A Comparison of Temperature Data from Automated and Manual Observing Networks in Georgia and Impacts of Siting Characteristics

Jason Allard Department of Physics, Astronomy and Geosciences, Valdosta State University, Valdosta, Georgia

Search for other papers by Jason Allard in
Current site
Google Scholar
PubMed
Close
,
Paul C. Vincent Department of Physics, Astronomy and Geosciences, Valdosta State University, Valdosta, Georgia

Search for other papers by Paul C. Vincent in
Current site
Google Scholar
PubMed
Close
,
Jeromy R. McElwaney Department of Physics, Astronomy and Geosciences, Valdosta State University, Valdosta, Georgia

Search for other papers by Jeromy R. McElwaney in
Current site
Google Scholar
PubMed
Close
, and
Gerrit Hoogenboom AgWeatherNet Program, Washington State University, Prosser, Washington

Search for other papers by Gerrit Hoogenboom in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The objectives of this study were to compare average monthly and seasonal maximum and minimum temperatures of the Georgia Automated Environmental Monitoring Network (AEMN) to those of geographically close (i.e., paired) manual observations from U.S. Historical Climatology Network (USHCN) stations and Cooperative Observer Program (COOP) stations for the period 2002–13, and to evaluate the extent to which differences in siting characteristics of paired AEMN–USHCN stations contribute to the temperature differences. Correlations for monthly and seasonal maximum and minimum temperatures of paired AEMN–USHCN and AEMN–COOP stations were high and almost always significant, although the correlations for seasonal minimum temperatures were slightly lower than those of maximum temperatures, especially for summer. Monthly maximum and minimum temperatures and seasonal maximum temperatures of paired AEMN–USHCN and AEMN–COOP stations were significantly different in only a few instances, while seasonal minimum temperatures were more often significantly different, particularly for summer. The stronger relationship between maximum temperatures than minimum temperatures for paired stations is logical given that minimum temperatures typically occur when a shallow, decoupled nocturnal boundary layer is more sensitive to local conditions [e.g., land use/land cover (LULC)]. Stepwise regressions confirmed that a portion of the variance of seasonal minimum temperatures of paired AEMN–USHCN stations was explained by differences in LULC, while the variance in seasonal maximum temperatures was explained better by differences in elevation. Despite the generally close relationships between temperatures of paired stations and a portion of the differences being explained, an abrupt change from manual networks to the AEMN without data adjustments would change the Georgia climate record on monthly and seasonal time scales.

Current affiliation: Institute for Sustainable Food Systems, University of Florida, Gainesville, Florida.

Corresponding author address: Dr. Jason Allard, Department of Physics, Astronomy and Geosciences, Valdosta State University, 1500 North Patterson Street, Valdosta, GA 31698-0055. E-mail: jmallard@valdosta.edu

Abstract

The objectives of this study were to compare average monthly and seasonal maximum and minimum temperatures of the Georgia Automated Environmental Monitoring Network (AEMN) to those of geographically close (i.e., paired) manual observations from U.S. Historical Climatology Network (USHCN) stations and Cooperative Observer Program (COOP) stations for the period 2002–13, and to evaluate the extent to which differences in siting characteristics of paired AEMN–USHCN stations contribute to the temperature differences. Correlations for monthly and seasonal maximum and minimum temperatures of paired AEMN–USHCN and AEMN–COOP stations were high and almost always significant, although the correlations for seasonal minimum temperatures were slightly lower than those of maximum temperatures, especially for summer. Monthly maximum and minimum temperatures and seasonal maximum temperatures of paired AEMN–USHCN and AEMN–COOP stations were significantly different in only a few instances, while seasonal minimum temperatures were more often significantly different, particularly for summer. The stronger relationship between maximum temperatures than minimum temperatures for paired stations is logical given that minimum temperatures typically occur when a shallow, decoupled nocturnal boundary layer is more sensitive to local conditions [e.g., land use/land cover (LULC)]. Stepwise regressions confirmed that a portion of the variance of seasonal minimum temperatures of paired AEMN–USHCN stations was explained by differences in LULC, while the variance in seasonal maximum temperatures was explained better by differences in elevation. Despite the generally close relationships between temperatures of paired stations and a portion of the differences being explained, an abrupt change from manual networks to the AEMN without data adjustments would change the Georgia climate record on monthly and seasonal time scales.

Current affiliation: Institute for Sustainable Food Systems, University of Florida, Gainesville, Florida.

Corresponding author address: Dr. Jason Allard, Department of Physics, Astronomy and Geosciences, Valdosta State University, 1500 North Patterson Street, Valdosta, GA 31698-0055. E-mail: jmallard@valdosta.edu

1. Introduction

The National Weather Service (NWS) Cooperative Observer Program (COOP) network, a volunteer network of citizens and institutions, has provided daily observations of temperature and precipitation across the United States for more than 100 years (National Research Council 1998, 2009). At the beginning of the twenty-first century, the COOP network consisted of more than 11 000 stations, with between 5000 and 6000 observers measuring air temperature, and between 7500 and 10 000 observers measuring precipitation (Fiebrich 2009; National Research Council 2009). The applications of these data have expanded substantially since the inception of the network and include, but are not limited to, agricultural planning and management, the design and maintenance of infrastructure, management of water resources, environmental impacts assessments, and meteorological forecasts (see Changnon and Kunkel 1999; National Research Council 2009). Because of the geographic distribution of its observations and its relative stability over time (i.e., some stations have periods of records with over 100 years of data), the COOP network is also considered an authoritative source of temperature and precipitation data for monitoring local-to-nationwide climate variations and trends (National Research Council 1998; Wu et al. 2005; Holder et al. 2006; National Research Council 2009).

Given the need for an accurate and unbiased long-term climate record for climate change detection for the United States, the National Centers for Environmental Information (NCEI), formerly known as the National Climatic Data Center, and the Global Change Research Program of the U.S. Department of Energy developed the U.S. Historical Climatology Network (USHCN) (Quinlan et al. 1987; Karl et al. 1990; Hughes et al. 1992; Easterling et al. 1996; Menne et al. 2009, 2014). The stations included in the USHCN are a subset of those in the COOP network, and they were originally selected due to their long period of record, small percentage of missing data, spatial coverage, and limited station movement, or other station changes that could affect data homogeneity (Menne et al. 2009, 2014). Adjustments to the data in the USHCN dataset were then made to account for systematic biases associated with missing values (Menne et al. 2009), changes in the time of observation (Karl et al. 1986; Vose et al. 2003), changes in instrumentation (Karl and Williams 1987; Quayle et al. 1991; Menne et al. 2009), and changes in station location and exposure (Menne and Williams 2009; Menne et al. 2009). As such, the USHCN consists of COOP stations that have been adjusted for systematic, nonclimatic changes that bias climate trends (Menne et al. 2009), and it provides a dataset to document climate trends and variability over the past century and beyond.

Despite the importance of the COOP network for monitoring air temperature and precipitation, and the USHCN for detecting long-term changes in climate, a review of the COOP network conducted by the National Research Council (NRC) in 1998 concluded that “over the past several years the Coop Network has been weakened by a combination of technological, organizational, and budgetary factors” (National Research Council 1998, p. 1). Errors associated with manual observations and data entry, the cumbersome nature and slowness of disseminating the climate data, and the steady decline in the number of COOP stations providing observations has put the network in danger of becoming unreliable for climate monitoring (National Research Council 1998; U.S. Department of Commerce 2004). Since reaching a peak of nearly 14 000 observers in 1972, the number of COOP stations had declined nearly 15% by the late twentieth century (National Research Council, 1998). Even USHCN stations—the most stable group of COOP stations—are losing approximately 1% of its sites annually due to station closures or relocation, necessitating revisions to the stations included in the network (National Research Council 1998; U.S. Department of Commerce 2004; Fiebrich 2009; Menne et al. 2009). The need to seek solutions and to address these issues to maintain the more than 100-yr legacy of the networks led, at least in part, to efforts to modernize the COOP network in the mid-2000s (e.g., high-quality sensors and observing standards, installation of automated NWS sites, rigorous quality assurance of data, increased temporal resolution of observations) (U.S. Department of Commerce 2004; Rothfusz et al. 2006; National Research Council 2009).

Efforts to modernize the existing COOP network and to develop additional weather and climate monitoring networks (e.g., automated mesonets) have also arisen from the dramatic increase in the applications of climate data in recent decades that necessitate finer spatial and temporal resolutions (Changnon and Kunkel 1999; U.S. Department of Commerce 2004; National Research Council 2009). As requirements for meteorological observations at the mesoscale increase to meet the needs of users (e.g., mesoscale weather and air quality forecasting, agricultural management practices, and decision-making in sectors such as transportation), data are also required at near–real time for more environmental variables than are collected at COOP stations (e.g., wind speed and direction, soil parameters, and solar radiation) (National Research Council 1998, 2009). With advancements in relatively inexpensive microelectronics, computers, and communication technologies, there has been an increase in the number of relatively dense networks of automated meteorological observing systems (i.e., mesonets) (Changnon and Kunkel 1999; Hubbard 2001; Fiebrich 2009; National Research Council 2009). These automated networks of stations have the benefit of increasing the frequency of observations, providing near-real-time observations of weather, permitting data collection from a wider variety of sensors, eliminating manual input errors, and in some cases increasing the spatial coverage of weather observations (National Research Council 1998).

The number of automated weather stations collecting climatic data has been increasing since the later part of the twentieth century at local, state, regional, and national levels (National Research Council 1998). By the beginning of the twenty-first century, the National Oceanic and Atmospheric Administration’s (NOAA) Automated Surface Observing System (ASOS) was the largest in the United States. Several statewide networks of automated weather stations have also been operating since the 1980s and 1990s, and numerous other states are forming or operating automated weather networks (Fiebrich 2009). Some of the states in the southeastern United States with active mesonets that report at least hourly observations for a range or climate variables include Alabama (Kimball et al. 2010); Florida (http://fawn.ifas.ufl.edu/); Georgia (Hoogenboom 2005; Rothfusz et al. 2006; www.weather.uga.edu); Kentucky (Brown et al. 2008; www.kymesonet.org); Louisiana (http://weather.lsuagcenter.com/); Mississippi (White and Matlack 2007); and North Carolina (Holder et al. 2006; www.nc-climate.ncsu.edu/econet).

