Abstract

A large portion of the lower Mississippi River alluvial valley (LMRAV) relies on irrigation from the regional alluvial aquifer for crop sustainability, which is expensive both in terms of water resources and farmer expenditures because of the large volume of water necessary to maintain crop production. As a result, knowledge of the seasonal frequency and distribution of precipitation over the LMRAV is critical for water resources management, the development of irrigation strategies, and economic planning. This project addresses the need for a detailed assessment of regional precipitation patterns through the use of rotated principal component analysis (RPCA) of high-resolution gridded radar-derived rainfall data, which provides quantification of the spatial and temporal characteristics of rainfall over the LMRAV from 1996 to 2011. Results of the project show that precipitation depths over the LMRAV are generally lower and more variable than adjacent eastern areas throughout the year, although there is substantial variability between seasons. This pattern seems to be influenced more by variations during the cool season (January–March), which has a higher overall precipitation depth and lower spatial variability than the warm season (July–September). Results further indicate that warm season rainfall is generally lower and less predictable over the LMRAV as compared to the cool season, which may be detrimental to regional water resources since irrigation planning and permitting is heavily based on seasonal rainfall predictions.

1. Introduction

Substantial soil and vegetation contrasts exist within the lower Mississippi River alluvial valley (LMRAV) because of extensive deforestation before 1940 (Fig. 1; MacDonald et al. 1979), and these regional soil and vegetation boundaries have been shown to influence local rainfall and temperature patterns through modification of the sensible and latent heat fluxes (Dyer 2011; Brown and Wax 2007; Raymond et al. 1994). This influence has been noted in other areas at various scales (Boyles et al. 2007; Pielke 2001; Chase et al. 1999; Koch and Ray 1997; Hong et al. 1995; Rabin et al. 1990; Segal et al. 1988; Mahfouf et al. 1987; Anthes 1984; Ookouchi et al. 1984; Shukla and Mintz 1982) and plays a large role in determining regional precipitation distribution and frequency, particularly during the warm season, when surface energy exchanges are maximized.

Fig. 1.

Map of vegetation and soil type over the southeastern United States. The study area is denoted by the solid white box, while the outlines of the regions used to compare precipitation inside and to the east of the Mississippi Delta are denoted by the dashed white boxes (adapted from Dyer 2011).

Fig. 1.

Map of vegetation and soil type over the southeastern United States. The study area is denoted by the solid white box, while the outlines of the regions used to compare precipitation inside and to the east of the Mississippi Delta are denoted by the dashed white boxes (adapted from Dyer 2011).

The widespread agriculture within the LMRAV makes it extremely sensitive to water resources because of heavy irrigation during the growing season (YMD 2008); therefore, precipitation variability (both depth and distribution) is of key concern to sustainable crop production and water resources. This is especially true in the warm season when biomass growth and water requirements are maximized while rainfall patterns are determined primarily by high-intensity convective storms with limited duration and extent. Since the location and duration of precipitation heavily influences water availability on the seasonal scale, it is critical that precipitation patterns over the LMRAV be quantified using detailed spatial analysis to determine the efficacy of proposed regional water management strategies.

Climate predictions show that the LMRAV will likely experience a general decrease in warm season precipitation over the coming decades, which will place additional stress on already limited regional surface and groundwater resources during the growing season (Reddy et al. 2002; Mearns et al. 2003; Parry et al. 2007). Additionally, other studies indicate that extreme precipitation events have increased in frequency in recent decades (Karl 1998; Milly et al. 2002), resulting in flash flooding and crop losses (Changnon et al. 1997; Pielke and Downtown 2000). Although these findings seem contradictory, both scenarios can occur simultaneously depending on the level of modification of local rainfall variability and the timing of these modifications (warm versus cold season); therefore, in terms of rainfall changes, of greater importance to agriculture are the potential spatial and temporal changes in precipitation variability and depth due to small-scale climatic forcings and land cover influences. Even if precipitation trends show no substantial temporal changes, a variation in spatial rainfall patterns could drastically alter the availability of water. For example, if patterns change such that rainfall does not fall over an aquifer recharge zone, that water will not be available for irrigation but will instead run off into the surface hydrologic system. Additionally, local-scale increases or decreases in precipitation intensity and/or duration could play a large role in erosional or biological processes, especially if those changes coincide with harvesting or critical crop growing phases (Rosenzweig et al. 2002).

Based on the concerns related to water resources management and precipitation, the primary objective of this research is to quantify the spatial patterns of rainfall over the LMRAV using high-resolution multisensor rainfall estimates. This information is critical to the generation of water resource planning and mitigation strategies since it will provide details regarding potential near-future water availability and rainfall probabilities. Because of the complicated feedback processes between surface and atmospheric energy and moisture fluxes, it is necessary to define historical small-scale precipitation patterns over the LMRAV to better identify specific areas where variability is highest. By finding these areas and quantifying the variability of rainfall, water resource managers will be better able to define and predict water availability in the near future, allowing for better assessment of current and proposed strategies for mitigating unsustainable water levels in the alluvial aquifer.

2. Data and methods

The study area for this project is centered on the Mississippi Delta; however, to incorporate regional precipitation patterns, especially downstream of the Mississippi Delta with respect to prevailing westerly winds, surrounding areas in Tennessee, Arkansas, Louisiana, and Alabama were included (Fig. 1). To most effectively define and quantify the local-scale patterns of rainfall over the LMRAV, multisensor precipitation estimates derived from hourly Weather Surveillance Radar-1988 Doppler (WSR-88D) data (details of the methods and limitations of the products can be found in Fulton et al. 1998) were used because of their high spatial (4 × 4 km) resolution. One drawback to the use of multisensor precipitation estimates is the short period of record (1996 to present), which is not long enough to satisfactorily define temporal trends in precipitation; however, the superior spatial resolution of the data as compared to existing gauge networks makes it more applicable for this study.

It should be noted that in 2003, the Office of Hydrologic Development (OHD) of the National Weather Service (NWS) made a transition from the Stage III processing algorithms to the updated Multisensor Precipitation Estimator (MPE) algorithm. The MPE algorithm includes an additional weighted adjustment based on surface gauge distance from a precipitation measurement, such that more weight is given to the radar estimate as the precipitation event occurs further from a rain gauge, allowing for adjustment based on within-storm variability (Westcott et al. 2005; Fulton et al. 1998; Seo 1998). Despite the fact that the algorithm used to calculate the precipitation estimates has changed during the study period for this project, no correction has been made to adjust the data since no quantification of the bias difference between MPE and Stage III is currently available. However, a bias in the data due to a change in the underlying algorithm should be spatially consistent, such that actual values of precipitation may change by some degree while the relative spatial patterns stay the same. Since this project is focused on the quantification of spatial rainfall variability and not absolute rainfall depth, this is considered a viable approach.

