errors.
4.2.1 NEXRAD
Next Generation Weather Radar (NEXRAD) is the largest collection of forecasting
radars accessible for research in North America. It is composed of 160 WSR-88D ground
radars across the United States (see Figure 2). NEXRAD has two levels of output data,
Level-II and Level-Ill. Level-II is raw meteorological data including reflectivity, radial
velocity, and spectrum width. Level-Ill is a set of computer-processed products that include
hourly precipitation and bias-corrected precipitation from rain gauges. NEXRAD
observations are taken every hour and can be gridded on a one-degree grid two to four days
after observations using a spatial weighting scheme. NEXRAD's station network has been
operational since 1994, which is not long enough for climatology research (Nelson et al.,
2010). Price et al. (2013) investigated whether NEXRAD data, corrected with rain gauge data
by the Multisensor Precipitation Estimation (MPE) algorithm, would improve simulations in
watershed models; they found that adjusted radar precipitation estimates using gauge data
consistently performed better than non-adjusted radar data. NEXRAD data archive can be
accessed at NOAA's website.
NEXRAD Coverage Below 10,000 Feet AGL
w/sx »*<
'« «jAt'
' i
k*«
3*"' »*«.
wSs
VCP12 Coverage
j 4,000 ft above ground level*
I 6,000 ft above ground level*
J 10.000 ft above ground level*
• Bottom of beam height (assuming Standard Atmospheric Refraction),
Terrain blockage indicated where 50% or more of beam blocked
0 125 250 500 750 Miles
A 30
t .A*;
"A>
djr."s" A
T :
... J- * ¦*
1
Jht
t,
i-iS
", *rJl>
¦ Yj
fZ
1
/AM
w
Figure 3. A map showing the NEXRAD site locations in the contiguous United States. Image from (NOAA, 2017)
17
-------
4.3 Satellite
Satellite-based precipitation data are derived from infrared and microwave
measurements taken from satellites in space. Infrared information gives cloud top
temperature which can be used in an algorithm to produce rainfall amounts. The microwave
measurement gives information about cloud depth and layer characteristics which is
physically related to the formation of precipitation (Duan et al., 2016). Satellites are the only
way to retrieve global homogeneous estimates of precipitation (Tapiador et al., 2012). They
provide an estimate of precipitation in millimeters per hour at high resolutions over their span
of orbits. Some datasets use a single satellite for specific coverage (TRMM), while others use
an array of satellites like GPM for global coverage (Table 4). Combining diverse satellite
observations lowers the potential to miss precipitation events. Missing data can be found in
satellite-derived data due to instrument error or lack of spatial coverage. Satellite
observations cannot detect frozen precipitation or snowfall accumulation very well due to the
complexity of the radiative properties of snowflakes and ice crystals (Kidd & Huffman,
2011). Satellite-derived precipitation data tends to underestimate rain from warm top clouds
due to the infrared sensory tools used (Awange et al., 2016). Multilayer cloud systems also
pose a threat to miscalculations because cloud layers can block the sensor's ability to detect
the precipitating layer (Tapiador et al., 2012). This technology has been around for at least 30
years but, since each satellite-derived dataset has different temporal and spatial resolution,
period of activation and methods of calculating precipitation, each dataset has information
for only the time the satellite was operational. Although there are many satellite precipitation
datasets and satellite-derived products, only the four most recommended and most recent
satellite datasets are described below. The newest precipitation dataset, the GOES-16 satellite
which was launched in November 2016, will provide atmospheric measurements of Earth
with more spectral bands than its predecessors. For further information about the use of
satellite data in modeling see Bitew and Gebremichael (2011) who used PERSIANN and
CMORPH datasets in the hydrological model MIKE-SHE and Tramblay et al. (2016) who
compared CMORPH, RFE, TRMM, and PERSIANN in the a hydrological model for water
resource management in ungauged areas.
Table 4. Description of precipitation datasets that use satellite technology.
