934 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 69 The Usefulness of Piecewise Potential Vorticity Inversion BJØRN RØSTING Norwegian Meteorological Institute, Oslo, Norway JÓN EGILL KRISTJÁNSSON Department of Geosciences, University of Oslo, Oslo, Norway (Manuscript received 29 April 2011, in final form 10 October 2011) ABSTRACT It is today widely accepted that potential vorticity (PV) thinking is a highly useful approach for understanding important aspects of dynamic meteorology and for validation of output from state-of-the-art numerical weather prediction (NWP) models. Egger recently presented a critical view on piecewise potential vorticity inversion (PPVI). This was done by defining a PV anomaly by retaining the observed PV field in a specific region, while changing the observed PV fields to zero elsewhere. Inversion of such a modified PV field yields a flow vastly different from the observed. On the basis of this result it was argued that PPVI is useless for understanding the dynamics of the flow. The present paper argues that the results presented by Egger are incomplete in the context of PPVI, since the complementary cases were not considered and that the results also depend on the idealized model formulations. The complementary case is defined by changing the observed PV to zero in the specific region, while retaining the observed PV field elsewhere. By including the complementary cases, it can be demonstrated that the streamfunction fields associated with the PV and boundary temperature anomalies presented by Egger add up to yield the observed streamfunction field, as expected if PPVI is to be valid. It follows that PPVI is indeed valid and useful in these cases. 1. Introduction Hoskins et al. (1985) provided a theoretical survey of the concept of potential vorticity (PV), along with applications of PV and low-level boundary temperature anomalies to the interpretation of atmospheric dynamics. Although the PV aspect of atmospheric dynamics had been well known for several decades, having been described by Rossby (1940), Ertel (1942), and Charney (1947), the paper by Hoskins et al. introduced, most likely for the first time, the concept of ‘‘PV thinking.’’ PV thinking means adopting PV in order to describe and understand the dynamics of the atmosphere, such as baroclinic and barotropic instability. This is achieved, for instance, through PV inversion, by which geopotential heights, winds, and temperature fields are Corresponding author address: Bjørn Røsting, Norwegian Meteorological Institute, Forecasting Division, POB 43, Blindern, 0313 Oslo, Norway. E-mail: bjorn.rosting@met.no DOI: 10.1175/JAS-D-11-0115.1 Ó 2012 American Meteorological Society obtained, provided that suitable balance and boundary conditions are specified. The classical approach to baroclinic instability, described by Eady (1949), is essentially based on the PV concept, in which the (quasigeostrophic) PV field confined between an upper lid (i.e., the tropopause) and the surface is specified, along with the temperature distribution at lower and upper boundaries. Analytical solutions of this problem yield growing waves with the largest growth rate for wavelengths on the order of 4000–6000 km. PV can also be applied in identifying errors in numerical weather prediction (NWP) analyses and simulations (e.g., Mansfield 1996). This is achieved by comparing the PV fields retrieved from NWP models with features observed in water vapor (WV) images. This method is based on the strong absorption of radiation in the 6.3–6.7-mm water vapor bands; hence PV-rich stratospheric air is easily identified in WV images (e.g., Santurette and Georgiev 2005). When errors in the PV fields are identified by an observed mismatch between PV fields and WV features, it is possible to manually correct the PV MARCH 2012 935 R Ø S T I N G A N D K R I S T J Á N S S O N fields. The corrected PV fields are then inverted, yielding a modified numerical analysis that is balanced and suitable for initializing numerical reruns (e.g., Demirtas and Thorpe 1999; Manders et al. 2007; Røsting et al. 2003; Røsting and Kristjánsson 2006, 2008; Verkley et al. 2005). These procedures represent an efficient use of PV thinking. An important aspect of PV thinking is piecewise PV