With the growth and increased use of automated weather stations and closures of COOP and USHCN stations, there is an increasing need to ascertain the comparability of automated data with manual observations (e.g., COOP) (Wu et al. 2005; Holder et al. 2006; Fiebrich and Crawford 2009). Ideally, changes in monitoring systems would maintain continuity with previous systems to ensure homogeneity in the datasets (Guttman and Baker 1996; Hubbard et al. 2004). The implementations of automated networks, however, were often designed to meet the specific needs of the organization installing the automated systems but not necessarily to maintain the long-term climate record (Brusberg and Hubbard 2001; Horel et al. 2002). The differences in sensors and shelters, observing practices, data processing algorithms, and site characteristics between the automated and manual observing systems can therefore produce differences in their time series that do not represent changes in climate (Guttman and Baker 1996; Peterson et al. 1998; Hubbard et al. 2004; Wu et al. 2005; Holder et al. 2006; Hale et al. 2008; Bodine et al. 2009; Fiebrich and Crawford 2009; Menne et al. 2010; Trewin 2010; Misra et al. 2012). As such, it is critical from a climatological standpoint to compare data from collocated automated and manual stations (i.e., geographically close stations are more likely to experience similar climates) to assess how a transition to automated observations affects the quality and homogeneity of the long-term climate records used to study climate change (Guttman and Baker 1996; Wu et al. 2005; Holder et al. 2006; Fiebrich and Crawford 2009).

Unfortunately, the number of studies that have compared in situ air temperature data from collocated automated and manual networks is less than would be ideal. Guttman and Baker (1996) compared hourly temperature data from 10 collocated COOP and ASOS stations for a 1-yr period. They found that differences in site characteristics (e.g., ground cover, topography) produced much larger discrepancies between the temperature measurements of the two systems (i.e., a couple of degrees Fahrenheit) than did differences in instrumentation (i.e., a few tenths of a degree). The authors concluded that creating a homogeneous dataset from the two instrument systems for climatological purposes would be a difficult task. Hubbard et al. (2004) conducted a 1-yr side-by-side comparison of U.S. Climate Reference Network (USCRN) temperature measurements and temperatures measured by the maximum–minimum temperature system (MMTS) used by COOP stations. Analyses of the data indicated that the MMTS temperatures performed with biases introduced by ambient air temperatures, wind speed, and solar radiation. The magnitudes of these biases ranged from a few tenths of a degree to over 1°C when compared to the USCRN temperatures. Leeper et al. (2015) compared 12 collocated (i.e., <400 m) USCRN–COOP station pairs, and found that the naturally aspirated COOP sensors generally had warmer (~0.48°C) daily maximum temperatures and cooler (~0.36°C) daily minimum temperatures, although there was substantial variability among stations. The authors partially attributed temperature differences to local siting characteristics (e.g., station exposure, ground cover, geographical aspect), and noted that COOP observer inconsistences (e.g., varying observation times, multiday observations, recording errors) and sensor errors led to biases that complicated network comparisons.

The number of studies that have compared temperature data from state-based mesonets with those from manual observing networks is similarly limited. Wu et al. (2005) compared daily temperature observations for 28 pairs of geographically close (<10 km) COOP and Automated Weather Data Network (AWDN) stations in Nebraska. The authors found that significant discrepancies exist between daily temperature observations for most paired stations (i.e., root-mean-square errors of ~6° and ~3°C for maximum and minimum temperatures, respectively) because of differing observation times, observation error, sensor error, and differences in microclimatic exposure. Similarly, Fiebrich and Crawford (2009) demonstrated that differences exist between daily temperatures recorded by the COOP network and the Oklahoma Mesonet over a 3-yr period for collocated stations (≤8 km between paired COOP and automated stations). Daily differences in temperature for paired stations, sometimes exceeding 5°C and differing by more than 1°C for more than 55% of the paired observations, were largely attributed to observer errors (e.g., transcription errors, incorrectly resetting manual sensors, and delaying the observation time). The effects of these errors diminished as the daily observations were averaged spatially and temporally, with differences in monthly mean temperatures typically reduced to less than 1°C at the climate division level. Overall, Fiebrich and Crawford (2009) concluded that the transition to automated observations would change the climate record for Oklahoma on the daily scale.

Holder et al. (2006) compared daily temperature data for 13 pairs of collocated (i.e., stations were typically within 500 m of each other) COOP and North Carolina Environment and Climate Observing Network (NC ECONet) stations from August 2001 through January 2004. Although the authors improved the agreement between the daily temperature observations of paired stations by adjusting for differing times of observation, differences in the daily temperature observations still existed (i.e., median root-mean-square errors of ~1.6° and ~2.5°C for maximum and minimum temperatures, respectively) that the authors attributed to differences in instrumentation between the two networks. Holder et al. (2006) also found that monthly averaging of daily observations reduced the differences in the daily temperature observations (i.e., median root-mean-square errors of ~0.7° and ~0.6°C for maximum and minimum temperatures, respectively). Bruton et al. (1998) found very high correlations for daily temperature observations (i.e., r2 values of 0.994 and 0.974 for maximum and minimum temperatures, respectively) for three pairs of very closely collocated (i.e., <4 m) COOP and Georgia Automated Environmental Monitoring Network (AEMN) stations over a 2-yr period. However, average monthly maximum (minimum) temperatures for AEMN stations ranged from 2.2°C (1.6°C) warmer to 1.1°C (1.1°C) cooler than those of COOP stations. The authors noted that the differences appeared to be random and that no bias could be identified for the differences in monthly temperatures.

The limited research conducted has shown discrepancies between the temperature measurements from manual observations and state-based mesonets. Most of the research, however, has focused on comparing daily observations over limited periods (i.e., 1–3 years), often attributing the differences to observer errors or differences in the time of observations, but also to sensor biases and differences in siting characteristics. Some of these studies found that converting the daily observations to monthly averages reduced the temperature differences between the two observing systems, although the monthly averaging was done for the same relatively short study periods. As such, the overall goal of this research was to compare monthly and seasonal temperatures from an automated network to those of manual observing networks over a longer period to understand better their compatibility from a climatological perspective, and to explore potential reasons for temperature differences between these networks. The specific research objectives were to 1) compare monthly and seasonal maximum and minimum temperatures derived from the Georgia AEMN to those of geographically close (i.e., paired) manual observing stations for the period 2002–13, and 2) quantitatively evaluate the extent to which differences in siting characteristics (i.e., geographic location and land surface characteristics) of paired stations contribute to the temperature differences.

2. Data and methodology

This study used average monthly maximum and minimum temperature data from the Georgia AEMN, the USHCN, and the COOP network. The main goal of the AEMN is to collect detailed weather data for agricultural and environmental applications from stations that represent the unique climate and soil conditions across the state of Georgia (Hoogenboom et al. 2003; Hoogenboom 2005). The network has grown from four stations in 1991 to the present 81 stations, with nearly half of the stations being installed during or after 2002. Stations were installed according to standards for sensor accuracy, placement, and exposure established for the AWDN in Nebraska (Hubbard et al. 1983; Meyer and Hubbard 1992). For AEMN stations, air temperature is recorded with a combination temperature and humidity sensor that incorporates a thermistor (models MP100 and MP101A, Rotronics Corp.; model HMP35C, Vaisala Inc.) housed in unaspirated Gill multiplate radiation shields (model 41002, R. M. Young Co.), with an accuracy of ±0.28°C between −65° and +60°C (Bruton et al. 1998; Rothfusz et al. 2006). The thermistor is sampled every second and averaged and recorded every 15 min, and the highest and lowest 15-min readings from midnight to midnight (i.e., 24 h) are recorded as the maximum and minimum daily air temperatures, respectively (Bruton et al. 1998; Hoogenboom 2005).

The USHCN and COOP stations selected for this research were equipped with MMTS sensors to avoid temperature biases associated with conversion from the liquid-in-glass (LIG) thermometers previously used (e.g., Quayle et al. 1991). The MMTS uses a single thermistor (Dale/Vishay 1140, Vishay Intertechnology, Inc.) with an accuracy of ±0.45° between −50° and 50°C that is housed inside an unaspirated multiplate, cylindrical, plastic radiation shield about 25 cm high and about 20 cm in diameter (Hubbard et al. 2004; Fiebrich and Crawford 2009). The MMTS samples air temperature every 2 s, and the extreme temperatures represent the daily maximum and minimum temperatures of the 24 h prior to the time of observation (Holder et al. 2006; National Weather Service 2014). The time of observations has historically been morning or afternoon readings as opposed to midnight to midnight. During the latter part of the twentieth century, there has been a widespread shift from afternoon to morning observation times that has artificially reduced the true temperature trend in the U.S. temperature record (Baker 1975; Karl et al. 1986; Vose et al. 2003; Pielke et al. 2007; Menne et al. 2009).

The monthly averaged daily maximum and minimum temperature data of AEMN stations (available at http://www.weather.uga.edu/) were compared to those of the fully adjusted version 2.5 USHCN data (available at http://www.ncdc.noaa.gov/oa/climate/research/ushcn/) and the Global Historical Climatology Network–Daily (GHCND) monthly summaries COOP data (available at https://www.ncdc.noaa.gov/data-access/land-based-station-data/land-based-datasets/global-historical-climatology-network-ghcn/). The AEMN data are verified by quality control procedures to check for internal (e.g., data values out of range, missing data), temporal, and spatial inconsistencies (Bruton et al. 1998; Hoogenboom 2005). The fully adjusted USHCN data have undergone intensive quality control to account for duplicate data, climatological outliers, and inconsistencies (i.e., internal, spatial, and temporal) (Menne et al. 2009; Durre et al. 2010; Menne et al. 2014). This version of the USHCN data has also been adjusted for changes to the time of observation (i.e., to 24-h observations), station relocation, missing data, and changes in instrumentation to reduce nonclimatic inhomogeneities in the dataset (Menne et al. 2009, 2012, 2014). The COOP data have similarly undergone quality assurance checks to detect duplicate data, and climatological outliers and inconsistences, but they have not been homogenized, unlike USHCN data (Durre et al. 2010; Menne et al. 2012). As such, the fact that the USHCN data have been homogenized and that the AEMN and COOP data have not been homogenized (e.g., the different time of observations) will need to be considered when evaluating the temperature differences between the automated and manual networks.

The selection of the specific stations and the study period were based upon four criteria: a pair of AEMN and USHCN stations had to be located in close proximity to each other to minimize differences in air temperature between paired stations due to spatial variability in climate; each station had to have at least 95% of the monthly data or not have more than one of each season missing during the study period; there had to be a sufficient number of stations with which to conduct the analyses; and the USHCN and COOP stations had to be equipped with MMTS sensors. The selection criteria limited the number of viable pairings because there are only 23 USHCN stations in Georgia, with many of them not in close proximity to AEMN stations, and nearly half of the AEMN stations were installed during or after 2002. As a compromise to the limitations of the available stations and data, temperature data from a total of 13 pairs of nearly collocated (i.e., <25 km separating the paired stations) AEMN and USHCN stations were compared for the period 2002–13 (Fig. 1; Table A1). The AEMN stations were also paired with their closest available COOP stations to increase the number of collocated stations, allowing for a test of the robustness of the results with different sample sizes. Based on the same criteria used to select paired AEMN and USHCN stations, a total of 25 paired AEMN and COOP stations within 25 km of each other were identified, with 5 of those station pairs within 10 km of each other for the period 2002–13 (Fig. 2, Table B1). To address the issue of missing data and to ensure more accurate comparisons between networks, if any monthly or seasonal data were missing from one network, then the same monthly or seasonal data were removed from the other network.

