Spatiotemporal variation in the microclimatic edge effect between wetland and farmland

Authors

  • Xiaoyu Liao,

    1. Northeast Institute of Geography and Agricultural Ecology, Chinese Academy of Sciences, Changchun, China
    2. Northeast Institute of Geography and Agricultural Ecology, University of Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Zhaoli Liu,

    Corresponding author
    1. Northeast Institute of Geography and Agricultural Ecology, Chinese Academy of Sciences, Changchun, China
    • Corresponding author: Z. Liu, Northeast Institute of Geography and Agricultural Ecology, Chinese Academy of Sciences, Shengbei Road 4888, Changchun 130102, China. (liuzhaoli@neigae.ac.cn)

    Search for more papers by this author
  • Yiyong Wang,

    1. Northeast Institute of Geography and Agricultural Ecology, Chinese Academy of Sciences, Changchun, China
    Search for more papers by this author
  • Jiming Jin

    1. Department of Watershed Sciences and Department of Plants, Soils, and Climate, Utah State University, Logan, Utah, USA
    Search for more papers by this author

Abstract

[1] The creation of wetland edges associated with anthropogenic fragmentation is common in wetland landscapes and brings about a unique microclimatic change that affects adjacent farmland. However, microclimatic edge effects within this kind of landscape are not well quantified. Thus, using a study site located in the northeastern Sanjiang National Nature Reserve, China, we investigated the spatiotemporal variations in the microclimatic variables of the surface layer across the wetland-farmland edge. Air temperature and relative humidity data were collected continuously along a horizontal transect at four different heights over different periods, and the relative humidity was converted to specific humidity. Sigmoid models were used to fit the horizontal gradients of microclimatic edge effects. Then, based on the parameters in the fitted models, the characteristics of the edge effects were quantified with two indices, the magnitude of edge effect (MEE) and the spatial range of edge effect (REE). Under a light wind condition, the microclimatic features across the edge generally presented sigmoid ecological gradients in the horizontal direction, with unique spatial and temporal patterns: (1) In a normal year, wetland patches cooled and moistened the adjacent farmlands during the daytime and had a warming-moistening effect during the nighttime. (2) Vertically, the absolute MEE value of the microclimatic variables decreased, but the REE increased with increasing height. (3) The MEE and REE for air temperature and specific humidity varied with the time of day, both shifting at dawn and dusk when the gradients were absent. (4) At the interannual scale, when compared to the farmland, the wetland was cooler at night in a dry year but warmer in a normal year. The detection of spatiotemporal microclimatic patterns in the wetland-farmland edge zone may enable further understanding of how wetland degradation affects adjacent farmland ecology under human disturbance and promote better management and restoration of fragmented wetlands.

1 Introduction

[2] Wetlands are the transition zones between terrestrial and aquatic ecosystems [Cowardin et al., 1979] and play a vital role in regulating local climate via both vegetation transpiration and water evaporation, which lead to lower air temperatures and higher air humidity than in nearby regions [Gao et al., 2002; Sun and Song, 2008]. The cooler and moister wetland microclimate inevitably affects adjacent ecosystems and brings about horizontal exchanges of water and energy, which produces horizontal gradients in air temperature and moisture around the edges of wetlands and adjacent areas. The meteorological phenomenon occurring on wetland edges may be defined as the wetland microclimatic edge effect.

[3] In the past few decades, extensive agricultural activities have converted vast natural wetlands into croplands in northeast China, particularly in Sanjiang Plain, in order to support the tremendous population growth there [Liu et al., 2005a, 2005b]. This has resulted in highly fragmented wetland landscapes, with an increase in both the number of patches and the complexity of their shapes. Agricultural activities have caused rapid growth along the edges between wetland and farmland patches, which has profoundly changed the regional microclimate and intensified the wetland microclimatic edge effect [Yan et al., 2002]. Therefore, it is essential to quantitatively explore the microclimatic edge effect in order to understand local climate change in such regions and its influence on neighboring crop production under the intensive exploitation of wetlands.

[4] Previous studies on microclimatic effects associated with wetlands and farmlands have shown horizontal microclimatic gradients, such as an air temperature increase and an air humidity decrease from wetland to farmland [e.g., Gao et al., 2003], but these investigations are far from a comprehensive quantitative interpretation of microclimatic edge effects. More intensive microclimatic edge effect studies have investigated forest ecosystems, focusing on the boundaries of forest patches and the abiotic factors affecting microclimatic edge gradients such as air temperature, relative humidity, and vapor pressure deficit [Chen et al., 1995; Gehlhausen et al., 2000; Pohlman et al., 2009; Wright et al., 2010]. It has been suggested that forest patch edges with abrupt transitions in vegetation type and structure generate unique microclimates [Murcia, 1995] that have more dramatic changes than those in forest patch interiors [Chen et al., 1995; Pohlman et al., 2009]. Many authors have further established numerical models to delineate the edge effects of forest microclimates by quantifying the spatial patterns and intensity of these effects [Cadenasso et al., 1997; Chen et al., 1992; Ewers and Didham, 2006; Hennenberg et al., 2008; Ries and Sisk, 2004]. The diurnal and seasonal variations in forest microclimatic edge effects have also been elaborated [Chen et al., 1993, 1995; Newmark, 2005; Pohlman et al., 2009; Wright et al., 2010]. However, these studies have been confined mostly to the edges of forest patches without regard for the microclimatic gradient of neighboring ecosystem patches. Although the interaction between the local climatic effects of oasis ecosystems and surrounding desert ecosystems has been discussed [Liu et al., 2005c], the research was conducted by comparing only a vertical section of their climatic variables rather than analyzing the horizontal microclimatic gradient between them.