inversion (PPVI). In PPVI, the PV field is partitioned, and the contributions of the different parts of the PV field to the flow are quantified by piecewise inversion (Davis 1992; Kristjánsson et al. 1999). The feasibility of such an approach for Ertel’s nonlinear PV was first demonstrated by Davis and Emanuel (1991). Criticism of PPVI pertaining to its usefulness for understanding atmospheric dynamics was recently put forward by Egger (2008, hereafter E2008). Although some of the arguments concerning PPVI in E2008 may be pertinent, particularly the discussion of inverted tendency fields, others call for comments. The criticism advanced in E2008 is based on experiments to be discussed in section 3 below, in which the observed PV is replaced by PV 5 0 in a selected region. Inversion of such a modified PV field yields a flow different from the observed flow. In E2008 it was claimed that such a result implies that PPVI in general is useless for understanding the dynamics of the flow. Below, we present evidence that this drastic claim is unjustified. Methven and de Vries (2008) addressed the limitations of the mathematical model describing the baroclinic cases of E2008 and pointed out the importance of dynamic evolution rather than pure diagnosis based on PPVI in assessing the benefits of PV thinking. This paper is structured as follows. In section 2 we give a brief presentation of the PV inversion technique used in many case studies. In section 3 we address some of the cases presented in E2008. A case study based on PPVI is presented in section 4, and discussion and conclusions are provided in section 5. We now recapture the PPVI technique, since knowledge of this technique is essential for the subsequent discussion. In general, studies requiring PV inversion adopt the Ertel PV: 1 h $u, r Dq ›q 1 1 [ 1 u $q 5 h $u_ 1 $ 3 Fr $u, Dt ›t r r (1) where r is density, h is the absolute vorticity vector, and $u is the three-dimensional gradient of potential temperature. PV is expressed in PV units (PVU; 1 PVU 5 1026 m2 s21 K kg21). (2) where u is the wind vector, Fr is the friction force, and u denotes diabatic heating. The mathematical expression for the Ertel PV can be simplified by using isentropic coordinates (x, y, u): ›u , q 5 (zu 1 f ) 2g ›p (3) where zu is relative vorticity on an isentropic surface and f is the Coriolis parameter (planetary vorticity), g is the acceleration of gravity, and 2›u/›p is an expression of static stability. Relations (2) and (3) show that for adiabatic and inviscid flow q is conserved for an air parcel moving on an isentropic surface. Potential temperature fields ub are specified at the lower and upper boundaries (e.g., at 950 and 150 hPa) and the ub anomalies can be regarded as PV anomalies; that is, a warm (cold) anomaly at the lower boundary can be regarded as a positive (negative) PV anomaly. Likewise, a warm (cold) anomaly at the upper boundary can be treated as a negative (positive) PV anomaly (Hoskins et al. 1985). Hence such anomalies can be inverted separately. The total PV and boundary potential temperature fields are decomposed into mean and perturbation terms. Hence the total PV is expressed as q 5 q 1 q9 (4) and the expression for the boundary potential temperature field is analogous: ub 5 ub 1 ub9 . 2. Piecewise PV inversion q5 PV is changed because of diabatic effects and friction as shown in the following equation: (5) The mean fields q and ub are temporal or spatial means, with temporal means taken as an average over, for example, 48 h. The perturbation terms in Eqs. (4) and (5) may be decomposed into several anomalies. Usually one is interested in specific PV and ub anomalies that appear to have a large influence on the phenomena of interest. Such anomalies are usually clearly defined as temporally and spatially coherent features. PV anomalies that are far away from the region of interest or difficult to define temporally are generally defined as parts of a residual term, denoted by qres. 936 JOURNAL OF THE ATMOSPHERIC SCIENCES Following Kristjánsson et al. (1999), the perturbation field in Eq. (4) is expressed as n q9 5 å qi , (6) i51 where qi denotes the anomalies. For N selected PV anomalies the above relation becomes N q9 5 å qi 1 qres , (7) i51 where the residual term is