1. Introduction
During the late 19th to early 20th century West Virginia (WV) was the frontier of American industrial capitalism where the rich natural resources of the Mountain State, especially timber and coal, were exploited for the benefit of aggressively populating eastern cities [
1,
2]. In 1880, it was estimated that two thirds of WV was covered by ancient growth hardwood forest [
2], but by 1909 approximately 72% of WV was agriculture or pasture lands [
3]. Widespread logging created a landscape characterized by vast acreages of dried slash that were easily ignited by passing locomotives [
3] and burned with unusual intensity during dry periods [
4]. By the 1920s, most lumbermen abandoned WV for lands further south and west [
2] and extensive fire prevention efforts resulted in the redevelopment of vast hardwood forests [
4]. When the first United States (US) Forest Service forest survey of WV was completed in 1949 an estimated 64% of WV lands were forested, which increased to 75% by the 1961 forest survey [
5] and was relatively steady (75% to 79%) through 2012 [
6]. Despite relatively steady areal coverage since the 1961 forest survey, timber volume increased through 2012 consistent with a maturing forest ecosystem [
6]. Given the magnitude of reforestation throughout the broader Appalachian region, WV is an ideal location to assess how the regrowth of native forests may have influenced (or been influenced by) local to regional climatic trends.
Changes in land use and land cover (LULC) alters radiative (e.g., absorption and reflection) and non-radiative (e.g., surface roughness and evapotranspiration) climate forcing that influence spatial and temporal patterns of temperature and precipitation [
7,
8]. Few studies have focused on climatic influence of reforestation in the northeast US and studies utilizing observations are less frequent than modeling efforts [
7,
9]. Considering results of model and observation based analyses are contradictory [
7,
9], spatiotemporal analysis of the climatic influence of extensive reforestation over a long period of record (POR) is needed in WV and other similar regions globally. For example, spatiotemporal patterns in minimum temperature, maximum temperature, and precipitation trends are valuable to agricultural interests [
10], natural resource managers (i.e., water, land, energy, wildlife; [
11,
12]), and other public and private stakeholders. More specifically, spatiotemporal climatic analyses are needed to better understand and prepare for terrestrial ecosystem vulnerabilities that may impact habitat suitability, human health, and food security [
12]. Impacts to human health and food security are particularly relevant to West Virginia given the broader Appalachian region classification as a food desert (i.e., low income and limited access to fresh produce) by the US Department of Agriculture, Food and Nutrition Services [
13]. Quantifying spatiotemporal climatic changes within WV will guide future climatic analyses, inform natural resource management decisions, and help preserve WV’s diverse terrestrial ecosystem.
West Virginia’s climate, water resources, and ecology are intricately linked and all three are influenced by WV’s complex terrain. Elevation gradually increases from the Ohio River (~300 m) on the western border of WV to the ridges of the Allegheny Mountains (>1000 m) in the eastern third of WV. Annual average precipitation is approximately 100 cm across the far western portions of the state (Wheeling, WV; 102.6 cm), gradually increasing to near 150 cm at the highest elevations (Snowshoe Mountain, WV; 152.1 cm), and abruptly decreases to near 90 cm immediately east of the Allegheny Mountains (Franklin, WV; 91.5 cm; [
14]). Complex climatic gradients resulting from WV’s rugged terrain make the Monongahela National Forest, located in the Allegheny Mountains, one of the most ecologically diverse forests in the US National Forest System [
15]. Stands of red spruce and populations of snowshoe hare, more typical of northern boreal forests, occur at higher elevations in WV. Localized cold air traps (i.e., deep rock crevices, talus slopes, and caves) provide refugium for organisms adapted to survival in cold and dark environments (e.g., bacteria and fungi; [
16]). Drier east-facing slopes located east of the Allegheny ridges support prickly pear cactus and rare plant species including the shale barren rock cress [
15]. Lowland locations, including mountain valleys, are characterized by stands of mixed hardwoods growing in relatively deep and rich soils more typical of southern Appalachia. Coupling of WV’s climatic gradients, biodiversity, and physiography suggests an opportunity to manage WV’s water resources using an integrated and multidisciplinary approach. However, before management practices can be advanced, investigation of spatiotemporal climatic changes during and after severe and widespread anthropogenic disturbance (e.g., deforestation and reforestation) is needed.
The overarching objective of the current work was to leverage state and federally funded WV climate observation datasets to quantify spatiotemporal characteristics of long-term (≥77 years) minimum temperature, maximum temperature, and daily precipitation trends across West Virginia, USA. Sub objectives included (a) assessing statistically significant (p < 0.05) trends using the Mann–Kendall trend test, (b) performing analyses during the first (1900–1958) and second (1959–2016) half of the time series, and (c) discussing the ecological and topographic dependence of climatic trends across WV.
