| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Dep. of Soil Science, Univ. of Saskatchewan, Agriculture Building, 51 Campus Dr., Saskatoon, SK, S7N 5A8 Canada
* Corresponding author (yates{at}sask.usask.ca)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: CV, concave • CW, cultivated wetland • CX, convex • MDCD, minimal detectable concentration difference • UW, uncultivated wetland • WFPS, water-filled pore space
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Nitrous oxide emission data is commonly observed to be highly skewed (Parkin, 1987; Corre et al., 1996). The high degree of skewness in soil N2O emission data is thought to be the result of the patchy distribution of denitrification in anaerobic microsites (Parkin, 1987). Corre et al. (1996) observed that the occurrence of anaerobic microsites or hotspots was landscape controlled and that this knowledge may be useful in the development of spatial models designed to estimate N2O emissions at a landscape level. Previous work by Yates et al. (2006) has demonstrated that the extreme flux values are part of an event-based/background emission pattern model as proposed by Brumme et al. (1999). In the study of Yates et al. (2006), high fluxes during an emission event were associated with specific locations in hummocky, agricultural landscape. However, during the transition from an event-based pattern to a background pattern, the magnitude, and number of extreme values changed and these changes altered the probability distribution of soil N2O emission data. The shape of the probability distribution, over multiple flux measurements, ranged from a highly skewed, reverse J-shaped through log-normal to a symmetrical or nearly normal shape. Yates et al. (2006) also observed that extreme values may be highly localized. This limits the ability of spatial analytical tools that rely on the assumption of stationarity (i.e., that the value of that soil property at each location is an estimate of the same mean and variance; see Trangmar et al. (1985), of a normally distributed field of values) to estimate the scale of the processes controlling soil N2O emission. To improve our understanding of the spatial scales of soil N2O emission and their change during and following a N2O emission event, it is necessary to use statistical tools that do not rely on the assumption of stationarity.
Wavelet analysis is a statistical technique that meets this requirement (Si, 2003). Wavelets provide location-specific information on scales and variance allowing a detailed study of nonstationary processes (Si and Farrell, 2004). Wavelets have been used to analyze spatial scales of soil properties (Si and Farrell, 2004) and specifically N2O emission from soil cores (Lark et al., 2004). However, the use of wavelets for scale analysis of soil N2O flux data, which has been collected from undisturbed field locations on multiple occasions, has, to the best of our knowledge, not been reported.
The first objective of this study was to use wavelet analysis to examine the distribution of localized variance for successive samplings of soil N2O flux from a hummocky, agricultural landscape. The second objective was to identify the important scales of variability of the processes controlling soil N2O emission and assess the change in these scales over the snowmelt period.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Locations in the cultivated area of the transect were classified as either convex (CX), concave (CV), or cultivated wetland (CW). Convex elements were topographically high positions with a positive profile curvature. Concave elements were positions with negative profile curvature, adjacent to wetlands. Cultivated wetlands were depressional positions, roughly circular in shape, which temporally collect rain or snowmelt water. The mean bulk density of the soils at CX positions was 1.292 ± 0.104 g cm–3, the mean bulk density at CV and CW positions was slightly lower (1.236 ± 0.136 g cm–3). Non-agricultural portions of the transect included vegetated depressions and were classified as uncultivated wetlands (UW). Mean bulk density in the UW elements was lower than that of the surrounding cultivated units (0.845 ± 0.289). Mean soil bulk densities are taken from unpublished data based on a minimum of 12 to 42 samples in the vegetated depressions and a minimum of 84 to 140 samples in the cultivated area of the site.
Soil types ranged from thin Typic Calciborolls on the top of CX elements through thicker Typic Haploborolls down-slope from CX to CV elements. Albic Argiborolls and Argic Cryaquolls were found in CW and UW elements. Haplic Calciborolls were found in CX elements as well as some CV elements. Soil profiles were described in the field, classified, and converted to the U.S. taxonomic system using the Canadian System of Soil Classification (Agriculture and Agri-Food Canada, 1998b). Overall, soil textures ranged from loam at CX elements to silt loam in CV and CW elements. The hydrological pattern of the site was such that precipitation and snowmelt generated runoff concentrated in the numerous depressions where the bulk of water infiltrated laterally through the wetland margins and was subsequently lost to evaporation and transpiration (Hayashi et al., 1998).