n qres 5 å i5N11 qi . (8) The expressions for the ub fields are similar. PPVI deals with obtaining uniquely the familiar fields of geopotential height or streamfunction, winds, and temperature distribution associated with specific PV and ub anomalies. To close the set of equations required for PV inversion a balance condition is required, as well as lateral and horizontal boundary conditions. We adopt Charney’s balance condition (Charney 1955) given by ›(›C/›x, ›C/›y) , = 2 F 5 $ ( f $C) 1 2 ›(x, y) (9) where F is the geopotential and C is the streamfunction. The reason why Ertel’s PV Eq. (1) and the Charney balance condition [Eq. (9)] are often preferred in PV inversion studies is that the flow is properly described by Eqs. (2) and (9), including flow with fairly large Rossby numbers (e.g., Ro ; 1), allowing for strongly curved flow. The disadvantage of adopting these equations directly is their nonlinearity. This means that the streamfunction fields associated with their respective PV and ub anomalies, as well as contributions from the mean and residual fields of PV and ub, do not add up to yield the total streamfunction field, as they do for quasigeostrophic PV (QPV). Davis and Emanuel (1991) and Davis (1992) demonstrated how the nonlinearity problem can be resolved. For completeness this will now be briefly recaptured. Adopting the hydrostatic relation on the right-hand side of Eq. (1) yields a relation for q containing products of second derivatives of C and F [see, e.g., relation (1.4) in Davis (1992) for details]. Adopting this relation, Davis and Emanuel (1991) demonstrated how Ertel’s PV VOLUME 69 can be inverted yielding unambiguous results, as for inversion of QPV. The method is referred to as the full linear (FL) method because all terms are retained through collecting the nonlinear terms in the coefficient of the linear operator. Davis (1992) presented two different PV inversion methods: subtraction from the total (ST) and addition to the mean (AM). He demonstrated that the average of these two methods yields results close to those of the FL method except for cases when the PV perturbations are much larger than the mean PV field. The above description of the FL, ST, and AM methods applies to boundary temperature anomalies as well. Rather than adopting the FL method we use the combination of the ST and AM methods in the case study presented in section 4 below. The method has also been successfully adopted in several case studies (e.g., Kristjánsson et al. 1999; Thorsteinsson et al. 1999; Røsting and Kristjánsson 2006). Although quasigeostrophic theory is a weak anomaly theory, the inversion of QPV is useful for demonstrating the principle of PV inversion and in theoretical studies of atmospheric flows (e.g., the Eady model). The results from such studies in E2008 lead to the conclusion that PPVI in many cases fails to contribute to our understanding of the dynamics of the flow. Since PPVI has been adopted in many recent case studies (e.g., Bracegirdle and Gray 2009; McInnes et al. 2009) and constitutes a central part of PV thinking, the criticism in E2008 needs to be addressed. We now discuss some of the results presented by E2008. The main purpose of the following discussion is to investigate to what extent PPVI is valid and useful in the cases adopted in E2008. 3. Results from elementary tests of piecewise PV inversion In E2008, in connection with Fig. 1, it was stated that the winds in region D2 are induced by the positive PV anomaly Z1 in region D1. However, basing our arguments on the PPVI technique (as described in the previous section), we note that the winds induced by Z1 only contribute to the total wind field in region D2 in the same way as one of the components qi in Eq. (6) is only one of many contributions to the total perturbation field. To what extent Z1 influences the winds in D2 depends on the distance between the PV anomalies Z1 and A2 and their amplitudes. The same argument is valid for the ‘‘complementary case’’ (i.e., the winds associated with the negative PV anomaly A2 contribute to the total wind field in region D1 and may influence Z1 to a certain extent). The total MARCH 2012 937 R Ø S T I N G A N D K R I S T J Á N S S O N FIG. 1. Schematic presentation of a domain D divided into