Name
Spatial
Coverage
Time
step
Spatial
Resolution
Time Frame
Category
Method
TRMM
35N, 35S
3hr
0.25x0.25
1998-2015
Single
Satellite
Microwave,
Infrared
GPM
60N 60S
0.5 hour
0.1x0.1
2014-present
Multi-
Satellite-
radar
Microwave,
Infrared, Satellite
Precip Radar
CMORPH
60N 60S
0.5, 3-
0.07277x0.07277
2002-present
Multi-
Morphing of
hour
, 0.25x0.25
Satellite
Microwave and
Infrared
PERSIANN
60N 60S
Hourly
0.04x0.04
2003-present
Multi-
Infrared & cloud
CCS
Satellite
segmentation
algorithm
18
-------
Name
Spatial
Coverage
Time
step
Spatial
Resolution
Time Frame
Category
Method
PERSIANN
60N 60S
Daily
0.25x0.25
1983-2015
Multi-
Infrared,
CDS
Satellite
Artificial Neural
Network
CHIRPS
50N, 50S
Daily
0.05x0.05
1981-present
Satellite-
gauge
Infrared with
station data
CMAP
Global
Monthly
2.5x2.5
1979-2006
Satellite-
gauge
Infrared with
station data
NLDAS
N.
America
Hourly
0.125x0.125
1979-present
Multi-
Satellite-
radar
Integration of
CMORPH and
RADAR
GLDAS
Global
3hr
0.25x0.25
2000-Dec 2016
Satellite-
gauge
Incorporation of
satellites and
ground-based
observations
4.3.1 TRMM
The Tropical Rainfall Measuring Mission (TRMM) was a single satellite used to
detect precipitation and tropical storms near the equator to better understand climate and
weather patterns. TRMM used microwave and infrared information to calculate precipitation
every three hours at a resolution of 0.25-degrees (see Figure 4). Although TRMM was
deactivated in 2015, its data from 1998 to 2015 has been used in numerous publications.
TRMM took tropical measurements of precipitation covering 35 degrees north to 35 degrees
south and had multiple reanalysis products. TRMM data can be accessed at NASA's website
(Skofronick-Jackson, 2017).
Figwe 4. TRAAl's orbit path captures the spatial resolution over the tropics on April 12th, 2012. Yellow areas depict probable
rainfall. Image from NASA.
19
-------
4.3.2 GPM
The Global Precipitation Mission (GPM) of 2014 is a continuation of TRMM's
mission. Partners from the United States, Japan, France, India and the European Union allow
GPM to track precipitation across the globe using 10 satellites. With a fine global resolution
of 0.1-degree taken every half hour, GPM provides precipitation measurements using
microwave, infrared, and radar technology. It has greater accuracy than its predecessor and
covers the earth between 60 degrees north and 60 degrees south. Due to calculation time, the
dataset can be accessed four to six hours after observation on NASA's website (Skofronick-
Jackson, 2017).
4.3.3 CMORPH
The Climate Prediction Center Morphing (CMORPH) product is another highly
recommended satellite dataset for precipitation. It gets its name from the method of
calculating precipitation by morphing microwave and infrared information from multiple
satellites at a 0.0730-degree resolution. Observations are taken every half hour and
accumulated every three hours with up to an 18-hour lag in data accessibility. CMORPH has
provided precipitation data since 2002 over the span of 60 degrees north to 60 degrees south.
CMORPH data can be accessed at website.
4.3.4 PERSIANN
The Precipitation Estimation from Remotely Sensed Information using Artificial
Neural Networks (PERSIANN) has satellite derived datasets calculated from infrared
imagery and artificial neural network algorithms (CHRS, 2004). This dataset's coverage is
between 60 degrees north and 60 degrees south. PERSIANN has two products that provide
precipitation data, one on an hourly time step (CCS) and one on a daily time step (CDR). The
PERSIANN CCS (Cloud Classification System) has a resolution of 0.04-degrees with data
from 2003 to the near present. Due to the complexity of algorithms, the CCS data can be
accessed one or two days after observation. PERSIANN CDR (Climate Data Record) has a
resolution of 0.25-degrees with data from 1983 to June 2016, with an even greater lag in
accessibility. All PERSIANN data can be accessed at CHRS's data portal.
4.4 Combining Datasets
Combining observed precipitation measurements often involves two or more types of
observations. An example of combined precipitation data are reanalysis products such as
North American Regional Reanalysis (NARR) and Climate Forecast System Reanalysis
(CFSR). Reanalysis products are generated when multiple precipitation datasets are merged
onto a regularly-spaced grid to produce a consistent spatiotemporal output (NOAA, 2017).
Blending and merging observed datasets can significantly improve precipitation estimates
(Ebert, 2007; NOAA, 2017). Since rain gauge data are often sparse in some areas and may
contain missing values, satellite and radar data has been combined with gauge data to fill the
20
-------
gaps, as in the Climate Prediction Center Merged Analysis of Precipitation (CMAP) and
Climate Hazards Group Infrared Precipitation with Station data (CHIRPS). In addition to the
previously stated datasets, there are many more datasets that combine multiple observational
methods including the popular LDAS and PRISM datasets that are described in detail below.