Fig. 1.
Fig. 1.

The locations of AEMN (triangles) and USHCN (circles) stations. The names of corresponding paired USHCN stations can be found in Table 1.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

Fig. 2.
Fig. 2.

The locations of AEMN (triangles) and COOP (circles) stations. The names of corresponding paired COOP stations can be found in Table 5.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

In the first part of the analysis, average monthly maximum and minimum temperatures of AEMN stations were compared to average monthly maximum and minimum temperatures of their paired USHCN and COOP stations (i.e., 144 months of data for each station). The monthly data were also averaged by season (i.e., DJF, MAM, JJA, SON), and the average seasonal maximum and minimum temperatures of the AEMN stations were compared to those of their paired USHCN and COOP stations (i.e., 12 averaged maximum and 12 averaged minimum temperatures for each season and each station). To conduct the comparisons, the average differences in temperatures between paired stations—AEMN–USHCN and AEMN–COOP—were derived by calculating the differences (i.e., AEMN minus manual) between the maximum and minimum temperatures from the AEMN stations and the corresponding temperatures of their paired stations for both manual observing networks for each month and each season in a given year. The monthly and seasonal temperature differences between each paired station for each year were then averaged over the whole study period to produce the average monthly and seasonal maximum and minimum temperature differences for each paired station. The average absolute differences in monthly and seasonal temperatures of paired stations were calculated in the same manner, except that the absolute values of the temperature differences were averaged. The average monthly and seasonal maximum and minimum temperatures of AEMN stations were also compared to those of both USHCN and COOP stations with t test and Pearson correlation techniques (confidence level of α ≤ 0.05) to quantify any differences in the temperatures of the networks, thereby allowing for an assessment of the compatibility of the automated and manual networks.

The second part of the analysis relates the average seasonal differences in temperatures of paired AEMN and USHCN stations to their average differences in site characteristics (i.e., geographic location and land surface characteristics). The USHCN stations, as opposed to COOP stations, were used because the adjusted USHCN dataset does not contain missing monthly data and the stations experienced little or no station movement over the study period. This would allow for more robust findings that would not be biased by missing data values or station movement. The latitude, longitude, and elevation of USHCN stations were obtained from metadata provided by NCEI (http://www.ncdc.noaa.gov/homr/), while the latitude, longitude, and elevation of AEMN stations were obtained from the AEMN website (http://www.weather.uga.edu/). Further information on the geographic location and land surface characteristics for each station were obtained from a variety of sources: elevation, slope, and aspect were obtained or derived from the 10-m National Elevation Dataset (Gesch et al. 2009), land-use and land-cover (LULC) data from the Enhanced Historical Land-Use and Land-Cover Data Set (Price et al. 2006; http://water.usgs.gov/GIS/dsdl/ds240/index.html), and soil order from the State Soil Geographic (STATSGO) Data Base (Schwarz and Alexander 1995). Information on geographic location and land surface characteristics were collected at multiple spatial scales surrounding each station (i.e., scales from 1 m to 10 km) because climate-forcing factors that affect maximum and minimum temperature may occur at different spatial scales (Gallo et al. 1996; Daly 2006; Bodine et al. 2009; Allard and Carleton 2010). Accordingly, to quantitatively evaluate potential reasons for differences in seasonal maximum and minimum temperatures of paired AEMN and USHCN stations, Pearson correlation, and stepwise regression techniques (confidence level of α ≤ 0.05) were conducted for the differences in temperature and the differences in geographic location and land surface characteristics at multiple spatial scales.

3. Results and discussion

a. Comparisons of AEMN with USHCN and COOP maximum and minimum temperatures

The correlations of average monthly maximum and minimum temperatures of paired stations were all high and statistically significant (i.e., averaged r values of 0.9972 and 0.9959 for maximum and minimum temperatures, respectively), and the correlations appear to be unrelated to the distance between paired stations (Table 1), indicating that both networks captured the same month-to-month variability in temperature. The differences between the monthly maximum temperatures of paired AEMN and USHCN stations were not statistically significant, and also do not appear to be related to the distance between paired stations (Table 2). The monthly maximum temperatures of AEMN stations, however, did range from 0.66°C warmer to 0.95°C cooler than their paired USHCN stations, with an average difference of 0.07°C and an average absolute difference of 0.62°C for all paired stations. The differences of monthly minimum temperatures of paired AEMN and USHCN stations were not statistically significant, with the exception of one pairing (i.e., Valdosta–Quitman 2 NW) (Table 2). The monthly minimum temperatures of AEMN stations ranged from 2.19°C warmer to 0.68°C cooler than their paired USHCN stations, with an average difference of 0.53°C and an average absolute difference of 1.00°C for all paired stations. Given that the average absolute differences in monthly minimum temperatures were ~0.38°C greater than those of monthly maximum temperatures, the results indicated that there was a better agreement between monthly maximum temperatures of paired stations than those of monthly minimum temperatures. Unlike monthly maximum temperatures, there may be a relationship between the absolute differences in monthly minimum temperatures and the distance between paired stations, with larger differences occurring more often with distances between paired stations exceeding 20 km.

Table 1.

Correlation (r values) of average monthly minimum temperatures Tmin and average monthly maximum temperatures Tmax of paired AEMN–USHCN stations. Values set in boldface denote a statistically significant correlation at α ≤ 0.05.

Table 1.
Table 2.

Average difference (avg) and average absolute difference (avg abs) for average monthly Tmin and Tmax (°C) of paired stations (AEMN–USHCN). Bold denotes a statistically significant difference in the means (average differences) of paired stations at α ≤ 0.05.

Table 2.

The comparisons of seasonal maximum temperatures were similar to those of the monthly maximum temperatures. However, the comparisons of seasonal minimum temperatures revealed some findings not evident in comparisons of monthly minimum temperatures (Table 3). The correlations of seasonal maximum temperatures for paired stations were all high and statistically significant (i.e., averaged r values of 0.9812 for winter, 0.9740 for spring, 0.9643 for summer, and 0.9250 for fall), and they did not show a consistent relationship with the distance between paired stations. The correlations of seasonal minimum temperatures for paired stations were statistically significant for all but one seasonal pairing (i.e., Brunswick–Brunswick for summer), and they also did not show a consistent relationship with the distance between paired stations. There was, however, a seasonal dependence on the strength of the correlations of seasonal minimum temperatures: the averaged r values were 0.9571 for winter, 0.9312 for spring, 0.7849 for summer, and 0.8952 for fall. The smaller sample sizes (i.e., N = 144 for monthly correlations and N = 12 for seasonal correlations) would not necessarily impact the magnitude of the correlation coefficients, although the significance of the correlations could be impacted. Seasonal maximum temperatures were also subject to the same sample sizes as those of seasonal minimum temperatures and were not substantially lower in summer, further supporting that sample size does not explain the lower correlations for summer minimum temperatures. Rather, these lower correlations for minimum temperatures for the warm season were more likely related to differences in siting characteristics (e.g., differences in land use/land cover) between the paired stations.

Table 3.

Correlation of average seasonal Tmin and Tmax of paired AEMN–USHCN stations. Bold denotes a statistically significant correlation at α ≤ 0.05.

Table 3.

The seasonal maximum temperatures of paired stations were not significantly different from each other in all but two instances (i.e., Brunswick–Brunswick for summer and Attapulgus–Bainbridge for fall) (Table 4). Again, the differences in seasonal maximum temperatures did not show a consistent relationship with the distance between paired stations, although it is of note that one of the two significant differences was for the paired stations closest to each other (i.e., Brunswick–Brunswick). Although the differences in seasonal maximum temperatures between paired stations were almost entirely not statistically significant, differences did exist between the seasonal maximum temperatures of paired stations. The average differences (absolute differences) between seasonal maximum temperatures were 0.08°C (0.51°C) for winter, 0.16°C (0.61°C) for spring, 0.06°C (0.57°C) for summer, and 0.02°C (0.56°C) for fall. These differences in temperatures were fairly uniform across seasons and therefore suggest that there was not a strong seasonal dependency on differences in maximum temperatures of paired stations.

Table 4.

Average difference (avg) and average absolute difference (avg abs) of paired stations (AEMN–USHCN) for average seasonal Tmin and Tmax (°C). Bold denotes a statistically significant difference in the means (average differences) of paired stations at α ≤ 0.05.

Table 4.

In contrast, the seasonal minimum temperatures of paired stations were significantly different for half the comparisons (i.e., 26 out of 52 possible seasonal comparisons), with significant differences between paired stations being more common for summer and fall than for winter and spring (Table 4). Again, these results indicate a seasonal dependence on differences in minimum temperatures, underscoring the importance of comparing seasonal temperatures of paired stations. If the differences had not been examined by season, then these significant seasonal differences, at least for minimum temperature, would not have been identified, as was the case in some previous studies (e.g., Bruton et al. 1998; Holder et al. 2006; Fiebrich and Crawford 2009). The average differences (absolute differences) between seasonal minimum temperatures of paired stations were 0.53°C (0.93°C) for winter, 0.67°C (0.98°C) for spring, 0.36°C (0.91°C) for summer, and 0.54°C (1.01°C) for fall. Moreover, the distance between paired stations affected the magnitude of the absolute differences in seasonal minimum temperatures, with the magnitudes generally increasing as the distance between paired stations increased. The average absolute differences for the seven paired stations closest to each other were 0.75°C for winter, 0.83°C for spring, 0.72°C for summer, and 0.83°C for fall; the average absolute differences for the six paired stations farthest from each other were 1.13°C for winter, 1.16°C for spring, 1.13°C for summer, and 1.22°C for fall. In addition, the average absolute differences in seasonal minimum temperatures for paired stations were ~0.43°C greater than those of seasonal maximum temperatures, again indicating a better agreement between maximum temperatures for paired stations than minimum temperatures.

The results of the comparisons of monthly air temperature data between paired AEMN and COOP stations were broadly similar to those of paired AEMN and USHCN stations. Correlations of monthly maximum and minimum temperatures of paired AEMN and COOP stations were high and all were statistically significant (i.e., the averaged r values of 0.9957 and 0.9942 for maximum and minimum temperatures, respectively) (Table 5). The average difference in temperatures between paired AEMN and COOP stations was 0.18°C for monthly maximum temperatures and 0.42°C for monthly minimum temperatures, and most of these differences were not significant statistically (Table 6). The average absolute differences were 0.68°C for monthly maximum temperatures and 0.98°C for monthly minimum temperatures. As with paired AEMN and USHCN stations, the absolute differences in monthly maximum temperatures were less than those of monthly minimum temperatures. There was also no statistically significant relationship for the distance between paired AEMN and COOP stations and average differences in monthly maximum and minimum temperatures or their correlations. However, there was a slight tendency for the magnitude of their average absolute differences to increase and the strength of their correlations to decrease as the distance between paired stations increased.