In May of 2004 the east side of the site, including the area of the transect, was seeded to grass by Ducks Unlimited Canada. The mix consisted of Agropyron elongatum (Host) Beauvois, Agropyron intermedium (Host) Beauvois, Bromus biebersteinii Roem. and Schult., Elymus dauricus Turcz. exgriseb., Festuca rubra L., Onobrychis viciifolia Scop., Elymus Canadensis L., Agropyron trachycaulum (Link) Malte and Medicago sativa L. Establishment of the grass cover was gradual and the change in management was not expected to have a significant impact on the N2O flux, as it would have been measured from the field in a cultivated state, until after the period of data collection reported here. However, to distinguish this transect from other site activities, this transect is referred to as the seeded-grass transect.
In total, N2O fluxes were measured on all transect points 18 times between June 2003 and 2005. It is not possible to address all 18 data sets due to space limitations and instead five consecutive transect samplings have been selected as a representative subset to illustrate the main relationships. The first of the five samplings occurred on 30 Mar. 2004. At this time the transect had a snow cover of approximately 80%. A snow survey conducted on February 24 found a mean snow depth of 18.4 cm and a mean snow water equivalent of 45.6 mm for the transect area. By the next sampling date, April 4, the snow cover was reduced to 5% and the transect was snow free by April 29, the third date. The transect was sampled again in the summer, June 3 and June 23.
Soil Nitrous Oxide Flux Measurements
Soil N2O flux was measured at each location along the seeded-grass transect using a two-piece, closed, vented chamber (International Atomic Energy Agency, 1992) consisting of a polyvinyl chloride (PVC) ring base and vented cap with sampling port. Vent tube design was similar to that proposed by Hutchinson and Mosier (1981). Before the start of data collection, the bases were pressed into the soil and secured using 20-cm spikes where they remained for the duration of the field season. At each sampling event, the chamber was placed onto the base and sealed by a rubber ring within the cap. When attached to the base, the head space of each chamber was 2.25 L and covered a soil surface area of 0.02 m2. Samples of the headspace gas were drawn from the chamber at time zero (t0) and at 8, 16, and 24 min, with a 20-mL syringe and injected into 12-mL Labco Exetainer (Labco Limited, UK) evacuated tubes for transport back to the laboratory. Gas sampling for each time interval was completed by a single individual moving from location to location every 45 s. This allowed for a complete transect sampling in <2 h. Gas sampling was timed to begin at midday. Samples were returned to the lab, placed in cold storage and analyzed in 1 to 2 wk.
Nitrous oxide concentrations were determined using a Varian CP3800 GC (Varian Canada Inc., Mississauga, ON) equipped with dual electron capture detectors (ECD). The injector temperature = 100°C, column temperature = 35°C, detector temperature = 370°C; separations were performed using Poraplot Q columns (12.5 m by 0.32 mm i.d. fused silica capillary column, DF (film thickness) = 8 µm; includes a 2.5-m particle trap) with ultra high purity He (14.4 mL min–1) as the carrier gas and P5 (95:5 v/v Ar/CH4 mix) as the make-up gas (12.0 mL min–1). The system was calibrated using standard gases (N2O in N2) obtained from PraxAir (Mississauga, ON). Internal calibration curves were obtained by applying linear, least squares regression to the gas concentration (ppbV N2O) versus peak area data; N2O concentrations in the headspace samples were then calculated automatically from the regression equations.