An example of the use of combined datasets is Radcliffe and Mukundan (2017) which
compared the effect of PRISM and CFSR, in a SWAT model of streamflow.
4.4.1 LDAS
The North American Land Data Assimilation System (NLDAS) combines North
American radar data and satellite data from CMORPH which allows higher resolutions and
better accuracy for detecting precipitation. NLDAS has an hourly time step on a 0.125 -
degree grid of North America and a maximum time lag of four days for data retrieval. The
Global Land Data Assimilation System (GLDAS) combines satellite data and ground-based
observational data to provide precipitation and other variables on a spatial resolution of 0.25-
degrees, covering the Earth between 90 degrees north and 60 degrees south. GLDAS data are
given every three hours and takes a least a month to process. EPA's BASINS system
combines NLDAS data with NCEI data for plugging in missing values in order to have a
near-seamless time series of precipitation data. Lee et al. (2010) compared NLDAS to NCDC
station data in the HSPF tool to improve streamflow predictions for water quality
assessments. NLDAS and GLDAS data can be accessed at NASA '$ website.
4.4.2 PRISM
The Parameter-elevation Relationship on Independent Slopes Model (PRISM) provides
climatology information by combining ground gauge stations from multiple sources and
radar products. The data is provided on a four by four kilometer spatial resolution covering
the contiguous United States from 1981 to present on a monthly or annul temporal resolution.
The method used to produce the gridded dataset is a combination of the Climatologically
Aided Interpolation (CAI) method, Digital Elevation Model (DEM), and radar interpolation.
With this methodology, the longer time-step is able to capture orographic precipitation
patterns in mountainous areas better than a daily interpolation (Daly et al., 2008). PRISM
data can be retrieved from the PRISM Climate Group website.
5. Simulated Data
Even after combining different types of observed data, there may still be missing
values in the datasets. Simulated data, based on computer models, can fill these gaps to
produce a continuous time series for model input. Weather prediction models are
mathematically-driven models that simulate precipitation from the past as well as the future.
Simulated precipitation measurements make computation easier for modelers because there
are no missing values nor time spent on data retrieval. Three main types of models simulate
precipitation data: Numerical Weather Predictors (NWP), stochastic models, and
nonparametric models, as described in Table 5. Simulating weather characteristics is difficult
21
-------
because climatic processes can occur below the grid size of the model which leads to
generalizations that introduce bias. Despite these limitations, some models closely mimic
true weather patterns and can be used to study and manage water quality and water supply
(Harmel et al., 2002). An example of the application of simulated precipitation data is the
paper by Golden et al. (2013). Golden et al. used simulated rainfall amounts from Global
Circulation Models in three watershed models (VELMA, GBMM, TOPLOAD) to determine
how future changes in climate may impact watershed mercury transport.
Table 5. Summary of simulated precipitation dataset characteristics.
Simulated
Method
Inputs
Spatial
Resolution
Spatial Extent
Error
Numerical
Cumulus and
Atmospheric
0.03x0.03-degree
Global or
Scheme
Weather
microphysics
conditions and
grid 0.703x0.703-
limited area
selection
Prediction
schemes
thresholds
degree grid
model (LAM)
Stochastic
Probability
20-year history of
precipitation
Site specific
Global or
delineated area
Skewed
distributions
Non-
Historical trends
Long historical
0.35x0.35-degree
Global or
Model drift
Parametric
records and
grid 1.4x1.4-
regional
from observed
emission
degree grid
downscaled
data
scenarios
5.1 Numerical Weather Prediction (NWP)
Numerical Weather Prediction models integrate differential equations that describe
fluid flows to predict rainfall and other atmospheric conditions. Two major types of
equations used to estimate precipitation describe microphysics and cumulus clouds.
Microphysics parameterization schemes resolve the process of rain production, and cumulus
parameterization schemes describe effects of cumulus clouds in rain events. Combining them
determines rainfall occurrences and amount (Yang et al., 2015). Many schemes are needed
to produce a set of rainfall predictions: for example, the Weather Research and Forecasting
(WRF) model uses seven microphysics parameterization schemes and three cumulus
parameterization schemes. More details on specific schemes can be found in Yang et al.
(2015). Scheme selection is a main source of error within NWP models because certain
schemes work better in certain circumstances. Some physical processes in rain production
occur at small scales or in specific climates and cannot be properly modeled over a larger
resolution. Two frequently used numerical weather prediction models are described below.