Table 5.

Correlation of average monthly Tmin and Tmax of paired AEMN–COOP stations. Bold denotes a statistically significant correlation at α ≤ 0.05.

Table 5.
Table 6.

Average difference (avg) and average absolute difference (avg abs) for average monthly Tmin and Tmax (°C) of paired stations (AEMN–COOP). Bold denotes a statistically significant difference in the means (average differences) of paired stations at α ≤ 0.05.

Table 6.

The comparisons of seasonal temperatures between paired AEMN and COOP stations were also generally similar to those of paired AEMN and USHCN stations. Seasonal maximum temperatures of paired AEMN and COOP stations were significantly correlated and high for all seasons (i.e., average r values are 0.9554 for winter, 0.9467 for spring, 0.9165 for summer, and 0.8971 for fall) (Table 7). The correlations of seasonal minimum temperatures of paired AEMN and COOP stations were statistically significant for all seasons, with the exception of five pairings for summer (Table 7), and were also lower than those of seasonal maximum temperatures, particularly for summer (i.e., average r values were 0.8924 for winter, 0.9092 for spring, 0.7502 for summer, and 0.8843 for fall). The average differences (absolute differences) between seasonal maximum temperatures were 0.18°C (0.57°C) for winter, 0.18°C (0.61°C) for spring, 0.17°C (0.70°C) for summer, and 0.17°C (0.56°C) for fall (Table 8). The average differences (absolute differences) between seasonal minimum temperatures were 0.58°C (1.03°C) for winter, 0.69°C (1.06°C) for spring, 0.07°C (0.71°C) for summer, and 0.32°C (0.90°C) for fall (Table 8). Comparable to those of paired AEMN and USHCN stations, the average absolute differences in seasonal minimum temperatures for paired stations were larger than those of maximum temperatures, and the differences in seasonal maximum temperatures and minimum temperatures of paired AEMN and COOP stations were statistically significant for ~7% and ~30% of the pairings, respectively. There did not appear, however, to be a robust relationship for the distance between paired stations and the magnitudes of the differences in seasonal maximum or minimum temperatures or their correlations when all paired stations were considered.

Table 7.

Correlation of average seasonal Tmin and Tmax of paired AEMN–COOP stations. Bold denotes a statistically significant correlation at α ≤ 0.05.

Table 7.
Table 8.

Average difference (avg) and average absolute difference (avg abs) of paired stations (AEMN–COOP) for average seasonal Tmin (°C) and Tmax (°C). Bold denotes a statistically significant difference in the means (average differences) of paired stations at α ≤ 0.05.

Table 8.

The similarity of the differences in temperatures and the correlations of paired AEMN and COOP to those of AEMN and USHCN pairings point toward the robustness of the results, but they also suggests that the homogenization of only the USHCN data had no obvious impact. With regard to the AEMN, no notable changes occurred to the stations over the study period (e.g., no station movement or instrumentation changes, constant 24-h observation times, and a very small percentage of missing data), which would introduce systemic biases with only one of the manual observation networks and not the other. Both the USHCN and COOP stations were equipped with MMTS sensors, so no bias would be expected between those two networks with respect to the AEMN due to differing sensors, as long as the sensors were functioning properly. Many of the COOP stations did experience some station movement and all but one did not have midnight-to-midnight observations times (see Table B1). While both of these have been shown to introduce temperature biases in other studies, there is no clear indication that not adjusting for these considerations affected the results. The lack of apparent differences between the USHCN and COOP stations with respect to the AEMN stations may have been related to station movements being too minimal to influence temperature trends, the fact that only four stations did not have constant observation times (i.e., Midville, Helen, and Colquitt changed from afternoon to morning observations, and Homerville changed from 0800 to 0600 local time observations), or that the time of observation biases were masked by longer averaging periods. The absence of clear differences related to homogenization may not necessarily prove that one did not exist, but rather that other factors were more dominant and prevented one from being detectable.

The differences in monthly and seasonal temperatures of paired automated and manual (both USHCN and COOP) network types also did not clearly indicate that a sensor bias existed between the two network types. As seen in Fig. 3 (AEMN–USHCN) and Fig. 4 (AEMN–COOP), the seasonal temperatures of AEMN stations were not consistently warmer or cooler than those of their paired USHCN or COOP stations, as would be expected for a systemic sensor bias between AEMN sensors and MMTS sensors. Therefore, if a bias between the sensor types existed, then it was also apparently obscured by other factors. These two figures do, however, graphically show the results presented earlier that seasonal maximum temperatures of paired stations are more in agreement than seasonal minimum temperatures. This may be because maximum temperatures typically occur during the daytime, when the land surface is coupled with a more commonly well-mixed boundary layer, and microclimate differences between closely located stations should be less evident (Christy et al. 2009; Menne et al. 2009; Misra et al. 2012). In contrast, minimum temperatures, which generally occur near sunrise, are measured when a shallow nocturnal boundary layer is decoupled from the rest of the atmosphere and is more sensitive to local conditions (e.g., land use/land cover) (Menne et al. 2009; Misra et al. 2012). Furthermore, variations in the amount of transpiring vegetation during the growing season could more strongly impact minimum temperatures, through alterations of the surface energy budget during the warm season, than during other parts of the year (Hanamean et al. 2003; Pielke et al. 2007), and may have resulted in the lower correlations and increased incidence of significant differences in the seasonal minimum temperatures between paired stations during the warm season. It is also possible that temperature differences between paired stations are related to meteorological differences, particularly the isolated nature of convective precipitation in the summer and subsequent impacts on soil moisture and associated radiation and energy budgets, and not necessarily siting characteristics. These differences, however, would not be limited to different network types but rather are a complication when comparing stations regardless of their network. Examinations of the differences between the siting characteristics of paired stations could clarify these assertions, as conducted in the next section.

Fig. 3.
Fig. 3.

Scatterplots showing the average seasonal temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs distance (km) between the same paired stations.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

Fig. 4.
Fig. 4.

Scatterplots showing the average seasonal temperature difference (°C) between paired AEMN–COOP stations (AEMN minus COOP) vs distance (km) between the same paired stations.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

b. Comparisons of differences in seasonal temperatures and site characteristics of paired AEMN and USHCN stations

The results from this study have shown that there were differences, whether significant or not, in the average seasonal temperatures of paired AEMN and USHCN stations, (i.e., average absolute differences of ~0.56° and ~0.96°C for seasonal maximum and minimum temperatures), and there was no clear evidence that these differences were due to sensor biases or only the USHCN data being homogenized. The differences in seasonal maximum temperatures of paired AEMN and USHCN stations also did not appear to be related to the distance between paired stations. However, the average absolute difference in seasonal minimum temperatures for the seven closest paired stations were less than those for the six farthest paired stations, suggesting that distance between paired stations may influence the magnitude of the differences in seasonal minimum temperatures. To examine the possibility that an increase in distance between paired stations could result in the paired stations being located in areas that differ climatologically and, thereby, increase the differences in temperatures, correlations were determined between the average differences in seasonal maximum and minimum temperatures and distances between those same paired stations. The correlations were consistently low, but differences for summer minimum temperatures of paired stations were significant at α ≤ 0.05 (r = 0.5771). The differences in minimum temperatures as a function of distance (Fig. 3), however, suggest that the statistical significance may not be meaningful. The summer minimum temperature differences between 20 and 24 km were similar to some of the differences between 5 and 15 km, and the differences between 0 and 5 km were similar to some of those between 10 and 20 km. It is likely that the correlation only achieved statistical significance because of the difference between a single pairing at about 25 km. As such, there is no strong evidence that the distance between paired AEMN and USHCN stations increased the magnitude of the differences in seasonal temperatures.

The differences in latitude and longitude of paired stations were also correlated to the differences in seasonal temperatures between paired AEMN and USHCN stations to evaluate further the possibility that temperature differences may be related to differences in geographic location. In these cases, the correlations were uniformly low and not statistically significant (α ≤ 0.05), suggesting that other factors besides differences in latitude and longitude between paired stations were responsible for most of the differences in seasonal maximum and minimum temperatures of paired stations. These findings are supported by scatterplots that show no clear relationship between differences in seasonal temperatures and differences in latitude and longitude (Figs. 5 and 6). In combination with the results relating differences in temperatures to differences in distance, the differences in seasonal maximum temperature between paired AEMN and USHCN stations did not appear to be related to differences in the geographic coordinates of the paired stations. For seasonal minimum temperatures, it is likely that an increase in distance between paired stations increased the likelihood that the paired stations had different siting characteristics, and that a reasonable distance (e.g., <25 km) between paired stations itself was not responsible for the differences in seasonal minimum temperatures of those paired stations.

Fig. 5.
Fig. 5.

Scatterplots showing the average seasonal minimum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in latitude (decimal degrees) and longitude (decimal degrees) between the same paired stations.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

Fig. 6.
Fig. 6.

Scatterplots showing the average seasonal maximum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in latitude (decimal degrees) and longitude (decimal degrees) between the same paired stations.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

Seasonal differences in the temperatures of paired stations were also compared to other differences in the geographic locations of paired stations (i.e., elevation, slope, and aspect) at multiple spatial scales (e.g., average elevations around individual stations for areas with radii that range from 1 m to 10 km). For seasonal minimum temperatures, differences in the elevation, slope, and aspect of paired stations did not show significant correlations (α ≤ 0.05) with the differences in seasonal minimum temperatures of the paired stations. Likewise, differences in seasonal maximum temperatures did not show consistent significant correlations with the differences in slope or aspect of paired stations. The lack of significant correlations can also be confirmed by visual interpretations of scatterplots depicting differences in seasonal temperatures versus differences in elevation, slope, and aspect (Figs. 7 and 8). The differences in seasonal maximum temperatures did, however, show consistently significant negative correlations with elevation (Table 9; Fig. 8), with stations located at higher elevations than their paired stations being correlated with lower maximum temperatures. The strength of the correlations varied by season and spatial scale considered, with the highest correlations occurring during the winter and at spatial scales of 1–2 km surrounding the stations. The highest correlations were likely found during the winter because other factors that could influence the differences in seasonal maximum temperatures (e.g., LULC) would be reduced, and the effect on elevation on maximum temperatures would have been most obvious during that time of the year.

Fig. 7.
Fig. 7.

Scatterplots showing the average seasonal minimum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in elevation (m), slope (°), and aspect [the compass direction of the topographic slope (°)] between the same paired stations for the spatial scales in Table A1.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

Fig. 8.
Fig. 8.