[5] In this study, we aim to investigate the spatiotemporal patterns in microclimatic edge effects between wetland and farmland under a light wind condition with high-quality in situ observations. More specifically, this objective is accomplished by (1) establishing horizontal models to simulate wetland microclimatic edge effects and then using these models to investigate horizontal patterns and the characteristics of these effects, (2) comparing microclimatic features at four different heights to explore the vertical patterns of such effects, and (3) quantifying the diurnal cycles and interannual variability of the edge effects.

2 Methods

2.1 Study Sites

[6] The study sites are located in Sanjiang National Nature Reserve, in the northeast part of Fuyuan County, Heilongjiang Province, China, situated at the confluence of the Amur and Ussuri Rivers; the total area is 19.8 × 104 ha. Fieldwork was conducted in the specific region of 48°9'21.50" –48°10'8.60" N, 134°36'55.30" –134°38'33.10" E (Figure 1). This region has a temperate, semihumid continental monsoon climate with an annual mean temperature of 2.2°C. Mean monthly minimum temperatures are below −18.1°C (January), and mean monthly maximum temperatures are 22 °C (July). Mean annual rainfall is 600 mm, up to 60% of which falls from June to August [Song et al., 2008; Liu and Ma, 2002]. There is a prevailing southeasterly wind in summer and a northwesterly wind in winter, and the mean annual wind speed is 3.6 m/s. The Ussuri River runs through the entire region from the northeast to the southwest and divides the region into two parts. The southeast portion consists of floodplains and various low depressions with wet meadows and swamps; the Changbai flora predominates, with considerable development of halophytes, hygrophytes, and phreatophytes, e.g., Calamagrostis angustifolia, Carex lasiocarpa, and Phragmites australis [Liu and Ma, 2002]. The northwest portion consists of low flats covered by vast farmlands that produce mainly soybeans. Sanjiang National Nature Reserve, one of the three main inland marsh wetlands in China, has changed significantly in land use/cover since 1954 [Song et al., 2008]. This change has resulted in a decrease in natural wetlands from 3.5 × 106 ha in 1954 to 0.96 × 106 ha in 2005 and an increase in farmland from 1.7 × 106 ha to 5.6 × 106 ha [Song et al., 2008]. This remarkable landscape fragmentation has had a profound effect on the area's ecosystems and agriculture. Our study sites, which include the typical landscape within the reserve, are very representative of the whole Sanjiang Plain.

Figure 1.

Study sites and location of the sampling transect in Sanjiang National Nature Reserve, northeastern China. (bottom left) Landsat 7 ETM false color image from 15 Jun 15 2007; (bottom right) ALOS AVNIR-2 false color image from 2 Jun 2007. The bar at the lower right shows the positioning of the sampling sites along Transect 1 and Transect 2 across the edge from the wetland interior to the farmland interior (km).

2.2 Experimental Design

2.2.1 Sampling Scheme

[7] Two transects were designed perpendicular to the wetland-farmland edge (Figure 1). Transect 1, the main sampling line, was selected to capture the spatiotemporal variation in the microclimatic variables at the edge. The sampling sites were more densely arranged at the edges and gradually became sparser in the interiors of the wetland and farmland with an unequally spaced pattern. A total of eight sampling sites were set up facing northwest (azimuth 310°) in our study area, with three in the wetland and five in the farmland. For convenience of observation, the actual locations of these sites deviated slightly from the designed sampling line (see Figure 1), but the observation sites were generally located on both sides of the line. The southeast end was marked as WL1, and the other sites were respectively marked as WL2, WL3, FL1, FL2, FL3, FL4, and FL5. The distances between the southeast end and the other seven sites were 0.86, 1.09, 1.42, 1.53, 1.81, 2.08, and 2.52 km along the designed sampling line, respectively, and the boundary between the two patches was 1.4 km. Because one of our observing poles in the wetland was broken in 2010, the observations from only two wetland sampling sites, WL1 and WL2 (originally WL3), were used for analysis in this study.