5.1.1 WRF Model
The Weather Research and Forecasting (WRF) model, known previously as the fifth
generation Mesoscale Model (MM5), is a numerical weather predictor used in climate
22
-------
forecasting models such as CMAQ4, NRCM5, and NCEP Eta6. The Kain-Fritsch Scheme is a
cumulus scheme in WRF that models precipitation based on condensation exceeding a
threshold value (Yang et al., 2015). WRF uses past observation data or idealized atmospheric
conditions and thresholds in schemes to generate rainfall. The daily precipitation output has a
fine resolution of 0.03-degree grid that can be scaled to fit the modeler's area of interest. The
source code for WRF can be found on LICAR's website.
5.1.2 ECHAM
The European Centre Hamburg Model (ECHAM) is a numerical weather prediction
method used as the atmospheric model in climate models such as MPI-ESM7 and ECHO-G8.
ECHAM6 is the latest version using parameterization schemes include mass flux in cumulus
convection and cloud microphysics to determine daily precipitation. A detailed description of
the model can be found in Stevens et al. (2013) and Giorgetta et al. (2013). ECHAM has
multiple sets of spatial resolutions with the finest resolution of 0.7031x0.7031 degrees for
experimental use. ECHAM source code is freely available to the public at the Max Planck
Institute for Meteorology website.
5.2 Stochastic
Stochastic models, among the simplest prediction models, use statistics and
probabilities associated with weather data to predict atmospheric parameters (Harmel et al.,
2002). Model output generates data that is statistically consistent with the observed data
input. These models generate daily weather at a single point location or through a more
complicated process of multi-site generation (Mehrotra et al., 2006). To generate
precipitation, a Markov Chain Model determines the probability of having a wet day or a dry
day, then finds the probability of a wet day following a dry or wet day (Wilks & Wilby,
1999). Historical precipitation measurements of 20 years or more is recommended to initiate
the Markov chain; then, an equation using mean daily rainfall, standard deviation of daily
rainfall, and a skew coefficient gives the amount of rainfall on a particular wet day.
Stochastic models often fail to accurately describe the length of dry or wet periods and model
output can be skewed based on historical input data, thus requiring statistical verification.
4 Community Multiscale Air Quality Modeling System littps://www.epa.gov/emaq
5 Nested Regional Climate Model https://rda.ncar.edu/datasets/ds601.0/
6 National Center for Environmental Prediction Eta model https://rda.ncar.edn/datasets/ds609.2/
7 Max-Planck-Institute Earth Systems Model (Stevens et al., 2013)
https://www.mpimet.mpg.de/eti/science/models/mpi-esm/
8 ECHAM4 Atmospheric model coupled withHOPE-G oceanic model ("Lawrence Livermore National Laboratory
Program for Climate Model Diagnosis and Intercomparison," 2005)
23
-------
5.2.1 Weather Generators (WGEN)
Weather Generators are used for statistically simulating atmospheric conditions in
many models. The Water Erosion Precipitation Project (WEPP) and the Soil and Water
Assessment Tool (SWAT) both use stochastic weather generators. SWAT's default is
Cooperative Observer Network (COOP) gauge data from 1960 to 2010, for historical
reference, to begin the Markov Chain Model (Tuo et al., 2016). Weather generators create
precipitation accumulation for the specified day, month, and year. SWAT can predict
precipitation accumulation over a delineated area of interest, out to year 2100. The WGEN
algorithm and Fortran code can be found in Richardson (1984); variations of this stochastic
model have been created with the same internal structure as WGEN— for instance,
WXGEN9, CLIGEN10, and GEM11. Comparing results of different weather generators
produces different predictions for each weather simulator due to their stochasticity
(Migliaccio & Srivastava, 2007).
5.3 Nonparametric Models
Nonparametric models resample historic data to find trends and weather
characteristics for future data. They can be thought of as smoothed, conditional bootstrapping
or kernel density estimates (Rajagopalan et al., 1997). Nonparametric simulations use large
numbers of observational data to create a probability density function that best describes the
data (Sharma, 2000). A Gaussian kernel function is commonly used to describe weather
patterns (Rajagopalan et al., 1997; Sharma, 2000). Nonparametric models produce only
values that occurred from the historical dataset, but data may be regenerated that violates a
boundary condition, such as the rain versus snow threshold temperature (Rajagopalan et al.,
1997). This can cause error in the model's product, so careful consideration of input data is
important. It is assumed (but cannot be guaranteed) that models which accurately predict
historic weather patterns are more likely to accurately predict future weather patterns (Rupp
et al., 2013).