The Next Generation Weather Radar (NEXRAD) Stage III/MPE dataset has been used over the Mississippi Delta to quantify average rainfall depth (Dyer 2008), and Dyer (2009) verified that the high-resolution multisensor estimates over the region generally agree with existing surface observations from various networks. It should be noted that warm season estimates showed a greater difference with respect to surface-observed precipitation values than cool season estimates, which agrees with work from Wu et al. (2011), who indicated that differences between radar-estimated and surface-measured rainfall vary by season and location. However, despite these differences, the variability of the biases remain relatively low for a specific grid point during any given season; therefore, even though NEXRAD-estimated precipitation depth may show some inaccuracy with respect to surface observations, the consistent nature of the biases over the period of record makes them particularly useful for defining spatial patterns. As a result, the use of the NEXRAD multisensor rainfall product over the LMRAV is considered viable and capable of identifying and quantifying local-scale rainfall variability, frequency, and distribution. To further minimize the temporal changes in bias associated with the multisensor data while maintaining the spatial resolution, the hourly radar-estimated precipitation data were used to generate lower-temporal-resolution monthly and seasonal rainfall totals on the same 4 km × 4 km grid as the initial data. It should be noted that missing data during December 1999 and August 2000 prevented those months from being included in the analysis.

The dominant spatial precipitation patterns will be identified by performing an S-mode (Richman 1986) Varimax rotated principal component analysis (RPCA; Wilks 2011) on monthly and seasonal rainfall values. An RPCA is a data transformation procedure that identifies leading modes of variability embedded within a dataset. The RPCA method proceeds as follows.

  1. Formulate a similarity matrix (correlation matrix). The 4 km × 4 km evenly spaced gridding of the MPE dataset is ideal for such a formulation, as irregularly spaced data (e.g., gauge data) will have variability in their correlation structure based on the proximity of the stations (e.g., closer stations will be more highly correlated because of their proximity, not their statistical characteristics).

  2. Eigenanalyze the similarity matrix (Richman 1986; Wilks 2011).

  3. Compute the Varimax rotated principal component (RPC) loading matrix and principal component (PC) score matrix. In this step, it is necessary to determine the correct number of RPCs to retain. Numerous methods have been devised to determine this (e.g., North et al. 1982; Richman and Lamb 1985; Richman 1986). We used the method of congruence, which matches the RPCs against the underlying correlation matrix and determines which PCs have a strong match with the correlation matrix. Values of 0.81 or higher for the congruence coefficient (Richman and Lamb 1985) suggested a good match with the correlation matrix and thus were retained. Note that both Promax (oblique) and Varimax (orthogonal) rotations were attempted with no appreciable difference.

The RPCA method will provide generalized patterns of precipitation that represent the most common rainfall patterns at the different time scales considered (monthly and seasonal). Time series of the RPC scores will provide insight into how each pattern has changed with time, allowing the investigators to determine if the phasing of each pattern is undergoing observable changes. The result of these RPC analyses will be a series of maps that reveal regions where precipitation amount and variability has been changing over the period of record of the MPE dataset, identifying regions of additional concern for water resource managers. It should be noted that the rotation of the loading matrix in the RPCA methodology means the associated loadings are no longer ranked, such that RPC 1 does not necessarily have a higher explained variance than RPC 2, and so on.

3. Results

a. Seasonal precipitation patterns

To better understand the spatial patterns and variations of precipitation over the LMRAV, it is first necessary to define the baseline precipitation depth defined over the NEXRAD period of record. This will provide a point of comparison for interpretation of subsequent RPCA results over the region.

There are four distinct seasons over the study area based on general precipitation patterns, centered on consistent cool season [January–March (JFM)] and warm season [July–September (JAS)] precipitation depth and distribution (Figs. 2a,c). The spring and fall seasonal periods of April–June (AMJ) and October–December (OND), respectively, exhibit lower interannual consistency over the period of record than the JAS and JFM seasons (in terms of annual precipitation depth) because of the varying influence of warm and/or cold season patterns during the transition months. Although the precipitation patterns in the selected seasons may shift outside the specified months, the 3-month periods were chosen to normalize the season length and minimize the influence of transitional processes on rainfall distribution.

Fig. 2.

Mean seasonal NEXRAD-estimated precipitation for (a) JFM, (b) AMJ, (c) JAS, and (d) OND.

Fig. 2.

Mean seasonal NEXRAD-estimated precipitation for (a) JFM, (b) AMJ, (c) JAS, and (d) OND.

The precipitation pattern during JFM over the study area is characterized by lower overall precipitation amounts (<375 mm) in southeast Arkansas and areas in and to the southwest of the Mississippi Delta (Fig. 2a). This area of lower precipitation increases in extent and magnitude from November through March, at which point the lowest mean precipitation depth in the Mississippi Delta decreases from roughly 85 to 60 mm month−1 (not shown). However, there is a narrow band of increased rainfall (~360 mm) along a line roughly parallel to the eastern edge of the Mississippi Delta through central Mississippi. Although apparent during OND (Fig. 2d), this region of high mean precipitation depths along the eastern edge of the Mississippi Delta becomes more prominent through the cool season and expands during JFM from the eastern edge of the Mississippi Delta to central Alabama and southeast Mississippi, where it increases from around 125 mm month−1 in November to a maximum of 160 mm month−1 in some areas in March (not shown).

The AMJ precipitation pattern deviates from this cool season setup into a more defined latitudinal gradient of higher precipitation to the north and lower precipitation to the south (Fig. 2b). This is likely a result of the shift of midlatitude cyclone tracks to the north with the transition between the cold and warm season, such that passing storm systems produce precipitation in the northern parts of the study region before surface temperatures are warm enough to sustain a consistent sea breeze along the coast. Although lumped together with April and May rainfall patterns, June produces another sharp change in precipitation distribution as the rainfall patterns now exhibit a more “warm season” characteristic, with higher precipitation depths along the coast and less consistent rainfall depths to the north (not shown).

Although the relative magnitude of precipitation over the study region changes through the JAS period, the same general zonal pattern as the AMJ season is maintained. Besides the maximum in coastal rainfall, a notable pattern during this period is a relative minimum in precipitation within and to the west of the Mississippi Delta, extending into eastern Arkansas (Fig. 2c). Along with this minimum, there is a corresponding increase in mean rainfall depths east of the Mississippi Delta. This increase to the east produces a relatively narrow band of higher rainfall amounts along a line slightly east of the Mississippi–Alabama border, at which point rainfall amounts again begin to decrease.