2.2.2 Microclimate Measurements

[8] Our vertical measurements were taken within 15 m above the surface. Because air pressure has a minimal effect on temperature and moisture within this layer of the atmosphere, we chose vertical air temperature and specific humidity to quantify the microclimatic features without considering the influence of air pressure on these variables. However, we chose to use specific humidity rather than relative humidity or vapor pressure deficit, which have been commonly used in previous studies of microclimatic edge effect [Chen et al., 1993, 1995, 1999; Gehlhausen et al., 2000; Pohlman et al., 2009; Wright et al., 2010], because specific humidity is not affected by air temperature and can truly reflect spatial moisture variations.

[9] Vertical air temperature and relative humidity were measured with JL-06 data loggers assembled by Qingsheng Electronic Technology Co., Ltd., Handan, China (the temperature sensors were from Heraeus Holding GmbH, Germany, and the humidity sensors were from Honeywell International, Inc., USA; their accuracies were within ±0.1°C and ±1.5%). Ex-factory calibration was performed for all of the 85 purchased sensors for 2 years of validity in 1997, and the inaccurate sensors were sent back for repair. These sensors were recalibrated at the end of 1998. In addition, a fundamental intercomparison was conducted among all the sensors under the same environment before the experiments, and 64 sensors with high-quality consistent readings (below 0.1°C and 1.5%) were selected for use in this study. To detect the vertical variation in these two variables, 15 m observing poles were placed at each sampling site, where the data loggers were mounted at 2 m, 5 m, 10 m, and 15 m above the ground and at a 1 m distance from the poles to minimize their influence on the measurements. Two groups of parallel data for air temperature and relative humidity were continuously measured at every height and logged by the setting intervals. The specific humidity used for analysis was derived from temperature and relative humidity data (accuracy 0.1 g kg-1).

[10] To observe the microclimatic edge effects of wetland vegetation, field data were collected during the growing seasons over the period of 2008 through 2010. A drought occurred in our study area during 2008 (total precipitation was 182 mm from June to September), and only a small amount of waterlogging was seen in the wetland, resulting in a transition from marsh wetland to meadow vegetation dominated by Calamagrostis angustifolia. During 2009, the Ussuri River region had abundant precipitation (a total of 293 mm), and the observing poles were submerged. It was difficult to place the data loggers at the sampling sites, and thus, some data were missing, and we could not use the data from this year. During 2010, the study sites had normal weather with moderate rain (a total of 220 mm), and the farmland was waterlogged during the observation period. Therefore, our observations were taken under dry and normal climate backgrounds (between 2008 and 2010), and the responses of wetland edge effects to these different climatic conditions were quantitatively investigated in this study. At the sampling sites, air temperature and relative humidity were measured hourly at heights of 2 m, 5 m, 10 m, and 15 m above the ground during 2008, and every 30 min at 2 m and 5 m during 2010. Simultaneously, we also measured surface water depth in the wetland and farmland patches as well as plant heights, vegetation type and characteristics, and local weather conditions every day during the observation periods (Table 1).

Table 1. Characteristics of the Underlying Surface at Sampling Sites in the Growing Season During Observation Periods
DateHeight (m)WetlandFarmlandWeather/WindSite
6–10 Jun 20082/5/10/15Calamagrostis angustifolia, no waterloggingSeeded soybeans, no germinationSunny, NWTransect 1
20–22 Aug 20082/5/10/15Calamagrostis angustifolia, no waterloggingSoybeans 90–100 cm high, in seed-filling periodSunny, SETransect 1
10 Sep 2009100Marsh waterloggingSoybeans 60–70 cm high, dry surface soilSunny, SETransect 2
8–11 Jun 20102/5Marsh waterloggingSeeded soybeans, no germinationSunny, WTransect 1
7–13 Jul 20102/5Marsh a little waterlogging, moist soilSoybeans 20–30 cm high, dry surface soilSunny, showers on 12 Jun, WTransect 1
11–14 Aug 20102/5Marsh 180 cm high, seep depth of 20–50 cm, moist soilSoybeans 20–30 cm high and seep depth of 10–20 cm in FL1, FL2, FL5; 50–60 cm high and 10–30 cm depth in FL3 and FL4; moist soilSunny, moderate rain on 12 Aug, WTransect 1
16–21 Sep 20102/5More waterlogging and marsh 1 m high in WL1, no waterlogging in WL2Soybeans 70–80 cm high, moist soil with sporadic waterloggingPartly sunny/ cloudy, WTransect 1

2.3 Statistical Analysis

2.3.1 Data Processing

[11] It was necessary to preprocess invalid data and data that were missing due to the damaged instruments and electric power outages. Abnormal values were first eliminated by comparing the daily variation curves of the variables; then, the daily means and standard deviations were calculated, and the values that differed from the average by more than three times the deviation were also removed; finally, the eliminated data in which the intervals were less than 1 h were replaced using linear interpolation. This simple method was easy to perform, and the results differed little from other interpolations at 1 h.