5.3.1 GCM
Global Circulation Models (GCM) use a combination of nonparametric trends and
numeric predictions to generate precipitation on a global scale. GCMs are good
representations of temporal trends on a large scale, but often vary when downscaled to a
regional level. The Coupled Model Intercomparison Project Phase 5 (CMIP5) was designed
to evaluate how realistic 20+ GCMs are at recreating past climate data, projecting future
climate change, and understanding differences among models (Taylor, 2009). Each model is
given inputs of historical climate data from 1800- 2005 and future emission scenarios to
9 Erosion/Productivity Impact Calculator Weather Generator (Wallis & Griffiths, 1995)
10 USDA's Climate Generator (Meyer, 2004)
11 Generation of weather Elements for Multiple applications (Harmel et al., 2002)
24
-------
simulate near-term (to year 2035) or long-term (to year 2100) weather conditions. The
Intergovernmental Panel on Climate Change (IPCC) provided CMIP5 with four future
atmospheric scenarios represented by radiative forcing values in year 2100 called
Representative Concentration Pathways (RCP) (Taylor, 2009). Each GCM used these
scenarios to predict weather conditions:
RCP2.6 assumes that greenhouse gas emissions will peak around 2030, then decline;
RCP4.5 assumes the peak will be around 2050, then level off at 4.5W/m2;
RCP6.0 assumes the peak will be around 2090, then level off at 6W/m2;
RCP8.5 estimates emission will continuously rise throughout the 21st century;
This set of scenarios is the newest decision from IPCC on future climate scenarios.
RCP 4.5, 6, and 8.5 are comparable to Special Report on Emission Scenarios (SRES) Bl, B2,
A1F1, respectively. Simulated precipitation data can be retrieved from a specific model or as
a multi-model mean. An assessment of CMIP5 models based on observational data for
reliability can be found in Rupp et al. (2013). The CNRM-CM5 model within CMIP5
performed with the least error in reproducing global precipitation, followed by CESM1-
CAM5 (Rupp et al., 2013). Please note the experimental design for CMIP6 has only been
recently published, and data is available but is still being evaluated (Eyring et al., 2016).
CM IPS data can be downloaded at the WorldClim website.
6. Discussion
Precipitation is a difficult variable to measure precisely. In calibrating the SWAT
model for river basin modeling, Tuo et al. (2016) found that precipitation is the main source
of uncertainty. Observed and simulated precipitation datasets have strengths and weaknesses
in providing an accurate representation of rainfall amounts. Simulated datasets from
numerical weather prediction, stochastic models, and nonparametric models provide a
seamless time series and perform well in cold weather, mountainous regions, and non-
extreme conditions. Simulated future data are applicable for managing and planning purposes
since there is a need for information about changes in future precipitation. Observed datasets
often include gaps in the time series due to lack of measurement, but they perform best in
warm conditions and reflect extreme weather events well. Direct rainfall measurement from
rain gauges is preferred by researchers since assumptions are not made and they have long
measurement records. Many studies comparing differences in precipitation datasets for
regional analysis have been performed, (e.g., Costa & Foley, 1998; Fekete et al., 2004;
Tapiador et al., 2012).
Precipitation plays a large role in the availability of drinking water, erosion, and
transportation of contaminants. Selection of precipitation data has crucial effects on
hydrological model performance; thus, choosing precipitation datasets based on method, time
step, and resolution needs to be carefully thought out (Tuo et al., 2016). Regulatory,
planning, and exploratory purposes require different levels of uncertainty. Regulatory
projects must have very little error and uncertainty while exploratory projects encourage
uncertainty. A need for advancements in precipitation accuracy, length of record, and free
availability is still a recurring problem in the modeling community. There is no "best"
25
-------
precipitation dataset, only the most appropriate for a given purpose. As Harm el et al. (2002)
said, "Historical data provide only one realization or 'picture' of a previous weather pattern
that may not represent future climate scenarios
26
-------
References
Awange, J. L., Ferreira, V. G., Forootan, E., Khandu, Andam-Akorful, S. A., Agutu, N. O., &
He, X. F. (2016). Uncertainties in remotely sensed precipitation data over Africa.
International Journal of Climatology, 36(1), 303-323. doi: 10.1002/joc.4346
Becker. (2013). A description of the global land-surface precipitation data products of the Global
Precipitation Climatology Centre with sample applications including centennial (trend)
analysis from 1901-present.