Although it is difficult to precisely define the exact causes for the warm season precipitation patterns across central and northern Mississippi, Dyer (2011) suggests that soil and vegetation characteristics across the area may lead to substantial sensible and latent heat flux gradients that effectively generate and drive mesoscale circulations. These circulations produce convective instability along the eastern border of the Mississippi Delta, which eventually advect to the east before producing rainfall. This could explain the roughly north–south line of precipitation depths in east-central and north-central Mississippi during JAS, as well as the general lack of precipitation over the more agricultural areas of the Mississippi Delta and southeast Arkansas.

b. RPCA of monthly precipitation

To identify trends in the spatial variability of precipitation associated with the warm and cool season, maps of RPC loadings and their associated RPC score time series are provided. Note that the product of an RPC score time series value and the RPC loading map yields units of standard anomalies. For interpretation purposes, a positive RPC score means that positive/negative anomalies are related to higher/lower precipitation, while a negative RPC score means that negative/positive anomalies are related to higher/lower precipitation. Thus, interpreting RPC loading maps without knowledge of the associated RPC score values provides no insight about local maxima or minima. It is important to note that trends associated with RPC scores are not associated with the robustness of the associated RPC loading map; they simply demonstrate changes to the RPC loading maps over time. That is, a nonsignificant RPC score time series does not suggest a meaningless RPC loading map, but instead that the RPC pattern is stable over time.

A congruence test of the RPC loadings for the monthly summed multisensor precipitation data over the study area revealed that the primary modes of variability within the precipitation signal are encompassed by the first four RPCs, which constituted a total of 37.6% of the total variance in precipitation (Fig. 3). The relatively low explained variance is likely a result of large changes in precipitation patterns between seasons and years, producing noise in the data that cannot be differentiated by monthly rainfall aggregations. Subsequent seasonal analysis should minimize this by focusing the RPCs on time periods with similar precipitation variability; however, the monthly analysis is necessary to provide a baseline for comparison to the seasonal analyses.

Fig. 3.

(a)–(d) Monthly PCs and related score time series for RPCs 1–4. Percent explained variance for each RPC is given below the respective map, and the p value for each score trend is given in the associated time series (trends significant at the 90% level are denoted by an asterisk).

Fig. 3.

(a)–(d) Monthly PCs and related score time series for RPCs 1–4. Percent explained variance for each RPC is given below the respective map, and the p value for each score trend is given in the associated time series (trends significant at the 90% level are denoted by an asterisk).

RPC 1 (Fig. 3a) shows a defined positive anomaly in precipitation in south-central Mississippi with a negative anomaly in Arkansas, while RPC 2 (Fig. 3b) shows a general negative anomaly in precipitation over much of the western and southern extent of the study area, with an area of positive anomaly over north-central Alabama and east-central Mississippi. The RPC score time series for RPC 2 indicates a significant increase (p < 0.10) throughout the study period, despite the relatively short period of record used for analysis, suggesting an increase of precipitation over time in central and northern Alabama; however, the large variation in the score time series around a zero value indicates that although the positive anomaly may be increasing in magnitude (i.e., higher rainfall), there is still a high level of variability in the overall pattern.

The third and fourth RPCs (Figs. 3c,d) show distinct areas of negative precipitation anomalies over west-central Mississippi and northeast Mississippi, respectively, extending into neighboring states, with the strongest positive anomalies in southern Alabama and Mississippi. However, the high variability in the score time series indicates that these patterns are not consistent, with substantial changes in anomaly magnitude and direction between seasons. Based on the roughly southwest to northeast orientation of the anomalies shown in RPCs 1, 3, and 4, it is likely that the RPC patterns are a result of changes in location and intensity of regional moisture convergence associated with synoptic-scale frontal systems, which would influence precipitation distribution and intensity. Because of the inherent spatial and temporal variability in synoptic-scale storm tracks and the associated mesoscale convective precipitation patterns over the southeast United States, especially during the spring and autumn transition seasons, the high variability in the score time series seems logical.

To further investigate the influence of possible seasonal storm track variability on the RPCA anomalies, the frequency of occurrence of anomalies greater than or less than 1 and −1, respectively, was calculated for each RPC score time series (Fig. 4). RPC score values of ±1 represent the actual pattern displayed on the RPC loading maps, although values of −1 represent the opposite to the anomaly patterns. Results show that the greatest frequency of positive anomalies for RPC 1 (Fig. 4a) and RPC 3 (Fig. 4c) was during the summer, while high-magnitude negative anomalies occurred primarily during the winter. This would indicate that the positive precipitation anomalies are associated with summertime convection, while negative anomalies are associated with synoptic-scale frontal systems, as mentioned above. The frequency of high-magnitude anomalies for RPC 2 (Fig. 4b) is actually opposite this pattern, such that positive anomalies are most frequent in the cool season and vice versa. However, the overall monthly frequencies are generally less than those for RPCs 1 and 3, indicating a more stable and consistent pattern. The frequency of high-magnitude events for RPC 4 (Fig. 4d) is also low, although there is a distinct peak of negative anomalies during the spring. Because of the influence of frontal precipitation in the southeast United States during this time of year, it is reasonable to suspect that synoptic-scale patterns play a large role in the rainfall patterns. Based on the anomaly map associated with RPC 4 (Fig. 3d), it seems that higher rainfall is more likely in the northern reaches of the study area during the spring, which could be related to the occurrence of stationary frontal systems.

Fig. 4.

Frequency of occurrence of months with precipitation anomalies greater than one (solid lines) or less than negative one (dashed lines) for (a) RPC 1, (b) RPC 2, (c) RPC 3, and (4) RPC 4.

Fig. 4.

Frequency of occurrence of months with precipitation anomalies greater than one (solid lines) or less than negative one (dashed lines) for (a) RPC 1, (b) RPC 2, (c) RPC 3, and (4) RPC 4.

c. RPCA of seasonal precipitation

To more effectively differentiate the seasonal patterns from one another within the RPCA, the winter (JFM) and summer (JAS) seasons were investigated independently. Also, since these seasons had more well-defined precipitation patterns than the fall (OND) and spring (AMJ) transition seasons, it is reasonable to figure that the associated RPCAs will produce the most stable results. In a physical sense, focusing on the summer allows for an interpretation of precipitation trends during the late growing season over the study area, which is when irrigation is usually applied to crops because of the hydrologic stress on the land surface, while a focus on winter is warranted because of the importance of rainfall during this time period on aquifer recharge.