subdomains D1 and D2; Z1 and A2 represent positive and negative PV anomalies, respectively. Reproduced from Egger (2008). streamfunction ctotal in D2 is then obtained through adding the contributions from A2 and Z1. The same procedure yields ctotal in D1. Hence performing inversions for both regions D1 and D2 is a necessary requirement for assessing the validity of PPVI. FIG. 2. Observed streamfunction cob in dotted lines, forming ridges R1 and R2 and troughs T1 and T2 in regions D1 and D2, respectively. Solid lines show c obtained from inversion of the PV field after replacing the observed PV with PV 5 0 in region D2. Reproduced from Egger (2008). regions shown in Fig. 2, where q9 5 0 in region D1 and q95 qob in D2: a. The barotropic case: Rossby waves in zonal mean flow q9 5 In the equation for PV in a barotropic flow (e.g., Holton 2004) we may define the planetary vorticity f0 1 by as the mean PV field. Introducing a height increment z9 5 g21f0c, we obtain through linearization an expression for perturbation PV: q9 5 = 2 c 2 d 2 c, (10) where d21 5 (gH)1/2 f021 is the Rossby radius of deformation, with H being the mean height of the fluid. We now consider a one-dimensional barotropic Rossby wave: here Eq. (10) becomes q9 5 ›2 c 2 d2 c. ›x2 (11) In this example from E2008, the observed streamfunction field is described by cob 5 P sin(kx) from which the observed QPV anomaly field is calculated by inserting cob in Eq. (11). The case presented in E2008 now proceeds as follows: region D1 is defined for 2L/2 # x # 0 and D2 for 0 # x # L/2 as shown in Fig. 2, reproduced from E2008. Following E2008, the PV field is modified by letting q9 5 0 in the region D2. Figure 2 shows, besides cob, the streamfunction c9 obtained through inversion of the modified PV (solid contours). To test the validity of PPVI for the flow in this example, we now define the complementary case for the 0 in D1 . 2(k2 1 d2 )P sin(kx) in D2 (12) Carrying out the inversion of PV in this case we obtain the streamfunction c0, assuming, as for the case of c9, that c0 and ›c0/›x are continuous at the origin. The solution is c0 5 2B sinh[d(x 1 L/4] P sin(kx) 1 B sinh[d(x 1 L/4] in D1 , in D2 (13) where B 5 2Pk/[2dcosh(dL/4)]; that is, B is identical to the coefficient A in the case presented in E2008. The distribution of c0 in this complementary case [Eq. (13)] is antisymmetric to that displayed in Fig. 2, with respect to the origin. Addition of the two complementary fields yields, for D1 and D2, c 5 c9 1 c0 5 P sin(kx) 5 cob. Hence addition of the fields associated with the PV in D1 and D2 yields the total observed PV field, as expected for piecewise PV inversion. b. Baroclinic waves in zonal shear flow In the baroclinic case described in E2008 the flow is confined between a surface plane and a rigid lid at the top, with the height of the model atmosphere being H. The dynamical development of the flow in this case is described by the equations of the Eady model (e.g., Holton 2004; Egger 2008). Figure 3, reproduced from E2008, shows the tropopause at Hs 5 3H/4. Following E2008, an observed field is given as a modal streamfunction: 938 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 69 FIG. 3. Idealized presentation of a two-dimensional atmosphere with depth H 5 20 km. The tropopause is at z 5 3H/4 5 Hs, whereas n 5 3p/H and k 5 2p/L are the vertical and horizontal wavenumber, respectively. Shown are (a) the observed streamfunction (positive values are solid contours, while negative values are dashed) and (b) the streamfunction obtained through inversion of the PV field after replacing the observed PV with PV 5 0 in the troposphere (D2). The contours, as in (a), show the streamfunction. Reproduced from Egger (2008). ^ sin(kx). cob 5 P sin(kx) sin[n(z 2 H)] 5 c(z) (14) ^ and ›c/›z ^ (ii) Continuity requirements for c at the tropopause z 5 3H/4. The horizontal wavenumber is k 5 2p/L, and n 5 3p/H is the vertical wavenumber. Figure 3a, retrieved from E2008, shows the structure of the observed streamfunction cob. Now using quasigeostrophic PV, the perturbation is expressed by ^ is shown in Fig. 3b. The distribution of c In the complementary case, not considered by E2008, q9ob is confined to the troposphere (D2), while q9ob 5 0 in the stratosphere (D1). This