Samples of ambient air were included in each analytical run as reference samples to check the ‘within run’ precision, calculate the minimum detectable concentration difference (MDCD), and correct for detector drift. The MDCD was calculated by (i) analyzing matched pairs of the reference gas samples at regular intervals during each analytical run; (ii) calculating the average difference between sample pairs (µp) as well as the standard deviation (
p); and (iii) calculating the MDCD using Eq. [1].
 | [1] |
The vertical flux of N2O at the soil–atmosphere interface (ng N2O-N m–2 s–1) was calculated as the slope of the tangent to the concentration (ng L–1) vs. time (min) curve at t0. That is, the flux at t0 was calculated as the first derivative of the second-order polynomial equation (y = at2 + bt + c) used to describe the concentration vs. time relationship. The flux (ng N L–1 min–1) was then converted to an area basis by multiplying by the chamber volume/(surface area x 60 s min–1). In cases where ‘rogue’ data points prevented the use of the second-order polynomial model, the flux was calculated as the slope of the linear, least-squares regression equation that best described the concentration vs. time relationship (Hutchinson and Mosier, 1981). Our preference to use a polynomial equation to describe the concentration vs. time relationship was based on Anthony et al. (1995) who concluded that the linear model may represent a serious source of measurement bias. Our experience with flux measurements in the Canadian Prairie region indicated that concentration vs. time relationship for soil N2O flux is usually nonlinear.
Wavelet Analysis
The use of wavelets in this study was as a tool of analysis and is not meant to be an exercise in wavelet analysis per se. This intention is reflected in the following basic explanation of the wavelet approach used in this study. For a comprehensive and mathematical understanding of wavelets see Kumar and Foufoula-Georgiou (1997). Si (2003) and Si and Farrell (2004) give an explanation of wavelets as applied to scale analysis of soil properties.
A wavelet is a mathematical function representing a small wave form. It is considered small because it is contained within a finite domain (Graps, 1995) in contrast to the sine and cosine functions used in Fourier analysis, which stretch out to infinity (i.e., have infinite span and are globally uniform in time (Lau and Weng, 1995). Thus, where Fourier analysis does a poor job in approximating sharp spikes in data—like that typical of soil N2O emissions, wavelet analysis is well suited.
A wavelet transform takes a data series that represents a physical process and, like the Fourier transform, decomposes the variance of this process into series of coefficients representing the distribution of the variance across different frequencies (scales) and space (location) or time (Percival, 1995; Lindsay et al., 1996). Unlike the Fourier transform, the spatial context of the data is retained.
Scale analysis using wavelets is performed by contraction and dilation of the wavelet function. Contraction and dilation changes the size of the interval in time or space (window) that the wavelet function is applied to in the data series. Increasing window size increases the scale at which the coefficient for a particular location is calculated. In this manner a wavelet coefficient is obtained for each location in a data series over a range of scales and thus, the scale of analysis is matched to the size of the feature(s) in the data series.
There are various types of wavelet transforms and generally are grouped as continuous or discrete (Si and Zeleke, 2005). Several types of discrete wavelets are orthogonal so there is no redundant information between wavelet coefficients and scales are discrete. This makes discrete wavelets very useful for data compression. Continuous wavelets are not orthogonal therefore, there is overlap between wavelet coefficients resulting in redundant information between scales and locations and making the continuous wavelet very useful for scale analysis (Si, 2003). The mother wavelet is rescaled by powers of two; hence, a data series that contains a binary number of points is optimal for computation speed (Si, 2003).
There are several types of wavelets (see Kumar and Foufoula-Georgiou, 1997) and it is important that one chooses a wavelet that is appropriate for the objectives of the proposed analysis. The wavelet chosen for this study was the Mexican Hat, which is a continuous wavelet that is sensitive to peaks in data (Si, 2003), such as one would expect in soil N2O emission. The Mexican Hat (Eq. [2]), a second derivative of a Gaussian function,
 | [2] |
() is the Gamma function.