[12] The average values of two groups of data were calculated after preprocessing, and air temperature and specific humidity values were separated into daytime (06:00–18:00 h) and nighttime (18:00–06:00 h) to calculate daily means in order to detect diurnal differences in the wetland radiation budget and evapotranspiration mechanism. To detect the microclimatic edge effect at different temporal scales, the average values at each sampling site and each height were calculated for each day and year to obtain daily and annual means of air temperature and specific humidity.

2.3.2 Selection of the Fitted Model

[13] Some models used in previous studies may not be suitable for simulating the horizontal gradients in microclimatic variables across the edge, e.g., linear models [Cadenasso et al., 1997; Cancino, 2005; Williams-Linera et al., 1998], which typically ignore the curvilinear and continuous nature of ecological responses, and exponential models [Chen et al., 1992; Laurance et al., 1998; Zheng and Chen, 2000], which describe microclimatic responses on only one side of the edge (Ewers and Didham, 2006]. Nevertheless, theoretical models of edge effects have repeatedly shown that there should be an asymptote on both sides of the boundary for microclimatic variables [Cadenasso et al., 2003a, 2003b; Ries and Sisk, 2004]. Therefore, general logistic models are an essential tool for quantifying wetland edge effects, which describe nonlinear continuous response functions to fit the variations across both sides of wetland boundaries and provide a powerful method to delineate the magnitude and range of edge effects in wetlands.

[14] In order to validate the horizontal patterns of microclimatic edge effects under a light wind day, the air temperature and specific humidity data were collected with high spatial resolution in southwestern Fuyuan County on 10 September 2009, not far from the main Transect 1. An 8.7 km transect (Transect 2) was delineated from the wetland interior to its neighboring farmland interior (Figure 1). Three data loggers, which were the same as those used to sample Transect 1, were mounted in an airship and used to measure air temperature and relative humidity every 1 min. A round-trip flight was accomplished from the boundary via the farmland interior to the wetland interior at an altitude of 100 m and an average speed of 10 m/s. Because Transect 2 crossed swamps and cropland and had underlying surface and environmental conditions similar to Transect 1 (Table 1), including the same light wind, it can validate the horizontal patterns of microclimatic edge effects (Figure 2). We selected observation periods with light wind to investigate the microclimatic horizontal gradients; thus, we mainly considered the influence of atmospheric turbulence in the surface layer on the wetland landscape and ignored the minor effect of advection. Thus, there appeared to be a symmetrical distribution of microclimatic variables across the wetland-farmland edge. By analyzing the measured data, we found the typical nonlinear sigmoid distribution in microclimatic variables across the edge (Pearson correlation coefficient r2 = 0.94, p < 0.01 for Ta; r2 = 0.92, p < 0.01 for Qa; Figure 2). In addition, to further justify this pattern, we also sampled along one side of the wetland edge into its interior at 2 m height with higher resolution (every 30 s) in Zhalong Nature Reserve, Heilongjiang Province. This reflected one side of the sigmoid distribution in microclimatic variables. Therefore, we selected the Boltzmann function, a form of general logistic model, to fit the microclimatic variation across the edge:

display math(1)

where y reflects the microclimatic variable at the edge, which drastically changes within a range of the x variable; x is the horizontal distance from the wetland interior WL1 (km); x0 is the center of the sigmoid curve, representing the distance between the boundary of the wetland and farmland patches where the change in the variable is strongest; dx is for the steepness of the change in the curve; and the constants Aw and Af characterize the upper and lower asymptotes of the sigmoid curve, respectively. These two parameters approximate the mean conditions of microclimatic variables in the wetland and adjacent farmland patch. This function is similar to that presented in Ewers and Didham [2006] and Hennenberg et al. [2008].

Figure 2.

Distribution and fitted results of cold-humid variables along Transect 2 from the wetland interior to the farmland interior. Filled circles represent air temperature (Ta, °C); filled triangles represent specific humidity (Qa, g kg-1). The left Y axis represents Ta, and the right Y axis is Qa; the X axis represents the distance across the edge from the wetland interior (the origin) to the farmland interior, ranging from 0 to 8.7 km.

2.3.3 Determining Edge Effect Indices

[15] We chose two quantitative indices, the magnitude of edge effect (MEE) and the range of edge effect (REE), to delineate the characteristics of microclimatic edge effects across the wetland edge under different spatiotemporal conditions. The indices calculated by the fitted parameters of the sigmoid model were able to fully describe the horizontal patterns of microclimatic features across the edge. The MEE is defined as the difference in ecological response variables from one patch interior through the adjacent patch interior, reflecting the gradients across the boundaries, which has been well documented in the literature [Cadenasso et al., 1997; Harper et al., 2005]. The REE is the horizontal distance over which ecological response variables vary along the edge; it is also referred to as the depth of edge influence [Chen et al., 1995] or the edge-effect penetration distance [Laurance and Yensen, 1991]. Previous studies used two methods to quantify the REE, one defining the REE as being 2/3 or 90% of the condition of the interior [Brand and George, 2001; Chen et al., 1992; Hylander, 2005] and the other using the first and second derivatives of the fitted models to estimate it [Ewers and Didham, 2006; Hufkens et al., 2008, 2009; Schmitz, 2004; Walker et al., 2003].