Bitew, M. M., & Gebremichael, M. (2011). Evaluation of satellite rainfall products through
hydrologic simulation in a fully distributed hydrologic model. Water Resources
Research, 47(6). doi: 10.1029/2010wr009917
Brewster, R. (2017). Paint.net 4.0.17 (Version 4.0.17). Retrieved from https://www.getpaint.net/
Chahine, M. T. (1992). The Hydrological Cycle and Its Influence on Climate. Nature, 359(6394),
373-380. doi: Doi 10.1038/359373a0
CHRS. (2004). Center for Hydrometeorology & Remote Sensing. Satellite Precipitation-
Activities Retrieved June 4, 2017, from
http://chrs.web.uci.edu/research/satellite precipitati on/acti viti esOO. htm 1
Costa, M. H., & Foley, J. A. (1998). A comparison of precipitation datasets for the Amazon
basin. Geophysical Research Letters, 25(2), 155-158. doi: Doi 10.1029/97gl03502
Daly, C., Halbleib, M., Smith, J. I., Gibson, W. P., Doggett, M. K., Taylor, G. H., . . . Pasteris, P.
P. (2008). Physiographically sensitive mapping of climatological temperature and
precipitation across the conterminous United States. International Journal of
Climatology, 25(15), 2031-2064. doi: 10.1002/joc.l688
Duan, Z., Liu, J., Tuo, Y., Chiogna, G., & Disse, M. (2016). Evaluation of eight high spatial
resolution gridded precipitation products in Adige Basin (Italy) at multiple temporal and
spatial scales. Science of the Total Environment, 573, 1536-1553. doi:
10.1016/j.scitotenv.2016.08.213
Ebert, E. J., John; Kidd, Chris. (2007). Comparison of Near-Real-Time Precipitation Estimates
from Satellite Observations and Numerical Models. American Meteorological Society.
doi: 10.1175/BAMS-88-1-47
ESRI. (2015). ArcGIS Release 10.3.1 Redlands, CA: Environmental Systems Research Institute.
Eyring, V., Bony, S., Meehl, G. A., Senior, C. A., Stevens, B., Stouffer, R. J., & Taylor, K. E.
(2016). Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6)
experimental design and organization. Geoscientific Model Development, 9(5), 1937-
1958. doi: 10.5194/gmd-9-1937-2016
Fekete, B. M., Vorosmarty, C. J., Roads, J. O., & Willmott, C. J. (2004). Uncertainties in
precipitation and their impacts on runoff estimates. Journal of Climate, 17(2), 294-304.
doi: Doi 10.1175/1520-0442(2004)017<0294:Uipati>2.0.Co;2
Gao, J. G., Sheshukov, A. Y., Yen, H., & White, M. J. (2017). Impacts of alternative climate
information on hydrologic processes with SWAT: A comparison of NCDC, PRISM and
NEXRAD datasets. CATENA, 156, 353-364. doi: 10.1016/j.catena.2017.04.010
Giorgetta, M., Roeckner, E., Mauritsen, T., Bader, J., Crueger, T., Esch, M., . . . Stevens, B.
(2013). The atmospheric general circulation model ECHAM6: Model Description.
Reports on Earth System Science( 135).
27
-------
Golden, H. E., Knightes, C. D., Conrads, P. A., Feaster, T. D., Davis, G. M., Benedict, S. T., &
Bradley, P. M. (2013). Climate change and watershed mercury export: a multiple
projection and model analysis. Environ Toxicol Chem, 32(9), 2165-2174. doi:
10.1002/etc.2284
Golden, H. E., Knightes, C. D., Cooter, E. J., Dennis, R. L., Gilliam, R. C., & Foley, K. M.
(2010). Linking air quality and watershed models for environmental assessments:
Analysis of the effects of model-specific precipitation estimates on calculated water flux.