The RPCA of cool season precipitation revealed four dominant RPCs that explained 61.3% of the total precipitation variability. The first cool season RPC (Fig. 5a) indicates a region of positive anomalies through northwest Mississippi and lower anomalies along the Gulf Coast and southern Alabama, while RPC 2 (Fig. 5b) shows a similar overall pattern, except the dominant feature is a strong negative anomaly in central Alabama and east-central Mississippi. This agrees with the observed seasonal rainfall patterns for the cool season (Fig. 2a). Based on these spatial patterns and the fact that the score time series for both RPCs follow the same significantly decreasing trend (p < 0.10), the two RPCs show a strong gradient of precipitation depth from northwest Mississippi through central Alabama. This indicates that rainfall over the Mississippi Delta and areas to the west in Arkansas and Louisiana is persistently lower than areas to the east and that the negative precipitation anomalies are increasing in magnitude over the period of record.

Fig. 5.

As in Fig. 3, but for JFM PCs.

Fig. 5.

As in Fig. 3, but for JFM PCs.

The lower rainfall over the Mississippi Delta is of particular importance since this area receives the least amount of precipitation based on monthly totals (Fig. 2a), meaning that the region could potentially be sensitive to cool season precipitation deficits and subsequent hydrologic stress during the early growing season. Since this area has a strong dependence on agriculture for economic sustainability, the lack of water resources early in the warm season through decreased aquifer levels could be a potentially critical situation.

RPC 3, which explains 13.4% of the total variance in winter precipitation patterns, shows a positive anomaly in the southern reaches of the study area, especially along the Alabama and Mississippi Gulf Coast, while the northern half of the study area is generally characterized with a negative anomaly (Fig. 5c). Based on the precipitation pattern shown, it is likely that this RPC is describing rainfall resulting from coastal influences to the south, such as convective rainfall produced by southeast flow during warmer wintertime conditions.

The fourth RPC (Fig. 5d) is similar to RPC 1 in that the greatest positive anomalies in precipitation depth are in the northwest area of the study period, through central Arkansas. However, the magnitudes of the anomalies are generally lower than RPC 1 and RPC 2. Because of the roughly northwest to southeast orientation of the gradient in precipitation anomaly over the study region, along with a dominant region of positive anomalies in Arkansas and a score time series with no significant slope, this RPC is indicative of normal cool season synoptic-scale rainfall associated with midlatitude cyclones passing to the north of the study area.

An RPCA of warm season precipitation yielded four viable RPCs explaining a total of 71.0% of the variability in precipitation over the study area. RPC 1 (Fig. 6a) is characterized by an area of negative anomalies through northern Mississippi and Alabama and into western Tennessee, while RPC 2 (Fig. 6b) shows a large area with a positive anomaly over southeast Arkansas and northern Louisiana. The score time series for RPCs 1 and 2 oscillate around a zero value and show no significant trend (p < 0.10), which means that the spatial precipitation anomalies are relatively consistent, and in the case of RPC 1, generally of a low magnitude. The exception is 2007, which was characterized by extremely high rainfall during July (3–5 times the climatological average) in northern Louisiana, southern Arkansas, and west-central Mississippi (not shown).

Fig. 6.

As in Fig. 3, but for JAS PCs.

Fig. 6.

As in Fig. 3, but for JAS PCs.

Since most of the precipitation in this region during the warm season is convective, it is logical that mechanisms related to convective initiation and precipitation generation (i.e., moisture convergence, isentropic forcing, etc.) should be the primary drivers for variations in rainfall patterns. Additionally, given the fact that the anomalies for RPCs 1 and 2 cover a relatively large area, the underlying mechanisms are likely based on synoptic-scale mechanisms. Because of the proximity of the region to the Gulf of Mexico, it is reasonable to speculate that the anomalies in rainfall in RPCs 1 and 2 are related to a shift in the moisture flux, such that a westward shift of the low-level moisture would lead to a decrease in rainfall over north Mississippi and an increase in Arkansas, and vice versa. RPC 3 shows a high-magnitude negative anomaly in southwest Mississippi (Fig. 6c), which is more indicative of the spring transition rainfall pattern (Fig. 2b) than of summertime rainfall patterns (Fig. 2c). Since frontal systems associated with extratropical cyclones produce a large percentage of the precipitation in this region during the early and late part of the season, especially along the northern reaches of the study area that are closer to the jet stream, a northward shift in the respective storm tracks could lead to a decrease in rainfall in the northern regions of the study area, as shown by this anomaly pattern.

RPC 4 (Fig. 6d), which explains 12.9% of the variance in precipitation over the region, is most notable because of the close agreement between the anomalies and the mean warm season precipitation depth (Fig. 2c). Based on this pattern, there is a general area of negative precipitation anomalies over western Mississippi, centered on the Mississippi Delta, with a line of positive anomalies just west of the Mississippi–Alabama border. The rainfall distribution given by this RPC agrees with the summertime precipitation pattern described in Dyer (2011), whereby convection along the eastern boundary of the Mississippi Delta resulting from vegetation contrasts between cultivated land to the west and forested land to the east generates a surface-based mesoscale convective boundary. Since this only occurs during synoptically benign conditions, it is reasonable that this pattern would only explain a limited percentage of the total variability in rainfall over the region. However, the fact that this pattern is driven by anthropogenic modification of the land surface warrants further research, especially considering the stability of the pattern over time as seen by the lack of a slope in the associated score time series with generally positive (although low) magnitude values in most years.

An important aspect of the data that must be addressed with respect to the seasonal precipitation analysis is the apparent increase in variability after 2003 in nearly all of the retained RPC score time series. As mentioned previously, there was a transition in 2003 from the Stage III algorithm to the MPE algorithm that included, among other modifications, a change in the weighting factor associated with each gauge; therefore, it is expected that the rainfall totals would be influenced as well, although there is not a long enough data record to fully quantify the associated bias. However, as shown by the various score time series, the spatial variability in the data seems to be suppressed prior to 2003, after which there are several high-magnitude jumps in the RPC scores. Despite the change in variability in the time series, though, in most cases the overall slope remains consistent. This indicates that the spatial rainfall patterns given by the RPCs remain valid.

To address more fully the potential physical atmospheric mechanisms driving the precipitation anomalies described by the RPCAs for the warm and cool season, the meteorological characteristics associated with the seasons with the highest-magnitude anomalies were studied. This was done by calculating the difference in atmospheric features related to precipitation generation (mean sea level pressure, 700-hPa specific humidity, and 850–500-hPa temperature difference) between the seasons with the highest and the lowest anomalies. For example, for the cool season RPC 1 (Fig. 5a), the difference between the 2006 and 2009 seasons was calculated to diagnose the primary differences in atmospheric patterns between positive and negative anomalies. Although there are a number of additional variables that could be analyzed, these values reflect pressure patterns, lower atmospheric moisture content, and midlevel lapse rates, respectively, which provide an overview of the major dynamic and thermodynamic features related to rainfall generation over the study region. Only the first two RPCs for the warm and cold season were analyzed since they had the most explained variance.