[16] Considering ecological and statistical significance, we delineated the MEE using differences in the two asymptotes in the sigmoid fitted model, approximating the difference between the maximum and minimum values of an ecological variable from the wetland interior to the adjacent farmland interior. First, the coefficients Aw and Af were computed by a least squares estimate for equation ((1)); then MEE = Aw – Af (Figure 3).

Figure 3.

Graphical representation of the magnitude of edge effect (MEE) and the range of edge effect (REE). For the positive trend, Bw = Aw + |Aw – Af |*0.025, and Bf = Af – |Af – Af|*0.025; for the negative trend, Bw = Aw – |Aw – Af |*0.025, and Bf = Af + |Aw – Af |*0.025.

[17] The second derivatives in the fitted models over different times and heights were first calculated, and the extent of the edge effects, defined as the distance between the local maxima and minima, was estimated by the inflection points of the fitted curves. We then found that the magnitude of the edge effects with this corresponding edge extent was only 50–65% of the difference in the two asymptotes |Aw—Af|. This method for estimating the REE [Ewers and Didham, 2006; Hufkens et al., 2008, 2009] did not reflect the real range of influence of the microclimatic variables, as very minor differences in the form of the sigmoid curve could have a profound effect on the distance. Therefore, we delineated the REE as being where the gradient difference was 95% of the magnitude of the edge effect (|Bw – Bf|/|Aw – Af| = 95%), depending on the shape of the curve. More specifically, the parameters Bw and Bf were calculated using the 95% criterion; then, the corresponding spatial distance Xw at Bw and Xf at Bf could be estimated by the fitted models; finally, REE = |Xw – Xf| (Figure 3). The significance of the effect of the independent variable, distance, on each dependent variable, Ta and Qa, was tested, and the edge effects (MEE and REE) on each microclimatic variable at each height during an entire day were examined using one-way ANOVA.

3 Results

3.1 Spatial Patterns of the Microclimatic Edge Effects

3.1.1 Horizontal Gradients of Microclimatic Features

[18] Due to the normal weather conditions in 2010, the daytime and nighttime temperature and specific humidity data at 2 m height for the 2010 growing season (June through September) were chosen to simulate and quantify the horizontal gradients of microclimatic variables (Figures 4 and 5). Air temperature at 2 m height in the wetland patch was lower than in the farmland patch during the day but higher at night (r2 = 0.85, p < 0.01 for daytime; r2 = 0.88, p < 0.01 for nighttime). The gradient of air temperature was steepest at the wetland-farmland boundary, and it gradually became weaker from the boundary into either the wetland or farmland interior, representing a characteristic sigmoid curve. Moreover, the air temperature gradient at the boundary varied more steeply during the daytime than during the nighttime (Figures 4a and 4b). For a quantitative description of the microclimatic edge effect (Figures 5a and 5b), the MEE for air temperature at 2 m height was −0.46°C during the daytime, equal to the value at night except for the sign. The REE during the day (600 m) was obviously smaller than at night (1200 m). This indicates that the wetland patch was the main determinant in regulating local microclimate, cooling during the day and warming at night, and the edge effects of the wetland patch on the adjacent farmland patch extended farther at night than during the day.

Figure 4.

Horizontal gradient patterns of (a and b) air temperature (Ta,°C) during the daytime and nighttime and (c and d) specific humidity (Qa, g kg-1) during the daytime and nighttime at 2 m height across the wetland-farmland edge during 2010. Filled squares denote mean values (±SE) for air temperature or specific humidity at 2 m height during 2010; the curve is fitted for 2 m height.

Figure 5.

Comparison of (a and c) MEE for air temperature (Ta, °C) and specific humidity (Qa, g kg-1) and (b and d) REE (m) for air temperature and specific humidity at 2 m height during both daytime and nighttime in 2008 and 2010.

[19] Specific humidity at 2 m height in the wetland patch was greater than in the farmland patch during both the daytime and nighttime (r2 = 0.92, p < 0.01 for daytime; r2 = 0.84, p < 0.01 for nighttime). Similar to air temperature, the horizontal gradient of specific humidity also presented a sigmoid curve, and the steepness of change was greater during the day (Figures 4c and 4d). The MEE values in the daytime were greater than in the nighttime, but the REE values were obviously less than during the nighttime (Figures 5c and 5d). This indicates that the wetland patch had an apparent moistening effect on the adjacent farmland throughout the day and that the extent of the influence was larger at night than during the day.