2. Materials and Methods
Daily summaries of total precipitation, maximum temperature, and minimum temperature were acquired from the National Center for Environmental Information (NCEI), a branch of the National Oceanic and Atmospheric Administration (NOAA; [
17]). Stations were selected with start dates as early as 1900, but no later than 1930, with a nearly continuous time series through the end of 2016. These temporal criteria ensured analyses included the rapid transition from primarily agricultural and pasture lands in the early 20th century to primarily forested land coverage by the middle 20th century and subsequent forest maturation [
3,
6]. All leap days (i.e., February 29th) were excluded from analyses and daily data were post-processed to ensure all missing dates were included resulting in a continuous time series with 365 days in each year. Annual averages of daily summaries of maximum temperature, minimum temperature, and total precipitation were estimated and counts of missing data points were tabulated annually for each of the 18 selected stations. The spatial distribution of the 18 selected stations is shown in
Figure 1 and additional site-specific data are included in Table 2. It was assumed that observers and their equipment followed station siting guidelines established by the United States Weather Bureau in 1890 (Fiebrich 2009). The World Meteorological Organization’s guidelines on missing data in the calculation of climatological normals recommends not calculating monthly averages when five or more daily values are missing [
18], which scales to 60 or more daily values for annual averages. For the purposes of this work years with data gaps exceeding 15% of possible daily observations (>54 daily observations) were removed from analyses. Of 117 possible annual averages or totals, each observation location had between 77 and 110 annual averages for maximum and minimum temperature and between 84 and 110 annual precipitation totals.
Investigating temporal trends of the central tendency of a data series is of interest in the context of identifying a changing climate [
20,
21]. Sen’s slope estimator paired with the Mann–Kendall trend test is a robust, non-parametric method to estimate the linear trend and assess statistical significance [
22] and was thus deemed appropriate for this work. Spatial characteristics of temporal trends were estimated and statistical significance (α = 0.05) assessed over the entire time series (1900–2016) and during the first (1900–1958) and second (1959–2016) halves of each time series at each observation location. Analyses for each half of the time series were performed because the data series was sufficiently long and the first half corresponded with reforestation whereas the second half corresponded with forest maturation and globally averaged warming exceeding 0.65 °C [
4,
5,
6,
23]. Except for New Cumberland (
Table 1,
Figure 1), each observation location had a more complete time series of each variable during the second half (93.8%) rather than the first half (71.7%) of the POR suggesting early data gaps (1900–1930) could have influenced results. As a result, the data should not be considered stationary and readers should use caution when extrapolating results into the future. Changes in temporal characteristics (i.e., seasonality) were estimated for all 365 days using a continuous time series of daily data averaged across all eighteen locations for the entire POR and the second half of the POR (1959–2016). Seasonal analyses were not performed for the first half of the POR due to larger and more frequent data gaps. For plotting purposes, centered three-week moving averages of daily Sen’s slope, upper and lower 95% confidence interval values were calculated to smooth day-to-day variability and better show seasonal changes of each variable. Centered moving averages (CMAs) represent a series of arithmetic means throughout a time series that is centered on the middle value (i.e., day 11 of the 21-day average). Three-week CMAs were selected because meteorological and astronomical definitions of seasonality are static in time and space and may be inadequate considering a swiftly changing climate [
24]. Ultimately, the methods described in the current work are easily applied and provide valuable information about spatiotemporal changes in climate that are needed for effective decision making.
4. Discussion
Extensive LULC changes occurred across West Virginia (WV) and the broader Appalachian region between 1900 and 2016 that greatly contrast global LULC changes [
26]. WV’s dominant LULC transitioned from agricultural land (72%) to native forest cover (64%) between 1909 and 1949 [
5]. Forest cover continued to increase in areal coverage (79%) and density (i.e., timber volume) through 2012 [
6]. LULC changes influence biogeophysical regulation of climate associated with changes in albedo, turbulent fluxes of energy, mass, and momentum at the land–atmosphere interface, and the hydrologic cycle [
26,
27,
28,
29]. Simulations of coupled land–atmosphere interactions indicated that temperate forests are characterized by lower albedo and evapotranspiration values relative to irrigated croplands resulting in significantly warmer temperatures [
29]. However, near surface eddy-covariance observations were compared between forest cover and an adjacent, non-irrigated grass field in nearby North Carolina, USA [
30]. Results of Reference [
30] showed the cooling effect of greater bulk aerodynamic conductance in forests was about three times larger than warming attributable to albedo effects. Therefore, the rapid transition from non-irrigated agricultural to forested land uses in WV may have resulted in a net cooling effect.