Scatterplots showing the average seasonal maximum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in elevation (m), slope (°), and aspect [the compass direction of the topographic slope (°)] between the same paired stations for the spatial scales in Table A1.

Citation: Journal of Atmospheric and Oceanic Technology 33, 7; 10.1175/JTECH-D-15-0161.1

Table 9.

Correlation of differences in average seasonal Tmax and differences in elevation of paired AEMN–USHCN stations at multiple spatial scales. Bold denotes a statistically significant correlation at α ≤ 0.05.

Table 9.

The differences in seasonal temperatures of paired stations were also compared with the differences in land surface characteristics (i.e., LULC and soil order) of paired stations at multiple spatial scales (i.e., 1 m to 10 km). No significant correlations (α ≤ 0.05) were found between differences in seasonal maximum or minimum temperatures and differences in soil order of paired stations. A limited number of consistent and significant correlations (α ≤ 0.05) were found between differences in LULC and the differences of seasonal temperatures (data not shown). For seasonal maximum temperatures, differences in winter maximum temperatures and differences in the percentage of land classified as agriculture were significantly negatively correlated at the spatial scales of 2 and 5 km (i.e., r = −0.5700 and −0.6432, respectively). This indicates that stations surrounded by more exposed agricultural areas during the winter were correlated with lower maximum temperatures. For seasonal minimum temperatures, differences in spring, summer, and fall minimum temperatures were significantly positively correlated with wetlands at the spatial scales of 500 m and 1 km (e.g., the highest r values were 0.6009 at 500 m, 0.5976 at 1 km, and 0.5497 at 500 m for spring, summer, and fall, respectively). The higher heat capacity of water in wetland areas likely reduced the cooling of minimum temperatures for stations located near a higher percentage of wetlands during these seasons. Winter and fall minimum temperatures also possessed significant positive correlations with urban areas at the spatial scales of 5 km and 10 km (e.g., the highest r values were 0.7289 at 10 km and 0.6217 at 10 km for winter and fall, respectively). It is possible that greater urbanization in areas surrounding some stations increased the nighttime minimum temperatures for these two seasons compared to those stations with less urbanization in the surrounding areas.

To determine the extent to which the averaged differences in seasonal temperatures can be explained by their differences in siting characteristics, stepwise regressions were conducted on the average differences in seasonal temperatures and differences in the site characteristics with which seasonal temperature differences were significantly correlated (Table 10). The results indicated that a large amount of the variance between seasonal maximum temperatures of paired stations could be explained by differences in their elevation, although differences in agricultural land use also contributed to explaining temperature differences for the winter. The amount of explained variance ranged from ~38% for spring to ~81% for winter. In contrast, the variance between seasonal minimum temperatures of paired stations was explained by differences in their LULC, and the amount of explained variance was lower. The amount of explained variance ranged from ~30% for summer to ~73% for fall. Overall, the differences between seasonal maximum temperatures of paired stations were less than those for seasonal minimum temperatures, and a greater proportion of the variance for seasonal maximum temperature differences could be explained.

Table 10.

Results of stepwise regression where differences in site characteristics are used to explain variance between paired AEMN–USHCN stations for average seasonal Tmin and Tmax. Bold denotes statistical significance at α ≤ 0.05.

Table 10.

The results presented earlier, however, suggested that the differences in seasonal minimum temperatures of paired stations were likely related to local conditions, but the differences in siting characteristics of paired stations were found to explain relatively little of the variance in seasonal minimum temperatures of paired stations. It is possible that stronger relationships between differences in seasonal minimum temperatures and differences in siting characteristics were not found because the datasets on the siting characteristics of the stations do not adequately capture the local conditions around stations. For example, there could be subtle differences in LULC of paired stations (e.g., more trees in the vicinity of one station), but both stations could be classified as being forested areas. These subtle differences in LULC could impact the microclimates around stations, and therefore their seasonal minimum temperatures, but not be identified in the LULC dataset. Similarly, changes in LULC might have occurred over time at one station but not its paired station. A static representation of LULC around stations would not capture changes in LULC and potential changes in their microclimates. Information on soil moisture was also not available for USHCN stations, and therefore a variable that can impact radiation and energy budgets, and subsequently differences in maximum and minimum temperatures, was not available for the analyses. In addition, this study was not able to consider local-to-mesoscale air circulations around stations that could impact their temperatures because USHCN stations do not record wind speed or direction. As such, these findings underscore the importance of detailed metadata for the site characteristics of stations, so as to understand better how local and larger-scale conditions around a station impact their seasonal maximum and minimum temperatures.

4. Summary and conclusions

This study compared air temperature data from collocated stations within automated and manual networks to assess the comparability of these different networks. The results showed that the correlations for both average monthly and seasonal maximum and minimum temperatures of paired AEMN and USHCN stations were high and almost always statistically significant, and did not possess a consistent relationship with distance between paired stations. The correlations for seasonal minimum temperatures, however, were slightly lower, particularly for summer. The differences between monthly and seasonal maximum temperatures of paired AEMN and USHCN stations were only significantly different in a few instances, and were unrelated to distance between paired stations. The differences between monthly minimum temperatures were only significantly different for one pairing, while the differences between seasonal minimum temperatures were significantly different half the time, with those significant differences being more common for summer and fall. The average absolute differences in monthly and seasonal minimum temperatures of paired AEMN and USHCN stations were greater than those of maximum temperatures. The magnitude of these average absolute differences of monthly and seasonal minimum temperatures also tended to increase as the distance between paired stations increased. The results of the comparisons of monthly and seasonal temperatures for the larger sampling of paired AEMN and COOP stations are broadly similar to those of the paired AEMN and USHCN stations, demonstrating the robustness of the results found in this research.

The comparisons of both monthly and seasonal temperatures indicated that there was a stronger agreement between paired stations for maximum temperatures than for minimum temperatures. This is reasonable given that maximum temperatures occur during the daytime when a well-mixed boundary layer would reduce microclimate differences for closely located stations, while minimum temperatures are typically experienced when a shallow nocturnal boundary layer is decoupled from the rest of the atmosphere and is more sensitive to local conditions (e.g., LULC), similar to the results reported by Leeper et al. (2015) and Yao and Zong (2009). The lower correlations and increased incidences of significant differences for summer minimum temperatures, when variations in the amount of transpiring vegetation would more strongly impact minimum temperatures, seems to have confirmed this assertion. In addition, increases in the magnitudes of the absolute differences in monthly and seasonal minimum temperatures with an increase in distance between paired stations, and not maximum temperatures as well, suggest that siting characteristics are important in explaining the differences. If an increase in distance between paired stations (i.e., paired stations being in climatologically dissimilar locations) was primarily responsible for the differences in temperatures between paired stations, then differences in both maximum and minimum temperatures should be more evenly impacted. Rather, an increase in distance between paired stations likely increases the probability that paired stations are situated in locations with different siting characteristics.

The comparison of differences in seasonal temperatures and differences in siting characteristics of paired AEMN and USHCN stations partially confirmed the asserted role of siting characteristics. Differences in latitude, longitude, and distance between paired stations were generally not significantly correlated with differences in their temperatures, supporting the assertion that distance between paired stations was not the main contributor to temperature differences. Although differences in elevation were not significantly correlated with differences in minimum temperatures, about a third to just over a half of the variance between seasonal maximum temperatures of paired stations was explained by differences in elevation. Moreover, differences in slope, aspect, and soil order were not significantly correlated with temperature differences. Land use/land cover explained a portion of the variance between winter maximum temperatures and minimum temperatures in all seasons between paired stations, although a majority of the variance between temperatures was typically not explained by LULC. It is likely that the LULC classification used in this study did not adequately capture the more subtle differences in LULC around stations or how LULC around stations changed over time. Some of the variance between temperatures may also have been explained by differences in soil moisture, local-to-mesoscale air circulations, and meteorological conditions such as cloud cover, but this information was unfortunately not available for USHCN stations. This underscores the need for continued and improved collection of metadata of the siting characteristics around stations for all types of observing networks. Continuous site visits, in conjunction with satellite observations, would improve the metadata on siting characteristics, which in turn would allow for a better understanding of the comparability of observing networks and even climate trend analyses for individual stations.

Even with the high correlations and the lack of significant differences between paired stations for many of the stations examined in the study, differences in monthly and seasonal temperature still existed [e.g., average (absolute) differences of ~0.16°C (~0.56°C) and ~0.53°C (~0.96°C) between paired AEMN–USHCN stations for seasonal maximum and minimum temperature, respectively]. Whether significant or not, these differences still represent important differences in the temperature, particularly given that global mean temperature has warmed by 0.85°C (0.65°–1.06°C) between 1880 and 2012 (IPCC 2013). The differences in monthly and seasonal temperature, though, are also similar in magnitude to documented (e.g., station relocations) and undocumented (i.e., not recorded in station metadata) shifts in USHCN mean monthly maximum and minimum temperature series (Menne et al. 2009, their Fig. 6). An abrupt change from manual observing networks to the AEMN without adjustments to the data would, therefore, change the climate record for Georgia on monthly and seasonal time scales. In contrast, AEMN stations may be a viable means of replacing some manual observing stations if certain criteria are met: the transition from USHCN to AEMN stations is gradual; temperature differences are systematic; and there is spatial and temporal redundancy in the observing networks without all stations undergoing changes at the same time (e.g., instruments or station movement). This would also require proper station siting and well-documented metadata for a station, its instruments, and siting characteristics to be most effective. Even then, the creation of a homogeneous dataset from the two types of networks for climatological purposes would be a challenging task due to differences in siting characteristics (e.g., microclimates). The results of this study, therefore, point to the need for continued maintenance of both types of networks: the manual observing network to provide the necessary data for long-term studies of climate change, the automated network to meet the finer spatial and temporal resolution requirements that applications of weather and climate data now demand, and both networks to provide a sufficient period of redundancy between networks. Moreover, these results are informative for other statewide mesonets because the study area spans differing terrains and climatic regions, the AEMN is equipped and configured similarly to other mesonets [e.g., the AWDN in Nebraska, the Florida Automated Weather Network (FAWN) in Florida, and Agricultural Weather Network (AgWeatherNet) in Washington], and automated and manual observing networks in other states are likely to encounter similar differences in siting characteristics.

APPENDIX A

Geographic Coordinates, Siting Characteristics, and COOP Identification Number for Observing Stations

Table A1.

Table A1.