3.1.2 Vertical Variation of Microclimatic Features

[20] The data sets from 2008 at four different heights were selected to explore how microclimatic features of wetlands varied with height above the surface. The air temperatures at 15 m height during the nighttime varied too slightly to be fitted with the model. Thus, they were replaced with the mean values. From the fitted results of the microclimatic variables at the different heights, air temperature decreased with increasing height during the day and showed a temperature inversion at night (r2 = 0.87, 0.71, 0.58, and 0.71 at 2m, 5m, 10m, and 15m, p < 0.01, during the day; r2 = 0.89, 0.83, 0.42, at 2m, 5m, 10m, p < 0.01, at night; Figures 6a and 6b); specific humidity decreased both in the daytime and nighttime, except for a humidity inversion at 15 m height in both patches at night (r2 = 0.90, 0.72, 0.63, and 0.42 at 2m, 5m, 10m, and 15m, p < 0.01, during the day; r2 = 0.86, 0.74, 0.72, and 0.55, p < 0.01, at night; Figures 6c and 6d), which was likely related to the horizontal transport of water vapor by turbulence. The horizontal patterns of air temperature and specific humidity with a sigmoid curve were similar among the four different heights. However, in general, the closer to the ground surface, the greater the change rate and the steeper the fitted curve regardless of whether it was daytime or nighttime. It is also noteworthy that the wetland patch was cooler than the farmland patch at night, contrary to the horizontal patterns in 2010.

Figure 6.

Vertical patterns of (a and b) air temperature (Ta, °C) during daytime and nighttime and (c and d) specific humidity (Qa, g kg-1) during daytime and nighttime across the wetland-farmland edge during 2008 at four heights. Filled squares, filled circles, filled diamonds, and filled inverted triangles represent mean values (±SE) for air temperature and specific humidity at 2 m, 5 m, 10 m, and 15 m height, respectively; the solid, dashed, dash-dot-dotted, and dash-dotted curves are fitted for 2 m, 5 m, 10 m, and 15 m height, respectively.

[21] MEE values during both the daytime and nighttime were negative for air temperature (p < 0.05) but positive for specific humidity (p < 0.01 for daytime, p < 0.05 for nighttime), and all absolute MEE values generally decreased depending on height from 2 m to 15 m, except for the trend at 15 m, which could possibly be attributed to an observation error or insufficient sampling data (Figure 7a). Regardless of whether it was daytime or nighttime, the REE of air temperature extended with increasing height (p < 0.001 for daytime, p < 0.01 for nighttime; Figure 7b), and the REE of specific humidity also increased with height (p < 0.01 for daytime and nighttime; Figure 7b).

Figure 7.

Comparison of (a) MEE for air temperature (Ta, °C) and specific humidity (Qa, g kg-1) and (b) REE (km) for air temperature and specific humidity at heights from 2 m to 15 m during both daytime and nighttime during 2008. Open and filled circles represent MEE or REE for air temperature in the daytime and nighttime, respectively; open and filled triangles represent MEE or REE for specific humidity in the daytime or nighttime, respectively. The left Y axis in Figure 7a represents MEE for Ta (°C), and the right Y axis is MEE for Qa (g kg-1); the X axis represents heights from 2 m to 15 m. The left Y axis in Figure 7b represents REE for Ta and Qa (km).

3.2 Temporal Variation in the Microclimatic Edge Effects

3.2.1 Diurnal Patterns

[22] The diurnal changes in microclimatic variables across the edge were illustrated by the average values for the time of day in 2010, which were interpolated to 1 h (Figure 8). Edge gradients in air temperature and specific humidity varied with the time of day (p < 0.05 and n.s. for MEE of Ta and Qa, p < 0.001 and p < 0.01 for REE of Ta and Qa). Early in the morning (06:00–7:00 h) and late in the afternoon (18:00 h), there was a transition between the day and night patterns, during which there were small edge gradients for the two variables across the edge, leading to the more poorly fitted results and the MEE and REE being set at zero. In the daytime, the absolute MEE values for air temperature and specific humidity were increased between 8:00 and 12:00 h, and decreased between 15:00 and 17:00 h, while the REE generally increased; specifically, the edge effects were greatest with the maxima of the absolute MEE and the minimum of the REE in the middle of the day (13:00–14:00 h) (Figures 8a and 8b). Between 18:00 and 6:00 h, the nocturnal patterns of the MEE were reversed for air temperature while increased for specific humidity; the REE for both increased again, reaching maxima at 21:00–22:00 h (Figures 8a and 8b). The diurnal pattern indicates that the wetland patch had a predominant cooling effect during the day and consistently moistened the farmland patch throughout the entire day, but these microclimatic edge effects may weaken and disappear during transition periods.

Figure 8.