The difference in atmospheric characteristics between the greatest positive and negative anomalies for the cool season RPCs (Figs. 7a,b) show that the difference in midlevel temperature lapse rates are small (less than 0.5 K), indicating that thermal instability is not a strong factor in precipitation generation. This is reasonable considering the importance of synoptic-scale frontal systems during this period, which is shown by the relatively large differences in mean sea level pressure over the study region (greater than 1.5 hPa). However, the greater influence in precipitation patterns seems to focus on differences in low-level moisture, which shows relatively well-defined differences that relate to the spatial location of the precipitation anomalies.

Fig. 7.

Difference in seasonal 700-hPa specific humidity (g kg−1; shaded), mean sea level pressure (hPa; solid black lines), and 850–500-hPa temperature (K; dashed white lines) between seasons with maximum and minimum anomalies for (a) cool season RPC 1, (b) cool season RPC 2, (c) warm season RPC 1, and (d) warm season RPC 2. Note that mean sea level pressure and 850–500-hPa temperature have consistent contour intervals between figures while the 700-hPa specific humidity shaded contours change for each figure.

Fig. 7.

Difference in seasonal 700-hPa specific humidity (g kg−1; shaded), mean sea level pressure (hPa; solid black lines), and 850–500-hPa temperature (K; dashed white lines) between seasons with maximum and minimum anomalies for (a) cool season RPC 1, (b) cool season RPC 2, (c) warm season RPC 1, and (d) warm season RPC 2. Note that mean sea level pressure and 850–500-hPa temperature have consistent contour intervals between figures while the 700-hPa specific humidity shaded contours change for each figure.

For cool season RPC 1 (Fig. 7a), there is generally more moisture through central Mississippi and areas to the west during the highest-magnitude positive anomaly, associated with increased precipitation in northwest Mississippi. This, along with lower mean sea level pressure relative to the highest-magnitude negative anomaly, suggests that lower pressure and increased moisture from the west lead to higher precipitation over the lower Mississippi River valley, while higher pressure and increased moisture through central Alabama leads to an eastward shift of precipitation maxima. Cool season RPC 2 (Fig. 7b), which shows a maximum negative precipitation anomaly through central Alabama, also seems to be influenced by a change in low-level moisture as seen by a smaller difference in specific humidity from the Gulf Coast along the Mississippi–Alabama border northeast through Alabama. The relatively large differences in mean sea level pressure further indicate that the change in moisture flux may be a result of variations in synoptic-scale features.

The atmospheric characteristics associated with differences in the warm season precipitation anomalies tend to focus on thermal and moisture characteristics, such that differences in mean sea level pressure over the study region are generally less than 1 hPa, while differences in 700-hPa specific humidity and 850–500-hPa temperatures show a noticeably larger range than during the cool season (Figs. 7c,d). For warm season RPC 1, the highest-magnitude positive anomaly is associated with a distinct area of increased low-level moisture from the Mississippi Gulf Coast northward, which means that the highest-magnitude negative anomaly has less moisture for precipitation generation over the area (Fig. 7c). For the negative precipitation anomaly, the greatest low-level moisture is shown to be in eastern Texas and the southern Appalachians through North Carolina and Virginia. Additionally, 850–500-hPa temperature differences indicate that midlevel lapse rates are lower during the negative anomaly period, suggesting less thermal instability and a possible decrease in convection through northern Alabama and Mississippi.

Atmospheric patterns associated with warm season RPC 2 show a strong zonal gradient of low-level moisture and midlevel lapse rates through the study area, with the highest-magnitude negative anomaly having more moisture and higher lapse rates through northern Louisiana and southern Arkansas (Fig. 7d). This seems counterintuitive considering the negative precipitation anomaly is associated with decreased rainfall over this area; however, the increased moisture flux through Alabama associated with the highest-magnitude positive anomaly, along with higher pressure through the northern study area, suggests that variations in the moisture flux from the Gulf of Mexico play a key role in convective development and associated rainfall for this pattern.

d. Local-scale rainfall differences over the Mississippi Delta

To quantify the differences in precipitation variability over the Mississippi Delta and surrounding regions, two equal-sized subregions within the study domain were defined. The first region was selected to include the Mississippi delta region (western region within Fig. 1), while the second domain was selected to be directly east (and thus synoptically downstream) of the Delta (eastern region within Fig. 1). To assess variability within these two subregions, pairwise bootstrap resampling (1000 replicates; Efron and Tibshirani 1993) of standard deviation and means of all grid points within each subregion was conducted for both the cool and warm seasons. Bootstrap resampling is advantageous as it allows for the comparison of multiple years simultaneously and removes any constraints associated with differing statistical distributions among the two samples. The 2.5th and 97.5th percentiles of the bootstrap replicates were taken to be the upper and lower confidence levels on the bootstrapped statistic, and these confidence intervals, in addition to the medians of each variable, were plotted against each other to determine how the Mississippi Delta region varies in relation to the adjacent eastern area. It should be noted that the warm season mean precipitation was not calculated for 2000 because of missing precipitation data in August of that year; therefore, to keep the methods consistent for both seasons, the cool season precipitation for 2000 was not included in the analysis.

The cool season bootstrap mean precipitation values between the Mississippi Delta and adjacent eastern region share the same overall pattern, although precipitation in the eastern region tends to be slightly higher than over the Mississippi Delta (Fig. 8a). This is especially true during periods when there is minimal interannual variability, such as from 2008 to 2011, although the exact causes for these differences cannot be clearly defined. Also, the 95% confidence limits remain extremely close to the annual values, indicating minimal variability in the bootstrap results. In general, the bootstrap mean precipitation over the Mississippi Delta region has a baseline of approximately 225–275 mm yr−1, although two peaks in 2002 and 2006 show substantial variance from this value. Over the adjacent eastern region, however, the bootstrap mean rainfall is more variable with no noticeable baseline value. It is often above 300 mm yr−1, although 2007 and the first few years of the study period show values closer to 200 mm yr−1.

Fig. 8.

JFM bootstrap (a) mean and (b) standard deviation values for each year for the Mississippi Delta domain (solid line) and adjacent eastern domain (dotted line). The error bars represent the 95% confident limits for each year.

Fig. 8.

JFM bootstrap (a) mean and (b) standard deviation values for each year for the Mississippi Delta domain (solid line) and adjacent eastern domain (dotted line). The error bars represent the 95% confident limits for each year.