In the continuous wavelet transform, after Si and Farrell (2004), y(x) is a spatial series with y representing the measurement and x representing the location (Eq. [3]).
 | [3] |
,
) is the wavelet coefficient that is the integral of the product of y(x) and 
,(x) as of function of the scaling parameter () where
 | [4] |
). The spatial translation () is the factor that shifts the function [,(x)] from one location (x) to another along the spatial series to obtain a wavelet coefficient for each location. Thus, a wavelet coefficient is calculated for location at each scale. The continuous wavelet transform is implemented using fast Fourier transform. The local wavelet spectrum is obtained by squaring the wavelet coefficient for each scale at each location and is plotted as a two-dimensional field of values of wavelet variance. Edge effects were controlled by zero padding of the data set to 256 from 128 points. Edge effects are the result of the start and end of the data set acting similar to an abrupt change in value, which can influence the coefficients determined for points away from the ends. This effect also tends to increase with increase in scale. The Mexican hat wavelet is also a good choice in this regard because it is less susceptible to edge effects because it is narrow and has fewer oscillations than other wavelets (Si and Farrell, 2004).
To test the significance of the local wavelet spectrum we used a permutation test (Pardo-Igúzquiza and Rodríquez-Tovar, 2000) that compared the global wavelet spectrum for each original data series against the global wavelet spectrum for multiple reordering of the data series. The premise of a permutation test is that if a spatial pattern exists in a data series then a random reordering of the series will destroy this pattern. If the reordering of the series is performed multiple times then the chance of recreating the data series with the original spatial pattern can be determined. In this case, the null hypothesis is that all scales in the power spectrum of the original series will not be significantly different from the power spectra, below a predetermined significance level, of the multiple random realizations of the series. A significance level (5%) can be applied by sorting the power spectra of the multiple random realizations into increasing order and using the highest power spectrum for the first 95% of these spectra as the level of power that is required to be surpassed to reject the null hypothesis.
The global wavelet spectrum is the average of local wavelet spectra [W(,)] over all locations of a spatial series and can be written as:
 | [5] |
The use of consecutive pairs instead of individual data points incorporates the concept of red noise into the permutation test. Red noise is a measure of the background variability that is inherent to soil landscapes because of the short-range dependency between adjacent locations (autoregressive process of 1 lag) that occurs simply because of their close proximity (Torrence and Compo, 1998) and resulting sedimentary, hydrological, and biological communication. As the distance between locations increases (increasing scale), the variance of the soil property increases, like that of red noise (Si and Farrell, 2004). The question is whether the increase in variance is due to the imposition of order on the distribution of soil properties by processes such as, for example, topographic variation, or is the increase in variance over increasing scale due simply to the red noise? Using individual data pairs would be assumed that background variation is completely random (white noise) and that there is no autocorrelation between soil properties. Other significance tests of the local wavelet power spectrum require comparison of Gaussian white or red noise with the wavelet spectrum (Si and Zeleke 2005; Si and Farrell, 2004). This assumes normally distributed data, which is an invalid assumption in our case. The permutations test is nonparametric (Pardo-Igúzquiza and Rodríquez-Tovar, 2005).
The local wavelet spectrum is presented as a contour map of wavelet coefficients that represent highs and lows in variance and which are plotted according to the distance (m) along the transect and the scale (m) of the variation. Darker shades are associated with higher wavelet variance. Each spectrum has been contoured with the same intervals and shades. The interpretation of the wavelet spectrum is visual. Peak in wavelet variance are represented by contours that form concentric patterns, like a bull's eye or target. Features of interest are areas of high variance that appear as repetitive peaks in the contours such that they form a series along the x axis of the figure within a narrow scale range (a global event) or as a single, nonrepetitive high in variance at a discrete location on the x axis (localized event) (Si and Farrell, 2004). A series of repeated features is an indication of cyclic behavior and the period of that cycle is the scale of the underlying process. Localized features do not cycle and therefore cannot be assigned to a particular scale. The scale range assigned to a series of repetitive features is that along which the centers of these features have a tendency to line up. Thus, a global feature consisting of a series of repetitive highs in wavelet variance may have peaks that occur at various scales (44, 53, 59 m, for example) however they will be described as a global feature that is controlled by an underlying process that cycles over a 40- to 60-m range in scale.