Environmental Modelling & Software, 25(12), 1722-1737. doi:
10.1016/j.envsoft.2010.04.015
Harmel, R. D., Johnson, G., & Richardson, C. W. (2002). The GEM experience: Weather
generator technology development in the USD A. Bulletin of the American
Meteorological Society, 83(1), 954-957. doi: Doi 10.1175/1520-
0477(2002)083<0954:Tgewgt>2.3.Co;2
Harmel, R. D., Smith, P. K., Migliaccio, K. W., Chaubey, I., Douglas-Mankin, K. R., Benham,
B., . . . Robson, B. J. (2014). Evaluating, interpreting, and communicating performance of
hydrologic/water quality models considering intended use: A review and
recommendations. Environmental Modelling & Software, 57, 40-51. doi:
10.1016/j.envsoft.2014.02.013
Hu, Q., Yang, D., Li, Z., Mishra, A. K., Wang, Y., & Yang, H. (2014). Multi-scale evaluation of
six high-resolution satellite monthly rainfall estimates over a humid region in China with
dense rain gauges. International Journal of Remote Sensing, 35(4), 1272-1294. doi:
10.1080/01431161.2013.876118
Huffman, G. J., Adler, R. F., Rudolf, B., Schneider, U., & Keehn, P. R. (1995). Global
Precipitation Estimates Based on a Technique for Combining Satellite-Based Estimates,
Rain-Gauge Analysis, and Nwp Model Precipitation Information. Journal of Climate,
8(5), 1284-1295. doi: Doi 10.1175/1520-0442(1995)008<1284:Gpeboa>2.0.Co;2
Kidd, C., & Huffman, G. (2011). Review Global Precipitation Measurement. Meteorological
Applications, 18(3), 334-353. doi: 10.1002/met.284
Kim, K. P., Katie; Whelan, Gene; Wolfe, Kurt; Duda, Paul; Gray, Mark; Pachepsky, Yakov.
(2014). Using Remote Sensing and Radar MET Data to Support Watershed Assessment
Comprising IEM.
Lawrence Livermore National Laboratory Program for Climate Model Diagnosis and
Intercomparison. (2005). Model Information of Potential Use to the IPCC Lead Authors
and AR4: ECHO-G. Retrieved June 6, 2017, from http://www-
pcmdi.llnl.eov/ipcc/model documentation/ECHO-G.htm
Lee, S., Ni-Mesister, W., Toll, D., Nigro, J., Gutierrez-Magness, A. L., & Engman, T. (2010).
Assessing the hydrologic performance of the EPA's nonpoint source water quality
assessment decision support tool using North American Land Data Assimilation System
(NLDAS) products. Journal of Hydrology, 387(3), 212-220. doi:
http s: //doi. or of 10.1016/j. j hydrol. 2 C» 10 0 1.009
Liu, Z., Kingery, W. L., Huddleston, D. H., Hossain, F., Chen, W., Hashim, N. B., & Kieffer, J.
M. (2008). Modeling nutrient dynamics under critical flow conditions in three tributaries
of St. Louis Bay. J Environ Sci Health A Tox Hazard Subst Environ Eng, 43(6), 633-645.
doi: 10.1080/10934520801893725
28
-------
Mehrotra, R., Srikanthan, R., & Sharma, A. (2006). A comparison of three stochastic multi-site
precipitation occurrence generators. Journal of Hydrology, 337(1-2), 280-292. doi:
10.1016/j.jhydrol.2006.05.016
Meyer, C. (2004). General Description of the CLIGEN Model and its History. West Lafayette,
IN: Retrieved from
https://www.ars.usda. e terFiles/50201000/WEPP/cligen/CLIGENDescri pti on .p
df.
Migliaccio, K. W., & Srivastava, P. (2007). Hydrologic components of watershed-scale models.
Transactions of the Asabe, 50(5), 1695-1703.
Nelson, B. R., Seo, D. J., & Kim, D. (2010). Multisensor Precipitation Reanalysis. Journal of
Hydrometeorology, 77(3), 666-682. doi: 10.1175/2010JHM1210.1
NOAA. (2017). National Oceanic and Atmospheric Administration. National Centers for
Environmental Information. Data Access. Retrieved 04/20, 2017, from
https://www.ncdc.noaa.eov/data-access
Price, K., Purucker, S. T., Kraemer, S. R., Babendreier, J. E., & Knightes, C. D. (2014).
Comparison of radar and gauge precipitation data in watershed models across varying
spatial and temporal scales. HydrologicalProcesses, 28(9), 3505-3520. doi:
10.1002/hyp.9890
Radcliffe, D. E., & Mukundan, R. (2017). PRISMvs. CFSR Precipitation Data Effects on
Calibration and Validation of SWAT Models. JAWRA Journal of the American Water
Resources Association, 53(1), 89-100. doi: 10.1111/1752-1688.12484
Rajagopalan, B., Lall, U., Tarboton, D. G., & Bowles, D. S. (1997). Multivariate nonparametric
resampling scheme for generation of daily weather variables. Stochastic Hydrology and
Hydraulics, 77(1), 65-93. doi: Doi 10.1007/Bft)2428426
Richardson, C. W. W., D. A. (1984). WGEN: A Model for Generating Daily Weather Variables.