Diurnal variation of (a) MEE for air temperature (Ta, °C; filled circles) and specific humidity (Qa, g kg-1; filled triangles) and (b) REE for air temperature (km; filled circles) and specific humidity (km; filled triangles) with time of day.

3.2.2 Interannual Patterns

[23] Under dry and normal climatic conditions (2008 and 2010), the wetland patch was usually cooler and moister than the farmland patch during the day, and the primary differences in the effects of the wetland patch on the adjoining farmland occurred at night. Specifically, air temperature at night in 2008 was lower in the wetland patch than in the farmland patch, while it was higher in 2010 (Figures 4 and 6). The absolute MEE values were generally greater during 2008 than during 2010, except for specific humidity at night (Figures 5a and 5c). This indicates that the magnitude of the gradient in the microclimatic edge effect during the day could be greater in dry years than in normal years. The REE for air temperature and specific humidity was usually less during 2008 than during 2010, except for air temperature in the daytime (Figures 5b and 5d). This suggests that the extent of the influence of wetland microclimatic features on adjacent farmland is usually larger in a normal year than in a dry year.

4 Conclusions and Discussion

4.1 Horizontal Gradients of the Microclimatic Edge Effects

[24] The wetland patch cooled and moistened the adjacent farmland patch during the day, and it was warmer and moister than the farmland at night (Figure 4). This result depended greatly on the difference in the heterogeneity of the underlying surface between the wetland and farmland patches (Table 1). The wetlands were composed of lush hydrophytes and shallow water or hydric soil, which had a different heat budget than the farmlands. During the day, their higher surface reflectance led to lower net radiation than that of the farmlands; meanwhile, intensive evapotranspiration led to an increase in humidity in the surface layer [Sun and Song, 2008], and their latent heat flux was the largest consumer of incoming energy in wetlands [Priban and Ondok, 1985; Wessel and Rouse, 1994, Geiger et al., 2009]. On the other hand, the farmlands, composed of crops and bare soil, absorbed more solar radiation, leading to an increase in the proportion of sensible heat flux and soil heat flux, and they had weaker evapotranspiration, which was reflected in the higher temperatures and lower humidity there [Geiger et al., 2009]. At night, the air temperature of the wetland patch decreased slowly due to its higher heat capacity, while it decreased more rapidly in the farmland; thus, the air temperature in the wetland patch was slightly greater than in the farmland patch.

[25] Although the horizontal patterns in air temperature and specific humidity across the wetland-farmland edge were affected by both large-scale advection and turbulence [Geiger et al., 2009], we aimed to simulate the horizontal gradients in the microclimatic variables under a simple light wind condition; thus, we considered mainly the influence of atmospheric turbulence on the wetland landscape and ignored the minor effect of wind during the observations. Because of the differences between the underlying surface properties of the wetland and the farmland, heat and moisture were exchanged and diffused through turbulence, which resulted in higher rates of change in heat and water vapor across the edge, and the closer to the edge, the more intensive these processes were. The measured data in the Sanjiang Nature Reserve (Figures 4 and 6) fully demonstrate that the microclimatic variables across the wetland-farmland edge presented nonlinear and continuous sigmoid ecological gradients, as suggested by Cairns and Waldron [2003], Ewers and Didham [2006], and Hufkens et al. [2008].

[26] Nonetheless, wind and wind direction are important physical factors when studying the spatial patterns of microclimatic edge effects [Chen et al., 1995]. It is evident that the horizontal patterns and the REE are highly dependent on wind speed and wind direction [Chen et al., 1995; Pohlman et al., 2009; Wright et al., 2010]. The wind may change the sigmoid shape, resulting in an asymmetrical distribution across the wetland-farmland edge, and prevailing southeasterly winds in summer will extend the REE. Therefore, more studies are needed that sample wind speed/direction along transects across the edge to investigate the microclimatic edge effect under the influence of wind.

[27] The MEE of the mean air temperature and specific humidity at 2 m height ranged from 0.46 to 0.78°C and from 0.3 to 0.81 g kg-1, respectively (Figure 5), and the MEE of mean air temperature was less than the mean MEE in most closed forest types during the growing season [Chen et al., 1995; Didham and Lawton, 1999; Gehlhausen et al., 2000; Newmark, 2005; Newmark, 2001]. The REE ranged from 175 to 1200 m for the mean air temperature and 430 to 880 m for specific humidity (Figure 5). These values were obviously higher than those recorded at fragmented forest edges, which were generally within 10–300 m from the edge to the interior and were mostly less than 50 m [Chen et al., 1993, 1995; Gehlhausen et al., 2000; Newmark, 2001]. This disparity may be because the forest edge microclimates were influenced by shading from the forest canopy [Didham and Lawton, 1999; Gehlhausen et al., 2000; Wright et al., 2010], which reduced wind speeds, leading to less air turbulence and to the accumulation of heat. The open character of wetland edges, on the other hand, can enhance air advection and promote water vapor and heat diffusion over a longer distance [Hennenberg et al., 2008], as proved by Chen et al. [1995].