The global wavelet spectrum for each sampling date to be discussed was determined and compared with the 95% significance level that was also determined for the same date (data not shown). The result was that on some days not all the scales presented in the local spectra were significant. On those spectra for which all the scales were not significant, the nonsignificant scales, and the anomalies included in those scales, were indicated in the figure caption. The scales (m) that are indicated as nonsignificant in the local spectrum were scales at which the global wavelet variance was equal to or less than the 95% significance level (i.e., the null hypothesis could not be rejected). The purpose of this was to be clear on the significance of the scales, but allow the reader to view the anomalies because, although nonsignificant, they may still be of interest.
Measurement of Soil Moisture and Soil Temperature
Time domain reflectometry (TDR), after Topp and Ferre (2002), was used at each location on each sampling date (except March 30) to measure soil moisture content over the 0- to 15-cm depth. On March 30 frozen ground prevented use of the TDR. Readings were taken manually using a Tektronix 1502 B cable tester (Tektronix, Wilsonville, OR). The soil temperature was measured at the 5- and 20-cm depths using buried T type thermocouples made with twisted wire pairs of copper and constantan (45% Ni and 55% Cu) and read with a Barnant DuaLogR thermocouple reader (Barnant Co., Barrington, IL).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
On April 4, individual fluxes increased by an order of magnitude or more at several locations (Fig. 2 ). Fluxes ranged from 0.0 to 510.9 ng N2O-N m–2 s–1. Fluxes were low in all CW and UW elements. High and extreme flux values were associated with CV elements. At approximately 145 and 320 m, the flux was highest. The local wavelet spectrum showed a broad, discontinuous band of features that peaked within a scale range of 30 to 50 m. At 320 m there was an intense anomaly in wavelet variance that cuts across nearly the full scale range. This was a local feature that was associated with the high flux value at that location. The permutation test (data not shown) indicated that scales < 30 and > 65 m were not significant (Fig. 2).
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The probability distribution of the N2O flux data and the mean N2O flux are both a product of the control that landform position and land use have on the distribution of flux measurements along the transect. This can be clearly observed in the local wavelet spectrum. On March 30, fluxes are measured from the snow surface and the small areas of exposed soil at CX elements where snow cover was thinnest and first to melt away. High variance features in the spectrum were associated with peaks in N2O flux at the exposed CX elements and the low or negative N2O fluxes in-between. This repetition of N2O flux pattern, due to the CX elements resulted in the cyclic pattern observed in the wavelet spectrum. The exposed areas on the CX elements were more numerous in the middle portion of the transect (175–300 m), thus there were higher fluxes in this portion of the transect resulting in the large variation observed in the local wavelet spectrum at a scales of 80 to 100 m. Compared with later dates, the flux values for March 30 were low (mean value 2.3 ± 4.8 ng N2O-N m–2 s–1) but there was a high number of locations at which a flux was measurable (Table 1).
By April 4, the majority of the transect was snow free and snowmelt water had concentrated in the CW and UW elements resulting in mean WFPS as high as 100% (Table 2). Mean N2O fluxes increased by an order of magnitude (Table 1) and a shift in the position of the peak fluxes from the CX elements down-slope to the CV elements occurred (Fig. 2). Extreme values in the data set also increased in magnitude resulting in a greater range in measured N2O flux values, severely skewing the data and the changing the probability distribution shape from a log-normal to a reverse J-shaped (Table 1). The distribution of variance (according to the local wavelet spectrum) showed periodicity at a scale of 30 to 50 m associated with both the low fluxes from the flooded CW and UW elements and the high to extreme flux values from the CV elements (Fig. 2). The spectrum also showed the influence of the extreme value at 320 m which appears as a nonrepetitive variance anomaly (i.e., localized feature).
The effect of a localization of variance is taken to the extreme on April 29. Overall fluxes decreased with the exception of the large CW element near the middle of the transect (Fig. 3). Here the soil was wettest (Table 2), probably because snowmelt runoff was retained here the longest, while at other locations the soil had dried. The result was a concentration of locations emitting a flux at the same order of magnitude as the previous date, although the activity along the rest of the transect was much lower. The concentration of variance in this one location overshadowed the cyclic pattern of low level of activity present along the majority of the transect.