The latitude (decimal degrees), longitude (decimal degrees), elevation (m), slope (°), aspect [the compass direction of the topographic slope (°)], COOP identification (COOP ID) number, LULC, and topography of AEMN stations and USHCN stations. The elevation, slope, aspect, and LULC of stations included in the table are for the spatial scales with the consistently highest correlations between seasonal temperature and siting characteristics of paired stations (see section 3b for a discussion of the significance of these correlations). The numbers assigned to the LULC of stations correspond to the Anderson level I classification, and are listed from the highest to lowest percentage (i.e., left to right) of the three LULC classifications with the largest percentages of area within a circle with a radius of those spatial scales. The Anderson level I classifications are 11—residential; 12—commercial; 13—industrial; 17—other urban; 21—crops and pastures; 22—orchards, vineyards, nurseries; 41—deciduous forests; 42—evergreen forests; 43—mixed forests; 51—streams and canals; 53—reservoirs; 61—forested wetlands; 62—nonforested wetlands; 75—strip mines and quarries. Topography information for USHCN stations was obtained from the Historical Observing Metadata Repository of the National Centers for Environmental Information (http://www.ncdc.noaa.gov/homr/); topography information for AEMN stations was obtained by visual inspection of photographs of AEMN sites (available at http://www.weather.uga.edu/).

Table A1.

APPENDIX B

Geographic Coordinates, Elevation, Time of Observation, and COOP Identification Number for Observing Stations

Table B1.

Table B1.

The latitude (decimal degrees), longitude (decimal degrees), elevation (m), COOP ID number, and time of observation (24-h format, local time) for AEMN stations and COOP stations. Two times of observation for a COOP station indicate that the observation time changed from the first to the second time during the study period. The COOP stations are paired with their corresponding AEMN stations.

Table B1.

REFERENCES

  • Allard, J., and Carleton A. M. , 2010: Mesoscale interactions between Midwest land surface properties and convective cloud development in the warm season. Phys. Geogr., 31, 107136, doi:10.2747/0272-3646.31.2.107.

    • Search Google Scholar
    • Export Citation
  • Baker, D. G., 1975: Effect of observation time on mean temperature estimation. J. Appl. Meteor., 14, 471476, doi:10.1175/1520-0450(1975)014<0471:EOOTOM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bodine, D., Klein P. M. , Arms S. C. , and Shapiro A. , 2009: Variability of surface air temperature over gently sloping terrain. J. Appl. Meteor. Climatol., 48, 11171141, doi:10.1175/2009JAMC1933.1.

    • Search Google Scholar
    • Export Citation
  • Brown, Z. G., Foster S. A. , Mahmood R. , Strudbig S. , Grabowski D. , Grogan M. , and Ferris M. M. , 2008: An overview of the Kentucky mesonet. 17th Conf. on Applied Climatology, Whistler, BC, Canada, Amer. Meteor. Soc., 1.3. [Available online at https://ams.confex.com/ams/13MontMet17AP/techprogram/paper_140983.htm.]

  • Brusberg, M. D., and Hubbard K. G. , 2001: Preliminary findings of the 1999 survey of (nonfederal) automated weather stations in the United States and Canada. Automated weather stations for applications in agriculture and water resources management: Current use and future perspectives; Proceedings of an international workshop, K. G. Hubbard and M. V. K. Sivakumar, Eds., WMO Tech. Doc. WMO/TD-1074, 229–240.

  • Bruton, J. M., Hoogenboom G. , and McClendon R. W. , 1998: Comparison of automatically and manually collected weather data. Preprints, 1998 ASAE Annual Int. Meeting, Orlando, FL, American Society of Agricultural Engineers, Paper 98-2188.

  • Changnon, S. A., and Kunkel K. E. , 1999: Rapidly expanding uses of climate data and information in agriculture and water resources: Causes and characteristics of new applications. Bull. Amer. Meteor. Soc., 80, 821830, doi:10.1175/1520-0477(1999)080<0821:REUOCD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Christy, J. R., Norris W. B. , and McNider R. T. , 2009: Surface temperature variations in East Africa and possible causes. J. Climate, 22, 33423356, doi:10.1175/2008JCLI2726.1.

    • Search Google Scholar
    • Export Citation
  • Daly, C., 2006: Guidelines for assessing the suitability of spatial climate data sets. Int. J. Climatol., 26, 707721, doi:10.1002/joc.1322.

    • Search Google Scholar
    • Export Citation
  • Durre, I., Menne M. J. , Gleason G. E. , Houston T. G. , and Vose R. S. , 2010: Comprehensive automated quality assurance of daily surface observations. J. Appl. Meteor. Climatol., 49, 16151633, doi:10.1175/2010JAMC2375.1.

    • Search Google Scholar
    • Export Citation
  • Easterling, D. R., Karl T. R. , Mason E. H. , Hughes P. Y. , and Bowman D. P. , 1996: United States Historical Climatology Network (U.S. HCN) monthly temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, ORNL/CDIAC-87, NDP-019/R3, 280 pp.

  • Fiebrich, C. A., 2009: History of surface weather observations in the United States. Earth-Sci. Rev., 93, 7784, doi:10.1016/j.earscirev.2009.01.001.

    • Search Google Scholar
    • Export Citation
  • Fiebrich, C. A., and Crawford K. C. , 2009: Automation: A step toward improving the quality of daily temperature data by climate observing networks. J. Atmos. Oceanic Technol., 26, 12461260, doi:10.1175/2009JTECHA1241.1.

    • Search Google Scholar
    • Export Citation
  • Gallo, K. P., Easterling D. R. , and Peterson T. C. , 1996: The influence of land use/land cover on climatological values of the diurnal temperature range. J. Climate, 9, 29412944, doi:10.1175/1520-0442(1996)009<2941:TIOLUC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gesch, D., Evans G. , Mauck J. , Hutchinson J. , and Carswell W. J. Jr., 2009: The national map—Elevation. U.S. Geological Survey Fact Sheet 2009-3053, 4 pp., accessed 24 February 2014. [Available online at http://viewer.nationalmap.gov/viewer/.]

  • Guttman, N. B., and Baker C. B. , 1996: Exploratory analysis of difference between temperature observations recorded by ASOS and conventional methods. Bull. Amer. Meteor. Soc., 77, 28652873, doi:10.1175/1520-0477(1996)077<2865:EAOTDB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hale, R. C., Gallo K. P. , and Loveland T. R. , 2008: Influences of land use/land cover conversions on climatological normals of near-surface temperature. J. Geophys. Res., 113, D14113, doi:10.1029/2007JD009548.

    • Search Google Scholar
    • Export Citation
  • Hanamean, J. R., Jr., Pielke R. A. Sr., Castro C. L. , Ojima D. S. , Reed B. C. , and Gao Z. , 2003: Vegetation greenness impacts on maximum and minimum temperatures in northeast Colorado. Meteor. Appl., 10, 203215, doi:10.1017/S1350482703003013.

    • Search Google Scholar
    • Export Citation
  • Holder, C., Boyles R. , Syed A. , Niyogi D. , and Raman S. , 2006: Comparison of collocated automated (NC ECONet) and manual (COOP) climate observations in North Carolina. J. Atmos. Oceanic Technol., 23, 671682, doi:10.1175/JTECH1873.1.

    • Search Google Scholar
    • Export Citation
  • Hoogenboom, G., 2005: The Georgia Automated Environmental Monitoring Network: Experiences with the development of a state-wide automated weather station network for Georgia. 15th Conf. on Applied Climatology, Savannah, GA, Amer. Meteor. Soc., JP1.16. [Available online at https://ams.confex.com/ams/15AppClimate/webprogram/Paper94108.html.]

  • Hoogenboom, G., Coker D. D. , Edenfield J. M. , Evans D. M. , and Fang C. , 2003: The Georgia Automated Environmental Monitoring Network: Ten years of weather information for water resources management. Proceedings of the 2003 Georgia Water Resources Conference, K. J. Hatcher, Ed., The University of Georgia, 896900.

  • Horel, J., and Coauthors, 2002: Mesowest: Cooperature mesonets in the western United States. Bull. Amer. Meteor. Soc., 83, 211226, doi:10.1175/1520-0477(2002)083<0211:MCMITW>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hubbard, K. G., 2001: The Nebraska and High Plains regional experience with automated weather stations. Automated weather stations for applications in agriculture and water resources management: Current use and future perspectives; Proceedings of an international workshop, K. G. Hubbard and M. V. K. Sivakumar, Eds., WMO Tech. Doc. WMO/TD-1074, 219–227.

  • Hubbard, K. G., Rosenberg N. J. , and Nielsen D. C. , 1983: Automated weather data network for agriculture. J. Water Resour. Plann. Manage., 109, 213222, doi:10.1061/(ASCE)0733-9496(1983)109:3(213).

    • Search Google Scholar
    • Export Citation
  • Hubbard, K. G., Lin X. , Baker C. B. , and Sun B. , 2004: Air temperature comparison between the MMTS and the USCRN temperature systems. J. Atmos. Oceanic Technol., 21, 15901597, doi:10.1175/1520-0426(2004)021<1590:ATCBTM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hughes, P. Y., Mason E. H. , Karl T. R. , and Brower W. A. , 1992: United States Historical Climatology Network daily temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, CDIAC-50, NDP-042, 140 pp.

  • IPCC, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp., doi:10.1017/CBO9781107415324.

  • Karl, T. R., and Williams C. N. Jr., 1987: An approach to adjusting climatological time series for discontinuous inhomogeneities. J. Climate Appl. Meteor., 26, 17441763, doi:10.1175/1520-0450(1987)026<1744:AATACT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., Williams C. N. Jr., Young P. J. , and Wendland W. M. , 1986: A model to estimate the time of observation bias associated with monthly mean maximum, minimum, and mean temperature for the United States. J. Climate Appl. Meteor., 25, 145160, doi:10.1175/1520-0450(1986)025<0145:AMTETT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., Williams C. N. Jr., and Quinlan F. T. , 1990: United States Historical Climatology Network (HCN) serial temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, ORNL/CDIAC-30, NDP-019/R1, 83 pp.

  • Kimball, S. K., Mulekar M. S. , Cummings S. , and Stamates J. , 2010: The University of South Alabama Mesonet and coastal observing system: A technical and statistical overview. J. Atmos. Oceanic Technol., 27, 14171439, doi:10.1175/2010JTECHA1376.1.

    • Search Google Scholar
    • Export Citation
  • Leeper, R. D., Rennie J. , and Palecki M. A. , 2015: Observational perspectives from U.S. Climate Reference Network (USCRN) and Cooperative Observer Program (COOP) network: Temperature and precipitation comparison. J. Atmos. Oceanic Technol., 32, 703721, doi:10.1175/JTECH-D-14-00172.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., and Williams C. N. Jr., 2009: Homogenization of temperature series via pairwise comparisons. J. Climate, 22, 17001717, doi:10.1175/2008JCLI2263.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Williams C. N. Jr., and Vose R. S. , 2009: The United States Historical Climatology Network monthly temperature data, version 2. Bull. Amer. Meteor. Soc., 90, 9931007, doi:10.1175/2008BAMS2613.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Williams C. N. Jr., and Palecki M. A. , 2010: On the reliability of the U.S. surface temperature record. J. Geophys. Res., 115, D11108, doi:10.1029/2009JD013094.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Durre I. , Vose R. S. , Gleason B. E. , and Houston T. G. , 2012: An overview of the Global Historical Climatology Network-Daily database. J. Atmos. Oceanic Technol., 29, 897910, doi:10.1175/JTECH-D-11-00103.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Williams C. N. Jr., and Vose R. S. , 2014: United States Historical Climatology Network (USHCN) version 2.5 serial monthly dataset. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, accessed 4 March 2014. [Available online at http://cdiac.ornl.gov/ftp/ushcn_v2.5_monthly/.]