The cool season bootstrap standard deviation values over the Mississippi Delta show a relatively consistent interannual value of roughly 50 mm yr−1, with a peak in 2006 of just under 100 mm yr−1 (Fig. 8b). This peak in 2006 is associated with a peak in bootstrap mean precipitation; however, the larger peak in bootstrap mean precipitation in 2002 does not show a corresponding peak in bootstrap standard deviation. The 2006 feature is due to a localized maximum in precipitation along the western edge of the Mississippi Delta region (not shown), which caused mean rainfall to increase while generating a large variance in precipitation depths across the domain. In 2002, however, the increase in rainfall was relatively evenly distributed across the Mississippi Delta domain (not shown), leading to the spike in mean rainfall with no associated increase in variability.

Although the bootstrap standard deviation rainfall values in the Mississippi Delta region during the cool season remain relatively consistent, the values over the adjacent eastern domain show considerable variability between years (Fig. 8b). The pattern shows spikes in variability in 2002, 2006, and 2009, which as stated above, are only mirrored over the Mississippi Delta domain in 2006. This implies that although the overall total precipitation depth over the two regions follows the same general patterns, the spatial distribution over the two regions may be different. In other words, when cool season precipitation over the Mississippi Delta is anomalously high, the overall spatial distribution remains largely unchanged with respect to years with average precipitation; however, over the adjacent eastern region, when mean seasonal precipitation is anomalously higher, there are certain areas where rainfall preferentially increases. Analysis of rainfall patterns during these specific time periods shows that the areas of increased precipitation vary by year (not shown), such that there is no discernible spatial consistency to where the rainfall occurs. With a longer study period it may be possible to more clearly ascertain if this pattern is statistically consistent or just a result of noise in the data; therefore, further investigation of this pattern is left as a topic of future research when more data are available.

As with the cool season bootstrap mean precipitation, the bootstrap mean warm season precipitation over the Mississippi Delta region is generally lower than the adjacent eastern region from 1996 to 2006; however, the pattern switches for a few years (2007–08) before returning back to the initial orientation (Fig. 9a). The difference in precipitation between the two study areas is roughly 50–100 mm for each year, despite substantial fluctuations in the mean precipitation in 2002 and 2009. Additionally, there is a fairly consistent baseline of rainfall within the two regions over the study period, such that precipitation in the Mississippi Delta stays above approximately 160 mm yr−1, while to the east the values stay above roughly 250 mm yr−1.

Fig. 9.

As in Fig. 8, but for JAS.

Fig. 9.

As in Fig. 8, but for JAS.

The variability of the precipitation over the two subregions, as defined by the annual bootstrap standard deviation values (Fig. 9b), indicates that the variability of the warm season rainfall is approximately 80 mm yr−1, although there are substantial spikes in 2002 and 1999 over the Mississippi Delta and adjacent eastern region, respectively. This implies that either the Mississippi Delta or adjacent eastern domain can expect a minimum depth of rainfall of approximately 200 mm yr−1 over the entire domain; however, this does not describe the specific spatial location of the rainfall, such that an even distribution of rainfall is just as likely as a highly variable distribution in both magnitude and location.

The most noticeable and important features of the seasonal bootstrap analysis of precipitation mean and standard deviation are the higher precipitation depths during the cool season, the more consistent mean rainfall values, and the lower (and also more consistent) standard deviations. This is logical considering the more convective nature of precipitation during the warm season and the dominance of thermodynamically driven mesoscale convective boundaries. However, for water resources management and planning, this suggests that seasonal warm season rainfall forecasts will be less reliable because of higher precipitation variability, with a low probability of accurate spatial forecasts because of variations in distribution within the region.

4. Conclusions

The LMRAV is characterized by widespread agriculture due to fertile soils, a relatively long growing season, and abundant rainfall; however, unequal distribution of rainfall throughout the year, especially in the warm season because of the convective nature of precipitation, can lead to substantial water stress on crops. As a result, irrigation from the regional alluvial aquifer is common, leading to additional stress on groundwater resources and increasing the importance of aquifer recharge. Since local-scale precipitation variability can lead to drastic changes in water availability in surface and subsurface hydrologic systems over the region, this paper works to quantify the spatial patterns of rainfall over the LMRAV to better understand current and future water resource issues and needs.

Mean seasonal precipitation values indicate that there is a clear differentiation between rainfall patterns within the alluvial valley and areas to the east and south, although the causes of these patterns likely change by season. During JAS, surface features such as land cover could play a role in determining the location and magnitude of rainfall since air mass convection is the dominant forcing mechanism for precipitation generation. As a result, thermodynamically driven boundaries such as the sea breeze along the Gulf Coast and circulations resulting from discontinuities in vegetation and soils along the alluvial valley boundary in northwest Mississippi (Dyer 2011) cause rainfall to have a regional minimum within the alluvial valley (Fig. 2c). During JFM there is also a minimum in precipitation over the alluvial valley, although there is more spatial variability than during the warm season (Fig. 2a). This is likely a combination of thermodynamically driven processes and dynamic midlatitude processes (i.e., frontal passages), although the latter may be the dominant precipitation generation mechanism because of the proximity of the jet stream to the study area during this time.

The rotated principal component analysis (RPCA) of monthly precipitation over the study area shows several dominant modes of variability, most of which are likely related to variations in the location and intensity of synoptic-scale frontal systems because of the general southwest to northeast orientation of the anomalies (Fig. 3). The high variability and general mean of zero of the associated score times series indicates substantial inconsistency in the spatial rainfall patterns, which further justifies the influence of synoptic-scale systems on rainfall patterns. In general, the spatial patterns given by the RPCs show positive rainfall anomalies in northeastern Mississippi and northern Alabama and negative anomalies over southern Arkansas, northern Louisiana, and west-central Mississippi. The only score time series with a significant trend (p < 0.10; Fig. 3b) indicates an increase in magnitude of a positive precipitation anomaly in north-central Alabama and a decrease in areas to the east and along the Gulf Coast. Despite the positive trend, there remains substantial variability around a zero anomaly.

An RPCA of seasonal rainfall shows that JFM rainfall is increasing through central Alabama and decreasing in northwestern Mississippi, based on the first two RPCs that explain 38.7% of the total cool season rainfall variability (Figs. 5a,b). The score time series associated with these RPCs have significant trends (p < 0.10), indicating that the magnitude of the precipitation anomalies is increasing over the 1996–2012 study period. The remaining two RPCs from the cool season RPCA show positive rainfall anomalies along the Gulf Coast and central Arkansas, although the related score time series show a consistent oscillation around a value of zero that suggests relatively weak and stable anomaly patterns (Figs. 5c,d).