The wavelet spectrum for June 3 was similar to that of March 30. On the June 3 wavelet spectrum of soil N2O flux on the transect indicated three scales of spatial variation, as they were on March 30th (Fig. 4). As well, there are no strong localized features present. The highs in variance on June 3 were at a scale of 40 to 60 m and were associated with peaks in flux that occurred primarily in CW and UW elements, although mean flux had dropped by one order of magnitude since the last sampling date (Table 1) and these highs in variance were nonsignificant. The lack of localized features indicates that the extreme values needed to skew the distribution to a reverse J-shape were absent. Hence, the distribution had returned to a log-normal, as was the data for March 30. The large-scale, significant, variation evident in the spectrum appears to be related to the high overall fluxes in the middle portion of the transect; however on this date the highest fluxes were from the CW elements and not the CX elements as was the case on March 30.
Between June 3 and June 23 soil N2O flux continued to decrease. The weak cyclic pattern in the <25-m scale range reflected the alternation along the transect between positive and negative fluxes and had no direct connection to the landscape elements, hence its nonsignificance. The low variance, random spatial pattern on this day is confirmed by a symmetrical distribution of flux values about zero and a net negative mean flux.
The local wavelet spectra demonstrated that landscape position can be an important control on the spatial variation in soil N2O emission. Landform imposed a periodicity to the distribution of the fluxes and controlled the occurrence of the nonstationary, localized features that were strong contributors to the mean N2O flux magnitude and variance. The local wavelet spectra also demonstrated that a certain type landscape position determines the spatial variation of N2O flux on a specific date, but the type of position important to the spatial variation will change between dates. For example, through the first four sampling dates, the landscape positions of importance changed in the order: CX, CV, CW, and UW/CW.
There is a definite down slope trend in the above order that indicated the importance of the redistribution of water in the landscape during spring snowmelt. Wind scouring during winter removes snow from high areas between depressions (catchment boundaries) where CX elements are found (Hayashi et al., 1998). Thin snow cover on CX elements lead to first exposure of soil at these positions. The air temperature at the Saskatoon Airport for March 30 ranged from –1.4 to 16.8°C (Environment Canada, 2005). Thus, snowmelt was proceeding, limited by cool temperatures, and the soil was still frozen at the 5-cm depth. Under these conditions a thin, wet, active layer formed on the soil surface because the snowmelt water was prevented from infiltrating, and created and the conditions for denitrification. As snowmelt accelerated and completed, water flowed into CW and UW elements resulting in flooded soils with a high WFPS, but the water surface prevented major fluxes from occurring at these positions; however, exposed, soils with a WFPS high enough to promote denitrification were possible at the CV elements on the margins of the pooled water. As the water receded at CW positions, the exposed soils were temporarily at a WFPS conducive to denitrification while soils at the topographically higher CV and CX positions had already dried and flux had diminished.
It is clear that the topography did control the timing and distribution of the major fluxes during the spring period at the St. Denis site; however, this trend does not explain the relatively low fluxes from CX elements on April 4 or June 3 when the mean WFPS for that element was high (Table 2). Although there is no data to support this, our assumption is that the available N at CX elements was limiting. Hydrolyzable N (Gianello and Bremner, 1986) is considered to be a measure the labile N pool and has been found to be lowest in soils on topographically high positions in hummocky terrains (Walley et al., 2002).
In addition to the control that the redistribution of water had on the spatial pattern of the N2O flux was the control that soil temperature and land use had on the spatial pattern. Frozen soil can act as a barrier to the infiltration of water if soil pores are blocked with ice, as observed in agricultural soils by Granger et al. (1984). In our study, the reduced drainage maintained a high WFPS and allowed formation of the thin, active layer at CX positions that lead to the March 30 spatial pattern of fluxes. The absence of major fluxes in the UW elements suggests that soils under long-term vegetated growth may not experience extreme flux events, perhaps due to high porosity and better soil aeration. As described above, soils in UW have much lower bulk densities. It is also possible that available N may have limited N2O emissions from UW elements as it may have at CX elements. Perhaps the perennial growth in UW elements results in differences in N cycling between cultivated and noncultivated elements.