Net cooling in WV was characterized by decreasing state-average maximum temperatures (−1.0 °C) and increasing minimum temperatures (+0.4 °C) between 1900 and 2016 [
9], but spatial and seasonal changes were not addressed. At the regional scale, changes in maximum and minimum temperatures contrasted the reported warming of 0.64 °C and 0.94 °C across the broader Northeast US region that included WV, respectively [
31]. However, temperature trends in WV were more consistent with the Southeast US region during the period of study where maximum and minimum temperatures increased 0.16 °C and 0.76 °C, respectively [
31]. In WV, minimum temperatures increased more at lower elevation locations (
Table 1) where old-growth hardwood forests were harvested, conceivably with relatively less effort than WV’s rugged mountainous terrain [
2]. Between 1900 and 2016, increasing daily minimum temperature trends were consistently positive during JJA when 84% of significant (
p < 0.05) increasing daily trends occurred. Similarly, 76% of the seventeen significantly positive (
p < 0.05) daily average trends between 1959 and 2016 occurred during JJA. Warming summertime minimum temperatures in WV’s humid subtropical climate suggests warmer dew point temperatures that may alter dewfall dynamics [
32] and contribute to WV’s increasingly humid climate [
9]. Maximum temperatures cooled significantly (
p < 0.05) on 72 days between 1900 and 2016 (
Figure 4) and annual trends were larger at higher elevations between 1959 and 2016 (
Table 2) where forest cover may be less disturbed and more mature [
3]. Cooling high-elevation maximum temperatures may be explained by topographically induced circulations focusing afternoon cloud development along high-elevation ridges [
33] that could be more frequent in an increasingly humid climate [
9]. Further investigation into the mechanism(s) responsible for differential temperature trends at low and high elevation locations is needed, but warmer minimum and cooler maximum temperatures are consistent with attenuation of the diurnal temperature range due to forest cover [
34] or cloud cover [
35]. In addition to LULC changes in WV, increased forest coverage and maturity, and agricultural intensification (i.e., fertilization and irrigation; [
36]) upwind of WV (i.e., Midwestern US) likely contributed to WV’s increasingly humid WV climate [
37]. Additional contributing mechanisms include (but are not limited to) increases in globally averaged atmospheric moisture associated with warming air and sea surface temperatures [
38] or regional precipitation recycling [
39]. Irrespective of the forcing mechanism(s), increased cloud cover aligns with an increasingly wet, temperate, and humid WV climate [
35] and suggests regional climate forcing associated with coupled land–atmosphere interactions exceeded radiative forcing in WV (i.e., greenhouse gases; [
40]).
Extrapolating annual trends over the entire POR then averaging across all eighteen locations indicated total annual precipitation increased by 3.8%, which is consistent with the national average reported in the Fourth National Climate Assessment [
31]. However, precipitation increased at all locations between 1959 and 2016 when linear trends averaged across all sites indicated a 14.2 cm (13.2%) increase in precipitation relative to the epoch average. This precipitation trend was more than triple the rate of increase during the entire POR suggesting precipitation trends may be accelerating as a result of global scale increases in water vapor content [
41] and super Clausius–Clapeyron scaling (SCCS) of precipitation extremes [
42]. Significant (
p < 0.05) increasing precipitation trends between 1959 and 2016 were estimated at lower elevation sites across western WV where forest harvesting, regrowth, and maturation was widespread [
2] implying greater feedbacks associated with LULC change. The methods documented in the current work could be used to address the need for data-driven assessments of feedbacks between climate and LULC change in many locations, globally [
43]. Understanding of LULC feedbacks on precipitation are particularly important in WV where some of the most extreme rainfall accumulations in the world were observed at time scales less than 6 h [
44]. Thus, WV may be particularly vulnerable to changes in the frequency and magnitude (i.e., SCCS) of precipitation extremes and the impact of LULC changes (e.g., reforestation) on precipitation extremes and SCCS should be investigated.
The combination of an increasingly wet and temperate climate should help to secure WV’s water resource quantity, but water quality problems exist in WV including (but not limited to) LULC change [
45], acid mine drainage [
46], and pathogenic water contamination [
47]. Additionally, WV’s increasingly temperate climate combined with decreasing temperature variance [
9] may prolong episodes of excessive soil wetness and humid conditions that increases the vulnerability of terrestrial ecosystems to pathogens and fungal-like oomycetes such as
Phytophtora and
Pythium [
48]. Similarly, increasing precipitation variance [
9] indicates wet years are becoming more extreme, which may exacerbate vulnerabilities associated with pests and pathogens [
49,
50,
51,
52,
53]. As a result, further investigation into changing climate variability may help to better explain observed changes in ecosystem biodiversity, carbon dynamics, and hydrology at a range of spatiotemporal scales [
54]. However, increasing average precipitation (i.e., freshwater availability) signals potential for increasing agricultural productivity to ameliorate the food desert crisis in WV through conservation practices sponsored by the United States Department of Agriculture’s (USDA) Environmental Quality Incentives Program (EQIP; [
55]). In particular EQIP’s High Tunnel System Initiative allows farmers to more efficiently deliver water and nutrients to plants, better protect plants from pests and pathogens, and minimize LULC change by increasing productivity per unit land area. Ultimately, climatic changes in WV may be distinct from the Northeast US climate region and Appalachia’s complex topography may therefore necessitate refined climatic regions based on climate trends observed over a long POR and physiographic properties [
56], including (but not limited to) elevation, slope and aspect, historic LULC changes, and forest age and composition.