PV distribution yields the following solution: in D1 (stratosphere) q9ob 5 =2 cob 1 ( f02 /N02 )(›2 cob /›z2 ), ^ 5 Ee2gz 1 Fegz , c (15) where N02 [ (g/u0 )(du0 /dz) is the squared Brunt–Väisälä frequency, expressing static stability, here assumed to be constant. By inserting the expression for cob into Eq. (15), an equation for the amplitude of the streamfunction in the x–z plane is obtained: ^ N02 q^ ›2 c 2^ c 5 2 g , ›z2 f02 (16) where g 5 kN0/f0 and q^ is the amplitude of the PV perturbation. With modifications of the observed PV—that is, changing PV to q9ob 5 0 in the troposphere [i.e., below Hs (D2)] while retaining the original q9ob in the stratosphere [i.e., above Hs (D1)]—Egger retrieved the solution of ^ through inversion of the modified PV field. c The coefficients are determined by the boundary conditions: ^ 5c ^ at z 5 0 and z 5 H. (i) c ob (17) and in D2 (troposphere) ^5c ^ 1 Ge2gz 1 Kegz c ob (18) The coefficients in Eqs. (17) and (18) may be determined by using the boundary conditions (i) and (ii). Adding the streamfunctions in E2008 to the ones in Eqs. (17) and (18), in regions D1 and D2 respectively, the ^ is obtained. Hence, as in the observed streamfunction c ob barotropic case, PPVI yields the desired results. c. Baroclinic waves in zonal shear flow: Impacts from the lower boundary temperature anomalies The final example from E2008 to be discussed addresses the impact on the flow from lower boundary temperature anomalies. With suitable boundary conditions the atmospheric flow (in region D1) associated with the temperature anomalies at the lower boundary (region D2) becomes c * 5 2nP sin(k x) sinh[g(z 2 H)] /g cosh(gH). (19) MARCH 2012 R Ø S T I N G A N D K R I S T J Á N S S O N The validity of PPVI in this case may be verified by again including the complementary case, where q9 5 q9ob in 0 , z # H. The atmospheric flow for the complementary case, in D1 (0 , z # H), is ^5c ^ 1 Ae2gz 1 Begz . c ob (20) The coefficients in Eq. (20) are determined by the following boundary conditions: ^5 (i) At the top of the model atmosphere (z 5 H), c ^ 5 0. c ob ^ (ii) At the lower boundary ›c/›z 5 0 as z / 0. Through application of the assumptions (i) and (ii) the inverted streamfunction for the complementary case is retrieved: c** 5 P sin(k x) sin[n(z 2 H)] 1 nP sin(kx) sinh[g (z 2 H)]/ g cosh(g H). (iii) c / 0 when r / ‘. The solution yields an induced cyclonic wind field that is strongest at the surface of the PV anomaly (i.e., at r 5 r0). Beyond the surface the winds become evanescent with increasing distance r [see Fig. 6.10 in Holton (2004), originally from Thorpe and Bishop (1995)]. We now assume that D1 is the region 0 # r # r0 and D2 is the adjacent region r . r0. We note from Eq. (23) that this problem avoids the requirement q9 5 0 in D1. This means that PV inversion in this case yields the c field associated with a specific positive PV anomaly. The solution of Eq. (23) is quite different from the ones presented in E2008, such as the solution presented in Fig. 2 for the barotropic case where the associated winds appear to remain strong far away from the PV anomaly. In fact, the results from the case described by Eq. (23) have a strong resemblance to NWP simulated flows associated with isolated PV anomalies. (21) The addition of expressions (19) and (21), which describe the two complementary cases, yields for region D1 c 5 c* 1 c** 5 P sin(k x) sin[n(z 2 H)] [ cob . 939 (22) Thus, we find that the two fields obtained from piecewise QPV inversion for the whole region, comprising D1 and D2, again add to yield the observed field. Hence, also for this case the criticism of PPVI by E2008 is refuted. d. The impact of an isolated QPV anomaly The cases presented in E2008 appear to depend strongly on the mathematical formulation of the problems, including the necessity of introducing the conditions c 5 0, implying q9 5 0 in a selected domain, as seen from, for instance, Eq. (12). We now compare the inversion of the PV anomalies (11) and (15) with an example discussed by Thorpe and Bishop (1995) and Holton (2004). Expressing the QPV anomaly (15) with a three-dimensional Laplacian operator through the transformation z^ 5 (N0 /f0 )z and introducing spherical coordinates, we get through symmetry q0 0 # r # r0 1 › 2 ›c r 5 . (23) 2 0 r . r0 ›r r ›r In this example the QPV anomaly is assumed to consist of a constant q0 within a ball-shaped region. Relation (23) is solved by integration with the following continuity and boundary conditions: (i) c is defined at the origin. (ii) Continuity of c and ›c/›r at the boundary (r 5 r0) of the PV anomaly. 