If higher soil porosity and better soil aeration as well as reduced available N is indeed responsible for the absence of major fluxes in the UW elements, then this has implications in regards to best management practices (BMPs) that might be employed to reduce soil N2O emissions produced from agricultural activities in these types of terrains. Our results that removing CW elements from cultivation and allowing some form of restoration to a land use represented by the UW element is a logical management practice that would reduce yearly emissions. The assumption is that such a change in land use would promote an increase in soil porosity at locations within former CW elements improving soil aeration and minimizing the peak in N2O flux that occurs at these locations in the spring. However, the degree of restoration required, a range from establishment of a permanent grass cover with periodic forage removal to no equipment activity with an introduction of tree species, is unknown. Also unknown is the time required under such a new management practice to see significant reductions in emissions. These unknowns open the door to new and valuable field research.
The role of landform was critical to the event-based emission pattern present in this terrain. The "event" was the spring snowmelt that began just before March 30. Landform controlled the pattern of emissions from the outset. As melt water began to concentrate in the landscape the high N2O fluxes that created the event-based pattern developed quickly and were markedly displayed in the wavelet spectra for April 4 and April 29. As melt water evaporated from and infiltrated into the soil, moisture conditions necessary for the extreme fluxes produced by the event diminished. The spatial effects of the event decayed toward a background emission pattern. Thus the event-based pattern itself evolved, and the course of this evolution was dependent on factors such as distribution of the snow pack (i.e., which locations are exposed first), redistribution of snowmelt and resulting soil moisture conditions, land use, and bulk density.
Although this pattern evolves, there is some continuity from date to date in the scales of variation associated with soil N2O emission. Although not always significant, the scale of variation that is related to the landscape ranged between 20 and 60 m across the five data sets depending on the landscape element controlling the highest fluxes. This is interesting because Velthof et al. (2000) found temporal stability in the spatial pattern of soil N2O flux in the short term (4 d), but observed that the range of spatial dependence increased over the duration of their study.
The short- and long-range variations were not always evident or significant in the wavelet spectrum due to the influence of localized features on the distribution of variance. The localized features were associated with different landscape elements on different dates, but their occurrence was of shorter duration than the cyclic pattern of emissions. The localized features essentially represented a pattern of emission that was super-imposed on the background emission pattern because of the snowmelt event.
The local wavelet spectra demonstrate that in a hummocky topography, such as at the St. Denis site, we may add spatial context to the event-based, background emission model proposed by Brumme et al. (1999) and this has implication for measurement of fluxes in these terrains for inventory and up-scaling purposes. The temporally stable, landscape controlled cyclic pattern of emissions could be monitored using a sampling design with a minimum number of locations based on the spatial scale. The temporary, localized pattern of emissions could be monitored by stratifying the landscape into elements that capture a range of CV and CW units and sampling a randomly chosen subset to monitor with particular attention to the critical spring snowmelt period. This may work because the spring flux event was not random, but occurred at specific elements in the landscape during a specific period of time. Certainly this research would need to be confirmed by repeat findings in other sites with hummocky topography.
The control of landscape position on the spatial variation of soil N2O emission and the change in spatial variation over time was tied to the control that landform had on the redistribution of water. Wavelet analysis provided us with spatial information on the cyclic and nonstationary processes controlling soil N2O over a period of active water redistribution and demonstrated that this technique can be applied to the relationship between landscape position, soil moisture and soil N2O emission. As well, the location and impact on peak fluxes of the variability in N2O emissions is clearly illustrated allowing for the development of management strategies that may reduce the emissions at peak times.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The results suggest that the distribution of fluxes in, general, have implications regarding best management practices designed to reduce soil N2O emissions, and that the cyclic and noncyclic pattern of emissions has implications for measurement of soil N2O in these landscapes. These are directions that should be taken in their own right with the data from St. Denis and future research projects.
|
ACKNOWLEDGMENTS |
|---|
Received for publication August 5, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  B. C. Si Spatial Scaling Analyses of Soil Physical Properties: A Review of Spectral and Wavelet Methods Vadose Zone J., May 27, 2008; 7(2): 547 - 562. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