  • Meyer, S. J., and Hubbard K. G. , 1992: Nonfederal automated weather stations and networks in the United States and Canada: A preliminary survey. Bull. Amer. Meteor. Soc., 73, 449457, doi:10.1175/1520-0477(1992)073<0449:NAWSAN>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Misra, V., Michael J.-P. , Boyles R. , Chassignet E. P. , Griffin M. , and O’Brien J. J. , 2012: Reconciling the spatial distribution of the surface temperature trends in the southeastern United States. J. Climate, 25, 36103618, doi:10.1175/JCLI-D-11-00170.1.

    • Search Google Scholar
    • Export Citation
  • National Research Council, 1998: Future of the National Weather Service Cooperative Observer Network. National Academy Press, 65 pp.

  • National Research Council, 2009: Observing Weather and Climate from the Ground Up: A Nationwide Network of Networks. National Academy Press, 234 pp.

  • National Weather Service, 2014: Cooperative station observations. National Weather Service Manual 10-1315, Surface Observing Program, Operations and Services, 138 pp.

  • Peterson, T. C., and Coauthors, 1998: Homogeneity adjustments of in situ atmospheric climate data: A review. Int. J. Climatol., 18, 14931517, doi:10.1002/(SICI)1097-0088(19981115)18:13<1493::AID-JOC329>3.0.CO;2-T.

    • Search Google Scholar
    • Export Citation
  • Pielke, R. A., Sr., and Coauthors, 2007: Unresolved issues with the assessment of multidecadal global land surface temperature trends. J. Geophys. Res., 112, D24S08, doi:10.1029/2006JD008229.

    • Search Google Scholar
    • Export Citation
  • Price, C. V., Nakagaki N. , Hitt K. J. , and Clawges R. M. , 2006: Enhanced historical land‐use and land‐cover data sets of the U.S. Geological Survey. U.S. Geological Survey Data Series 240. [Available online at http://pubs.usgs.gov/ds/2006/240/.]

  • Quayle, R. G., Easterling D. R. , Karl T. R. , and Hughes P. Y. , 1991: Effects of recent thermometer changes in the cooperative station network. Bull. Amer. Meteor. Soc., 72, 17181723, doi:10.1175/1520-0477(1991)072<1718:EORTCI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Quinlan, F. T., Karl T. R. , and Williams C. N. Jr., 1987: United States Historical Climatology Network (HCN) serial temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, NDP-019, 33 pp.

  • Rothfusz, L. P., Crawford K. , Hoogenboom G. , Stooksbury D. E. , and Knox P. N. , 2006: The Georgia mesonet: Concepts and systems to demonstrate a new cooperative climate network. 22nd Conf. on Interactive Processing Systems for Meteorology, Oceanography, and Hydrology, Atlanta, GA, Amer. Meteor. Soc., J5.8. [Available online at https://ams.confex.com/ams/Annual2006/techprogram/paper_102169.htm.]

  • Schwarz, G. E., and Alexander R. B. , 1995: State soil geographic (STATSGO) data base for the conterminous United States. USGS Open-file Rep. 95-449, accessed 29 March 2014. [Available online at http://water.usgs.gov/GIS/metadata/usgswrd/XML/ussoils.xml.]

  • Trewin, B., 2010: Exposure, instrumentation, and observing practice effects on land temperature measurements. Wiley Interdiscip. Rev.: Climate Change, 1, 490506, doi:10.1002/wcc.46.

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Commerce, 2004: COOP modernization: Building the national cooperative mesonet; Program development plan. NOAA/NWS, 73 pp.

  • Vose, R. S., Williams C. N. Jr., Peterson T. C. , Karl T. R. , and Easterling D. R. , 2003: An evaluation of the time of observation bias adjustment in the U.S. Historical Climatology Network. Geophys. Res. Lett., 30, 2046, doi:10.1029/2003GL018111.

    • Search Google Scholar
    • Export Citation
  • White, L. D., and Matlack E. , 2007: The Mississippi mesonet: Phase 2. 16th Conf. on Applied Climatology, San Antonio, TX, Amer. Meteor. Soc., JP1.6. [Available online at https://ams.confex.com/ams/87ANNUAL/techprogram/paper_119231.htm.]

  • Wu, H., Hubbard K. G. , and You J. , 2005: Some concerns when using data from the cooperative weather station networks: A Nebraska case study. J. Atmos. Oceanic Technol., 22, 592602, doi:10.1175/JTECH1733.1.

    • Search Google Scholar
    • Export Citation
  • Yao, W., and Zhong S. , 2009: Nocturnal temperature inversions in a small, enclosed basin and their relationship to ambient atmospheric conditions. Meteor. Atmos. Phys., 103, 195210, doi:10.1007/s00703-008-0341-4.

    • Search Google Scholar
    • Export Citation
Save
  • Allard, J., and Carleton A. M. , 2010: Mesoscale interactions between Midwest land surface properties and convective cloud development in the warm season. Phys. Geogr., 31, 107136, doi:10.2747/0272-3646.31.2.107.

    • Search Google Scholar
    • Export Citation
  • Baker, D. G., 1975: Effect of observation time on mean temperature estimation. J. Appl. Meteor., 14, 471476, doi:10.1175/1520-0450(1975)014<0471:EOOTOM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bodine, D., Klein P. M. , Arms S. C. , and Shapiro A. , 2009: Variability of surface air temperature over gently sloping terrain. J. Appl. Meteor. Climatol., 48, 11171141, doi:10.1175/2009JAMC1933.1.

    • Search Google Scholar
    • Export Citation
  • Brown, Z. G., Foster S. A. , Mahmood R. , Strudbig S. , Grabowski D. , Grogan M. , and Ferris M. M. , 2008: An overview of the Kentucky mesonet. 17th Conf. on Applied Climatology, Whistler, BC, Canada, Amer. Meteor. Soc., 1.3. [Available online at https://ams.confex.com/ams/13MontMet17AP/techprogram/paper_140983.htm.]

  • Brusberg, M. D., and Hubbard K. G. , 2001: Preliminary findings of the 1999 survey of (nonfederal) automated weather stations in the United States and Canada. Automated weather stations for applications in agriculture and water resources management: Current use and future perspectives; Proceedings of an international workshop, K. G. Hubbard and M. V. K. Sivakumar, Eds., WMO Tech. Doc. WMO/TD-1074, 229–240.

  • Bruton, J. M., Hoogenboom G. , and McClendon R. W. , 1998: Comparison of automatically and manually collected weather data. Preprints, 1998 ASAE Annual Int. Meeting, Orlando, FL, American Society of Agricultural Engineers, Paper 98-2188.

  • Changnon, S. A., and Kunkel K. E. , 1999: Rapidly expanding uses of climate data and information in agriculture and water resources: Causes and characteristics of new applications. Bull. Amer. Meteor. Soc., 80, 821830, doi:10.1175/1520-0477(1999)080<0821:REUOCD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Christy, J. R., Norris W. B. , and McNider R. T. , 2009: Surface temperature variations in East Africa and possible causes. J. Climate, 22, 33423356, doi:10.1175/2008JCLI2726.1.

    • Search Google Scholar
    • Export Citation
  • Daly, C., 2006: Guidelines for assessing the suitability of spatial climate data sets. Int. J. Climatol., 26, 707721, doi:10.1002/joc.1322.

    • Search Google Scholar
    • Export Citation
  • Durre, I., Menne M. J. , Gleason G. E. , Houston T. G. , and Vose R. S. , 2010: Comprehensive automated quality assurance of daily surface observations. J. Appl. Meteor. Climatol., 49, 16151633, doi:10.1175/2010JAMC2375.1.

    • Search Google Scholar
    • Export Citation
  • Easterling, D. R., Karl T. R. , Mason E. H. , Hughes P. Y. , and Bowman D. P. , 1996: United States Historical Climatology Network (U.S. HCN) monthly temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, ORNL/CDIAC-87, NDP-019/R3, 280 pp.

  • Fiebrich, C. A., 2009: History of surface weather observations in the United States. Earth-Sci. Rev., 93, 7784, doi:10.1016/j.earscirev.2009.01.001.

    • Search Google Scholar
    • Export Citation
  • Fiebrich, C. A., and Crawford K. C. , 2009: Automation: A step toward improving the quality of daily temperature data by climate observing networks. J. Atmos. Oceanic Technol., 26, 12461260, doi:10.1175/2009JTECHA1241.1.

    • Search Google Scholar
    • Export Citation
  • Gallo, K. P., Easterling D. R. , and Peterson T. C. , 1996: The influence of land use/land cover on climatological values of the diurnal temperature range. J. Climate, 9, 29412944, doi:10.1175/1520-0442(1996)009<2941:TIOLUC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gesch, D., Evans G. , Mauck J. , Hutchinson J. , and Carswell W. J. Jr., 2009: The national map—Elevation. U.S. Geological Survey Fact Sheet 2009-3053, 4 pp., accessed 24 February 2014. [Available online at http://viewer.nationalmap.gov/viewer/.]

  • Guttman, N. B., and Baker C. B. , 1996: Exploratory analysis of difference between temperature observations recorded by ASOS and conventional methods. Bull. Amer. Meteor. Soc., 77, 28652873, doi:10.1175/1520-0477(1996)077<2865:EAOTDB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hale, R. C., Gallo K. P. , and Loveland T. R. , 2008: Influences of land use/land cover conversions on climatological normals of near-surface temperature. J. Geophys. Res., 113, D14113, doi:10.1029/2007JD009548.

    • Search Google Scholar
    • Export Citation
  • Hanamean, J. R., Jr., Pielke R. A. Sr., Castro C. L. , Ojima D. S. , Reed B. C. , and Gao Z. , 2003: Vegetation greenness impacts on maximum and minimum temperatures in northeast Colorado. Meteor. Appl., 10, 203215, doi:10.1017/S1350482703003013.

    • Search Google Scholar
    • Export Citation
  • Holder, C., Boyles R. , Syed A. , Niyogi D. , and Raman S. , 2006: Comparison of collocated automated (NC ECONet) and manual (COOP) climate observations in North Carolina. J. Atmos. Oceanic Technol., 23, 671682, doi:10.1175/JTECH1873.1.