The RPCA results for JAS show considerable spatial variability, with no significant trend in any of the associated score time series (Fig. 6). The spatial variability is likely associated with the predominately convective nature of rainfall during the warm months, which is often driven by surface boundaries such as the sea breeze along the Gulf Coast or land cover changes near urban areas and forest–agriculture transitions. However, high-magnitude anomalies in the RPC score time series are usually associated with increased moisture convergence related to synoptic-scale flow; therefore, the overall spatial patterns still contain a general northwest-to-southeast gradient indicative of frontal passages through the region.

Analysis of the general atmospheric characteristics associated with the highest-magnitude positive and negative anomalies for the warm and cool season RPCs show that variations in precipitation patterns during the cool season are more closely associated with synoptic-scale flow (as indicated by differences in mean sea level pressure), while variations during the warm season are influenced primarily by moisture flux and midlevel lapse rates (Fig. 7). This result is reasonable and does allow for some conjecture regarding the general conditions necessary for the precipitation anomalies to be realized; however, it should be noted that these results can only be used in a general sense because of the smoothed nature of the seasonal data and the inability to completely isolate the patterns related to a single RPC. Because of the importance of recognizing the meteorological characteristics associated with anomalously high and low precipitation, especially for seasonal climate assessments, future work will include a more detailed analysis of the mechanisms related to the rainfall patterns shown by the RPCA over the southeast United States.

A local-scale assessment of the RPCA results shows that precipitation in and adjacent to the Mississippi River alluvial valley (also known as the Mississippi Delta) is predominately associated with cool season patterns, while rainfall trends and variability in the warm season over the study period are more indicative of changes in frontal precipitation and coastal processes. Because of the importance of rainfall trends within and adjacent to the alluvial valley in terms of water resources, mean precipitation depth and variance over the Mississippi Delta and the adjacent eastern area were analyzed using a bootstrap methodology. It was shown that precipitation patterns were generally similar over the period of record both within the alluvial valley in northwest Mississippi and the adjacent eastern area (Fig. 1), although rainfall was consistently higher to the east than over the alluvial valley (Figs. 8a, 9a). Also, despite the relatively large amount of variability between years, seasonal rainfall depths were generally higher during the cool season than the warm season.

Regarding the variability of rainfall over the two regions, there was substantial deviation in rainfall amounts over each of the regions, especially during the warm season (Fig. 9b). The lowest variability occurred within the alluvial valley region during the cool season (Fig. 8b), which is important considering that is also when most of the precipitation occurs. This implies that cool season rainfall over the LMRAV is higher and more spatially consistent than during other seasons or locations within the study area, leading to a greater potential for viable seasonal forecasts.

Acknowledgments

The authors thank the Mississippi Water Resources Research Institute and the U.S. Geological Survey for funding to complete this research through Award G11AP20088. Additionally, the authors thank the anonymous reviewers for their invaluable comments and suggestions to the original manuscript.