4.2 Vertical Variation in Microclimatic Edge Effects

[28] Vertically, the absolute values of MEE for air temperature and specific humidity declined, while their REE values increased with increasing height from the surface (Figure 7). This indicates that the influences of the wetland and farmland patches on microclimatic variables in the surface layer were reduced with increasing height. Specifically, the differences in water and heat fluxes between the wetland and farmland decreased with increasing height through turbulence exchange and diffusion, and then the extent of diffusion increased [Geiger et al., 2009]. At a certain height, water vapor and heat were mixed to homogenization, leading to the disappearance of the differences and a MEE close to zero (e.g., air temperature at 15 m in the nighttime). Nevertheless, we were still unable to determine other homogeneous mixed heights due to lack of data from more heights.

4.3 Temporal Variation in Microclimatic Edge Effects

[29] The diurnal patterns in the MEE and REE varied with the time of day [Chen et al., 1993, 1995; Newmark, 2001, 2005; Pohlman et al., 2009], which reflects the diurnal variation in the turbulent exchange and diffusion of water vapor and energy between the wetland and farmland (Figure 8). It depends greatly on solar angles. When solar angles tended toward zero near dawn (06:00–7:00 h) and dusk (18:00 h), the gradients in water vapor and energy disappeared, and then the MEE and REE went to zero. These two periods became the transition between the day and night patterns, as mentioned by Chen et al. [1995]; when solar angles reached a maximum in the middle of the day (13:00–14:00 h), the edge effects were greater than at other daytime periods, as mentioned by Newmark [2001, 2005] and Pohlman et al. [2009].

[30] Interannual variation in microclimatic edge effect is often associated with local climate conditions during different years. During 2008, the study area experienced a drought, with little waterlogging in the wetland, resulting in the growth of meadow vegetation (e.g., Calamagrostis angustifolia). This situation resulted in the weakening of evapotranspiration in the wetland and then a change in energy budget between the wetland and farmland patches. Thus, our data at 2 m above the surface show that the wetland patch was cooler at night in the dry year, as confirmed by Gao et al. [2002]. The results also showed that the edge effects were stronger in the dry year than in the normal year, which may be due to a greater contrast in energy and water between the wetland and farmland, similar to the results from Pohlman et al. [2009].

[31] Wetland succession is closely related to regional climate change [Poiani et al., 1996]. Previous research has found that the cold-humid microclimatic effect of wetlands in Sanjiang Plain is diminishing, as the area goes from cooler and moister to warmer and drier and as vast wetlands are reclaimed for croplands [Song et al., 2008; Yan et al., 2002]. This has led to an increase in wetland-farmland edges, which has intensified the wetland microclimatic edge effect; meanwhile, it has brought a change in the heat budget and a dramatic sensible heat exchange at the underlying surface, which has greatly improved soil moisture and heat to favor crop growth. However, with further wetland degradation, the number of wetland patches has declined, leading to a decrease in edges and a disappearance of the microclimatic edge effect. Thus, it has weakened the unique cold-humid effect of the wetland and has changed the regional ecological balance, having a negative influence on agricultural productivity.

[32] In summary, (1) air temperature and specific humidity presented a sigmoid ecological gradient in the horizontal direction under a light wind condition; wetland patches cooled and moistened the adjacent farmland patches during the day but had a warming, and still moistening, effect at night in the normal year. (2) The absolute MEE values for microclimatic variables decreased, but the REE increased with height from the surface. (3) The MEE and REE for air temperature and specific humidity varied with the time of day, both shifting at dawn and dusk when the gradients were absent. (4) The edge effect patterns changed under different local climate conditions; the wetland patches were cooler than the farmland patches at night in the dry year but warmer in the normal year.

[33] Our investigation focused on quantifying the gradient patterns in microclimatic features from the wetland interior to the neighboring farmland interior, namely by adapting two-sided microclimatic edge effects [Fonseca and Joner, 2007]. It differed from previous microclimatic edge effect studies, which have investigated mostly one-sided ecological gradients [Ries and Sisk, 2004]. The vast natural wetlands in Sanjiang Plain have been greatly transformed into a typical wetland-farmland landscape due to extensive human activities in the past few decades. Thus, we believe that studies on the spatiotemporal patterns and characteristics of the wetland microclimatic edge effect will enable further understanding of how wetland degradation affects the microclimate in neighboring farmland; at the same time, it provides a scientific basis for the management and restoration of fragmented wetlands and farmland ecosystems and the protection of biodiversity.

Acknowledgments

[34] We would like to thank Ke Li and Wen Hong for their assistance with fieldwork, and Lina Li for preliminary data processing. This work has been jointly supported by the Innovation Key Program of the CAS (KZCX2-YW-425) and the National Science and Technology Infrastructure Program (2012BAC19B05-4).