Rupp, D. E., Abatzoglou, J. T., Hegewisch, K. C., & Mote, P. W. (2013). Evaluation of CMIP5
20th century climate simulations for the Pacific Northwest USA. Journal of Geophysical
Research: Atmospheres, 775(19), 10,884-810,906. doi: 10.1002/jgrd.50843
Sharma, A. (2000). Seasonal to interannual rainfall probabilistic forecasts for improved water
supply management: Part 3 - A nonparametric probabilistic forecast model. Journal of
Hydrology, 239(1-4), 249-258. doi: Doi 10.1016/S0022-1694(00)00348-6
Shedekar, V. S., King, K. W., Fausey, N. R., Soboyejo, A. B. O., Harmel, R. D., & Brown, L. C.
(2016). Assessment of measurement errors and dynamic calibration methods for three
different tipping bucket rain gauges. Atmospheric Research, 178, 445-458. doi:
10.1016/j.atmosres.2016.04.016
Skofronick-Jackson, G. (2017). Precipitation Measurement Missions. Retrieved 6/30/2017,
from https://pmm.nasa.gov/precipitation-measurement-missions
Stevens, B., Giorgetta, M., Esch, M., Mauritsen, T., Crueger, T., Rast, S., . . . Roeckner, E.
(2013). Atmospheric component of the MPI-M Earth System Model: ECHAM6. Journal
of Advances in Modeling Earth Systems, 5(2), 146-172. doi: 10.1002/jame.20015
Tapiador, F. J., Turk, F. J., Petersen, W., Hou, A. Y., Garcia-Ortega, E., Machado, L. A. T., . . .
de Castro, M. (2012). Global precipitation measurement: Methods, datasets and
applications. Atmospheric Research, 104-105, 70-97. doi:
10.1016/j.atmosres.2011.10.021
Taylor, K. S., Ronald; Meehl, Gerald. (2009). A Summary of the CMIP5 Experiment Design. 1-
33.
29
-------
Thornton, P. E., M.M. Thornton, B.W. Mayer, Y. Wei, R. Devarakonda, R.S. Vose, and R.B.
Cook. (2017, 2017-04-19). Daymet: Daily Surface Weather and Climatological
Summaries. Daymet V3. from https://daymet.oml.eov/
Tramblay, Y., Thiemig, V., Dezetter, A., & Hanich, L. (2016). Evaluation of satellite-based
rainfall products for hydrological modelling in Morocco. Hydrological Sciences Journal,
61(14), 2509-2519. doi: 10.1080/02626667.2016.1154149
Tuo, Y., Duan, Z., Disse, M., & Chiogna, G. (2016). Evaluation of precipitation input for SWAT
modeling in Alpine catchment: A case study in the Adige river basin (Italy). Science of
The Total Environment, 573, 66-82. doi: 10.1016/j.scitotenv.2016.08.034
U.S. Geological Survey (USGS). 2011. National Elevation Dataset (NED). Sioux Falls, SD. In:
U.S. Environmental Protection Agency (USEPA) and the U.S. Geological Survey
(USGS). 2012. National Hydrography Dataset Plus - NHDPlus, V2.10. Available at:
https://www.epa.gOv/waterdata/nhdplus-national-hydrographv-dataset-plus#v2datamap
Wallis, T. W. R., & Griffiths, J. F. (1995). An assessment of the weather generator (WXGEN)
used in the erosion/productivity impact calculator (EPIC). Agricultural and Forest
Meteorology, 73(1), 115-133. doi: http://dx.doi.ore/10.1016/0168-1923(94)02172-G
Wilks, D. S., & Wilby, R. L. (1999). The weather generation game: a review of stochastic
weather models. Progress in Physical Geography, 23(3), 329-357. doi: Doi
10.1177/030913339902300302
Yang, M.-x., Jiang, Y.-z., Lu, X., Zhao, H.-l., Ye, Y.-t., & Tian, Y. (2015). A weather research
and forecasting model evaluation for simulating heavy precipitation over the downstream
area of the Yalong River Basin. Journal ofZhejiang University SCIENCE A, 16(1), 18-
37. doi: 10.1631/jzus.A1400347
-------
SEPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research
and Development
(8101R)
Washington, DC
20460
Official Business
Penalty for Private Use
$300
31
------- |