REFERENCES

REFERENCES
Anthes
,
R. A.
,
1984
:
Enhancement of convective precipitation by mesoscale variations in vegetative covering in semiarid regions
.
J. Climate Appl. Meteor.
,
23
,
541
554
, doi:10.1175/1520-0450(1984)023<0541:EOCPBM>2.0.CO;2.
Boyles
,
R.
,
S.
Raman
, and
A.
Sims
,
2007
: Sensitivity of mesoscale surface dynamics to surface soil and vegetation contrasts over the Carolina sandhills. Pure Appl. Geophys.,164, 1547–1576, doi:10.1007/s00024-007-0227-2.
Brown
,
M. E.
, and
C. L.
Wax
,
2007
:
Temperature as an indicator of landform influence on atmospheric processes
.
Phys. Geog.
,
28
,
148
157
, doi:10.2747/0272-3646.28.2.148.
Changnon
,
S. A.
,
D.
Changnon
,
E. R.
Fosse
,
D. C.
Hoganson
,
R. J.
Roth
, and
J. M.
Totsch
,
1997
:
Effects of recent weather extremes on the insurance industry: Major implications for the atmospheric sciences
.
Bull. Amer. Meteor. Soc.
,
78
,
425
431
, doi:10.1175/1520-0477(1997)078<0425:EORWEO>2.0.CO;2.
Chase
,
T. N.
,
R. A.
Pielke
,
T. G. F.
Kittel
,
J. S.
Baron
, and
T. J.
Stohlgren
,
1999
:
Potential impacts on Colorado Rocky Mountain weather due to land use changes on the adjacent Great Plains
.
J. Geophys. Res.
,
104
,
16 673
16 690
, doi:10.1029/1999JD900118.
Dyer
,
J. L.
,
2008
:
Basin-scale precipitation analysis for southeast U. S. watersheds using high-resolution radar precipitation estimates
.
Phys. Geogr.
,
29
,
320
340
, doi:10.2747/0272-3646.29.4.320.
Dyer
,
J. L.
,
2009
:
Evaluation of surface and radar estimated precipitation data sources over the lower Mississippi River alluvial plain
.
Phys. Geogr.
,
30
,
430
452
, doi:10.2747/0272-3646.30.5.430.
Dyer
,
J. L.
,
2011
:
Analysis of a warm-season surface-influenced mesoscale convective boundary in northwest Mississippi
.
J. Hydrometeor.
,
12
,
1007
1023
, doi:10.1175/2011JHM1326.1.
Efron
,
B.
, and
R. J.
Tibshirani
,
1993
: An Introduction to the Bootstrap. Chapman and Hall/CRC, 436 pp.
Fulton
,
R. A.
,
J. P.
Breidenback
,
D.-J.
Seo
,
D. A.
Miller
, and
T.
O'Bannon
,
1998
:
The WSR-88D rainfall algorithm
.
Wea. Forecasting
,
13
,
377
395
, doi:10.1175/1520-0434(1998)013<0377:TWRA>2.0.CO;2.
Hong
,
S. Y.
,
M. J.
Leach
, and
S.
Raman
,
1995
:
Role of vegetation in generation of mesoscale circulation
.
Atmos. Environ.
,
29
,
2163
2176
, doi:10.1016/1352-2310(94)00241-C.
Karl
,
T. R.
,
1998
:
Secular trends of precipitation amount, frequency, and intensity in the USA
.
Bull. Amer. Meteor. Soc.
,
79
,
231
241
, doi:10.1175/1520-0477(1998)079<0231:STOPAF>2.0.CO;2.
Koch
,
S. E.
, and
K. A.
Ray
,
1997
:
Mesoanalysis of summertime convective zones in central and eastern North Carolina
.
Wea. Forecasting
,
12
,
56
77
, doi:10.1175/1520-0434(1997)012<0056:MOSCZI>2.0.CO;2.
MacDonald
,
P. O.
,
W. E.
Frayer
, and
J. K.
Clauser
,
1979
: Documentation, Chronology and Future Projections of Bottomland Hardwood Habitat Losses in the Lower Mississippi Alluvial Plain. 2 vols. U.S. Fish and Wildlife Service, 135 pp.
Mahfouf
,
J.-F.
,
E.
Richard
, and
P.
Mascart
,
1987
:
The influence of soil and vegetation on the development of mesoscale circulations
.
J. Climate Appl. Meteor.
,
26
,
1483
1495
, doi:10.1175/1520-0450(1987)026<1483:TIOSAV>2.0.CO;2.
Mearns
,
L. O.
,
F.
Giorgi
,
L.
McDaniel
, and
C.
Shields
,
2003
:
Climate scenarios for the southeastern US based on GCM and regional model simulation
.
Climatic Change
,
60
,
7
35
, doi:10.1023/A:1026033732707.
Milly
,
P. C. D.
,
R. T.
Wetherald
,
K. A.
Dunne
, and
T. L.
Delworth
,
2002
:
Increasing risk of great floods in a changing climate
.
Nature
,
415
,
514
517
, doi:10.1038/415514a.
North
,
G. R.
,
T. L.
Bell
,
R. F.
Calahan
, and
F. J.
Moeng
,
1982
:
Sampling errors in the estimation of empirical orthogonal functions
.
Mon. Wea. Rev.
,
110
,
699
706
, doi:10.1175/1520-0493(1982)110<0699:SEITEO>2.0.CO;2.
Ookouchi
,
Y.
,
M.
Segal
,
R. C.
Kessler
, and
R. A.
Pielke
,
1984
:
Evaluation of soil moisture effects on the generation and modification of mesoscale circulations
.
Mon. Wea. Rev.
,
112
,
2281
2292
, doi:10.1175/1520-0493(1984)112<2281:EOSMEO>2.0.CO;2.
Parry
,
M. L.
,
O. F.
Canziani
,
J. P.
Palutikof
,
P. J.
van der Linden
, and
C. E.
Hanson
, Eds.,
2007
: Impacts, Adaptation and Vulnerability. Cambridge University Press, 976 pp.
Pielke
,
R. A.
,
2001
:
Influence of the spatial distribution of vegetation and soils on the prediction of cumulus convective rainfall
.
Rev. Geophys.
,
39
,
151
177
, doi:10.1029/1999RG000072.
Pielke
,
R. A.
, and
M. W.
Downtown
,
2000
:
Precipitation and damaging floods: Trends in the United States, 1932–1997
.
J. Climate
,
13
,
3625
3637
, doi:10.1175/1520-0442(2000)013<3625:PADFTI>2.0.CO;2.
Rabin
,
R. M.
,
S.
Stadler
,
P. J.
Wetzel
,
D. J.
Stensrud
, and
M.
Gregory
,
1990
:
Observed effects of landscape variability on convective clouds
.
Bull. Amer. Meteor. Soc.
,
71
,
272
280
, doi:10.1175/1520-0477(1990)071<0272:OEOLVO>2.0.CO;2.
Raymond
,
W. H.
,
R. M.
Rabin
, and
G. S.
Wade
,
1994
:
Evidence of an agricultural heat island in the lower Mississippi River floodplain
.
Bull. Amer. Meteor. Soc.
,
75
,
1019
1025
, doi:10.1175/1520-0477(1994)075<1019:EOAAHI>2.0.CO;2.
Reddy
,
K. R.
,
P. R.
Doma
,
L. O.
Mearns
,
M.
Boone
,
H.
Hodges
,
A.
Richardson
, and
V.
Kakani
,
2002
:
Simulating the impacts of climate change on cotton production in the Mississippi Delta
.
Climate Res.
,
22
,
271
281
, doi:10.3354/cr022271.
Richman
,
M. B.
,
1986
:
Rotation of principal components
.
J. Climatol.
,
6
,
293
335
, doi:10.1002/joc.3370060305.
Richman
,
M. B.
, and
P. J.
Lamb
,
1985
:
Climatic pattern analysis of three- and seven-day summer rainfall in the central United States: Some methodological considerations and a regionalization
.
J. Climate Appl. Meteor.
,
24
,
1325
1343
, doi:10.1175/1520-0450(1985)024<1325:CPAOTA>2.0.CO;2.
Rosenzweig
,
C.
,
F. N.
Tubiello
,
R.
Goldberg
,
E.
Mills
, and
J.
Bloomfield
,
2002
:
Increased crop damage in the US from excess precipitation under climate change
.
Global Environ. Change
,
12
,
197
202
, doi:10.1016/S0959-3780(02)00008-0.
Segal
,
M.
,
R.
Avissar
,
M. C.
McCumber
, and
R. A.
Pielke
,
1988
:
Evaluation of vegetation effects on the generation and modification of mesoscale circulations
.
J. Atmos. Sci.
,
45
,
2268
2293
, doi:10.1175/1520-0469(1988)045<2268:EOVEOT>2.0.CO;2.
Seo
,
D.-J.
,
1998
:
Real-time estimation of rainfall fields using radar rainfall and rain gauge data
.
J. Hydrol.
,
208
,
25
36
, doi:10.1016/S0022-1694(98)00140-1.
Shukla
,
J.
, and
Y.
Mintz
,
1982
:
Influence of land-surface evapotranspiration on the earth's climate
.
Science
,
215
,
1498
1501
, doi:10.1126/science.215.4539.1498.
Westcott
,
E. N.
,
S. E.
Hollinger
, and
K. E.
Kunkel
,
2005
:
Use of real-time multisensor data to assess the relationship of normalized corn yield with monthly rainfall and heat stress across the central United States
.
J. Appl. Meteor.
,
44
,
1667
1676
, doi:10.1175/JAM2303.1.
Wilks
,
D. S.
,
2011
: Statistical Methods in the Atmospheric Sciences. 3rd ed., Elsevier, 676 pp.
Wu
,
W.
,
D.
Kitzmiller
, and
S.
Wu
,
2011
: Evaluation of radar precipitation estimates from the National Mosaic and Multisensor Quantitative Precipitation Estimation System and the WSR-88D precipitation processing system over the conterminous United States. J. Hydrometeor.,13, 1080–1093, doi:10.1175/JHM-D-11-064.1.
YMD
,
2008
: 2008 annual report. Yazoo Mississippi Delta Joint Water Management District, 100 pp. [Available online at http://www.ymd.org/pdfs/2008annualreportfinal.pdf.]