    • Search Google Scholar
    • Export Citation
  • Hoogenboom, G., 2005: The Georgia Automated Environmental Monitoring Network: Experiences with the development of a state-wide automated weather station network for Georgia. 15th Conf. on Applied Climatology, Savannah, GA, Amer. Meteor. Soc., JP1.16. [Available online at https://ams.confex.com/ams/15AppClimate/webprogram/Paper94108.html.]

  • Hoogenboom, G., Coker D. D. , Edenfield J. M. , Evans D. M. , and Fang C. , 2003: The Georgia Automated Environmental Monitoring Network: Ten years of weather information for water resources management. Proceedings of the 2003 Georgia Water Resources Conference, K. J. Hatcher, Ed., The University of Georgia, 896900.

  • Horel, J., and Coauthors, 2002: Mesowest: Cooperature mesonets in the western United States. Bull. Amer. Meteor. Soc., 83, 211226, doi:10.1175/1520-0477(2002)083<0211:MCMITW>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hubbard, K. G., 2001: The Nebraska and High Plains regional experience with automated weather stations. Automated weather stations for applications in agriculture and water resources management: Current use and future perspectives; Proceedings of an international workshop, K. G. Hubbard and M. V. K. Sivakumar, Eds., WMO Tech. Doc. WMO/TD-1074, 219–227.

  • Hubbard, K. G., Rosenberg N. J. , and Nielsen D. C. , 1983: Automated weather data network for agriculture. J. Water Resour. Plann. Manage., 109, 213222, doi:10.1061/(ASCE)0733-9496(1983)109:3(213).

    • Search Google Scholar
    • Export Citation
  • Hubbard, K. G., Lin X. , Baker C. B. , and Sun B. , 2004: Air temperature comparison between the MMTS and the USCRN temperature systems. J. Atmos. Oceanic Technol., 21, 15901597, doi:10.1175/1520-0426(2004)021<1590:ATCBTM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hughes, P. Y., Mason E. H. , Karl T. R. , and Brower W. A. , 1992: United States Historical Climatology Network daily temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, CDIAC-50, NDP-042, 140 pp.

  • IPCC, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp., doi:10.1017/CBO9781107415324.

  • Karl, T. R., and Williams C. N. Jr., 1987: An approach to adjusting climatological time series for discontinuous inhomogeneities. J. Climate Appl. Meteor., 26, 17441763, doi:10.1175/1520-0450(1987)026<1744:AATACT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., Williams C. N. Jr., Young P. J. , and Wendland W. M. , 1986: A model to estimate the time of observation bias associated with monthly mean maximum, minimum, and mean temperature for the United States. J. Climate Appl. Meteor., 25, 145160, doi:10.1175/1520-0450(1986)025<0145:AMTETT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Karl, T. R., Williams C. N. Jr., and Quinlan F. T. , 1990: United States Historical Climatology Network (HCN) serial temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, ORNL/CDIAC-30, NDP-019/R1, 83 pp.

  • Kimball, S. K., Mulekar M. S. , Cummings S. , and Stamates J. , 2010: The University of South Alabama Mesonet and coastal observing system: A technical and statistical overview. J. Atmos. Oceanic Technol., 27, 14171439, doi:10.1175/2010JTECHA1376.1.

    • Search Google Scholar
    • Export Citation
  • Leeper, R. D., Rennie J. , and Palecki M. A. , 2015: Observational perspectives from U.S. Climate Reference Network (USCRN) and Cooperative Observer Program (COOP) network: Temperature and precipitation comparison. J. Atmos. Oceanic Technol., 32, 703721, doi:10.1175/JTECH-D-14-00172.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., and Williams C. N. Jr., 2009: Homogenization of temperature series via pairwise comparisons. J. Climate, 22, 17001717, doi:10.1175/2008JCLI2263.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Williams C. N. Jr., and Vose R. S. , 2009: The United States Historical Climatology Network monthly temperature data, version 2. Bull. Amer. Meteor. Soc., 90, 9931007, doi:10.1175/2008BAMS2613.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Williams C. N. Jr., and Palecki M. A. , 2010: On the reliability of the U.S. surface temperature record. J. Geophys. Res., 115, D11108, doi:10.1029/2009JD013094.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Durre I. , Vose R. S. , Gleason B. E. , and Houston T. G. , 2012: An overview of the Global Historical Climatology Network-Daily database. J. Atmos. Oceanic Technol., 29, 897910, doi:10.1175/JTECH-D-11-00103.1.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., Williams C. N. Jr., and Vose R. S. , 2014: United States Historical Climatology Network (USHCN) version 2.5 serial monthly dataset. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, accessed 4 March 2014. [Available online at http://cdiac.ornl.gov/ftp/ushcn_v2.5_monthly/.]

  • Meyer, S. J., and Hubbard K. G. , 1992: Nonfederal automated weather stations and networks in the United States and Canada: A preliminary survey. Bull. Amer. Meteor. Soc., 73, 449457, doi:10.1175/1520-0477(1992)073<0449:NAWSAN>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Misra, V., Michael J.-P. , Boyles R. , Chassignet E. P. , Griffin M. , and O’Brien J. J. , 2012: Reconciling the spatial distribution of the surface temperature trends in the southeastern United States. J. Climate, 25, 36103618, doi:10.1175/JCLI-D-11-00170.1.

    • Search Google Scholar
    • Export Citation
  • National Research Council, 1998: Future of the National Weather Service Cooperative Observer Network. National Academy Press, 65 pp.

  • National Research Council, 2009: Observing Weather and Climate from the Ground Up: A Nationwide Network of Networks. National Academy Press, 234 pp.

  • National Weather Service, 2014: Cooperative station observations. National Weather Service Manual 10-1315, Surface Observing Program, Operations and Services, 138 pp.

  • Peterson, T. C., and Coauthors, 1998: Homogeneity adjustments of in situ atmospheric climate data: A review. Int. J. Climatol., 18, 14931517, doi:10.1002/(SICI)1097-0088(19981115)18:13<1493::AID-JOC329>3.0.CO;2-T.

    • Search Google Scholar
    • Export Citation
  • Pielke, R. A., Sr., and Coauthors, 2007: Unresolved issues with the assessment of multidecadal global land surface temperature trends. J. Geophys. Res., 112, D24S08, doi:10.1029/2006JD008229.

    • Search Google Scholar
    • Export Citation
  • Price, C. V., Nakagaki N. , Hitt K. J. , and Clawges R. M. , 2006: Enhanced historical land‐use and land‐cover data sets of the U.S. Geological Survey. U.S. Geological Survey Data Series 240. [Available online at http://pubs.usgs.gov/ds/2006/240/.]

  • Quayle, R. G., Easterling D. R. , Karl T. R. , and Hughes P. Y. , 1991: Effects of recent thermometer changes in the cooperative station network. Bull. Amer. Meteor. Soc., 72, 17181723, doi:10.1175/1520-0477(1991)072<1718:EORTCI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Quinlan, F. T., Karl T. R. , and Williams C. N. Jr., 1987: United States Historical Climatology Network (HCN) serial temperature and precipitation data. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, NDP-019, 33 pp.

  • Rothfusz, L. P., Crawford K. , Hoogenboom G. , Stooksbury D. E. , and Knox P. N. , 2006: The Georgia mesonet: Concepts and systems to demonstrate a new cooperative climate network. 22nd Conf. on Interactive Processing Systems for Meteorology, Oceanography, and Hydrology, Atlanta, GA, Amer. Meteor. Soc., J5.8. [Available online at https://ams.confex.com/ams/Annual2006/techprogram/paper_102169.htm.]

  • Schwarz, G. E., and Alexander R. B. , 1995: State soil geographic (STATSGO) data base for the conterminous United States. USGS Open-file Rep. 95-449, accessed 29 March 2014. [Available online at http://water.usgs.gov/GIS/metadata/usgswrd/XML/ussoils.xml.]

  • Trewin, B., 2010: Exposure, instrumentation, and observing practice effects on land temperature measurements. Wiley Interdiscip. Rev.: Climate Change, 1, 490506, doi:10.1002/wcc.46.

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Commerce, 2004: COOP modernization: Building the national cooperative mesonet; Program development plan. NOAA/NWS, 73 pp.

  • Vose, R. S., Williams C. N. Jr., Peterson T. C. , Karl T. R. , and Easterling D. R. , 2003: An evaluation of the time of observation bias adjustment in the U.S. Historical Climatology Network. Geophys. Res. Lett., 30, 2046, doi:10.1029/2003GL018111.

    • Search Google Scholar
    • Export Citation
  • White, L. D., and Matlack E. , 2007: The Mississippi mesonet: Phase 2. 16th Conf. on Applied Climatology, San Antonio, TX, Amer. Meteor. Soc., JP1.6. [Available online at https://ams.confex.com/ams/87ANNUAL/techprogram/paper_119231.htm.]

  • Wu, H., Hubbard K. G. , and You J. , 2005: Some concerns when using data from the cooperative weather station networks: A Nebraska case study. J. Atmos. Oceanic Technol., 22, 592602, doi:10.1175/JTECH1733.1.

    • Search Google Scholar
    • Export Citation
  • Yao, W., and Zhong S. , 2009: Nocturnal temperature inversions in a small, enclosed basin and their relationship to ambient atmospheric conditions. Meteor. Atmos. Phys., 103, 195210, doi:10.1007/s00703-008-0341-4.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    The locations of AEMN (triangles) and USHCN (circles) stations. The names of corresponding paired USHCN stations can be found in Table 1.

  • Fig. 2.

    The locations of AEMN (triangles) and COOP (circles) stations. The names of corresponding paired COOP stations can be found in Table 5.

  • Fig. 3.

    Scatterplots showing the average seasonal temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs distance (km) between the same paired stations.

  • Fig. 4.

    Scatterplots showing the average seasonal temperature difference (°C) between paired AEMN–COOP stations (AEMN minus COOP) vs distance (km) between the same paired stations.

  • Fig. 5.

    Scatterplots showing the average seasonal minimum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in latitude (decimal degrees) and longitude (decimal degrees) between the same paired stations.

  • Fig. 6.

    Scatterplots showing the average seasonal maximum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in latitude (decimal degrees) and longitude (decimal degrees) between the same paired stations.

  • Fig. 7.

    Scatterplots showing the average seasonal minimum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in elevation (m), slope (°), and aspect [the compass direction of the topographic slope (°)] between the same paired stations for the spatial scales in Table A1.

  • Fig. 8.

    Scatterplots showing the average seasonal maximum temperature difference (°C) between paired AEMN–USHCN stations (AEMN minus USHCN) vs differences in elevation (m), slope (°), and aspect [the compass direction of the topographic slope (°)] between the same paired stations for the spatial scales in Table A1.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 460 101 4
PDF Downloads 334 48 3