4. Case study We now present a case study describing a real weather situation, in which the results from PPVI, using the technique described in section 2 (i.e., combining the ST and AM methods) explain properly important aspects of cyclogenesis. This implies explaining the mutual intensification of the upper-level PV anomaly and the lowlevel (boundary) temperature anomaly, as well as the strong frontogenesis that took place. The fields in Fig. 4 are based on a successful NWP simulation of the severe North Atlantic winter storm of 10–11 January 2006. The position of the surface low is indicated by L. Two snapshots of the development are shown: the first one at 0600 UTC 10 January shown in Figs. 4a,b and the second 6 h later in Figs. 4c,d. PPVI reveals the following sequence of events: (i) A positive upper-level PV (UPV) anomaly is shown in Fig. 4a. The winds obtained from PPVI are strong at the edge of the UPV anomaly and weaken with distance away from it. There is pronounced lowlevel temperature advection by the wind associated with the UPV anomaly (Fig. 4b). Strong cold advection takes place west and south of the cyclone center, while warm advection is pronounced in the warm frontal region north and northwest of the surface low. This pattern of temperature advection enhances the lower boundary warm anomaly observed at the cyclone center denoted by L. (ii) Six hours later (Fig. 4c) the UPV anomaly has become stronger because of the impacts from the winds induced by the strengthened lower boundary warm anomaly. Figure 4d shows that the warm front has 940 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 69 FIG. 4. Fields based on a simulation by the Norwegian High-Resolution Limited-Area Model (HIRLAM12). The position of the surface low is indicated by L. (a) A positive UPV anomaly (contours every 0.5 PVU) and the associated winds obtained through PPVI, presented at the 400-hPa level. Fields are valid at 0600 UTC 10 Jan 2006. (b) Potential temperature (contours every 2 K) at 950 hPa and the winds (obtained through PPVI) at 900 hPa associated with the UPV anomaly shown in (a). The fields are valid at 0600 UTC 10 Jan 2006. (c) As in (a), but valid at 1200 UTC 10 Jan 2006. (d) As in (b), but valid at 1200 UTC 10 Jan 2006. intensified and the cold advection west and south of the cyclone center has become stronger over the 6-h period through the intensification of the winds related to a stronger positive UPV anomaly. In turn the cold advection to the south of the cyclone center enhances the lower boundary warm anomaly observed at L in Fig. 4d. Hence the mutual interaction between the warm anomaly at the lower boundary and the UPV anomaly continues and becomes stronger. Figure 4 only shows one selected UPV anomaly and the low-level boundary temperature fields. Nevertheless, the UPV and boundary temperature anomalies presented in Fig. 4 have a crucial impact on the initial stage of the cyclogenesis. This example illustrates that PPVI is very useful for analyzing instantaneous flow patterns, as well as the dynamics of cyclogenesis. 5. Discussion and conclusions By presenting some examples of PV inversion, Egger (2008) claimed that PPVI is generally not useful for understanding atmospheric dynamics. Here, by reanalyzing some of the cases presented in E2008 we demonstrate that there are no reasons to discard PPVI as a valid diagnostic tool. In E2008 the observed PV field was changed to qob 5 0 in a selected domain (D2), while qob was retained in the remaining adjacent domain (D1). Regarding the PV field in the adjacent domain (D1) as the anomaly, inversion of this anomaly for the whole integration region (D1 1 D2) yields a streamfunction different from the one that is observed. We have argued that this procedure is misleading, and that in order to test the validity of PPVI it is necessary to invert the PV field also in the complementary cases. The complementary cases are obtained by retaining qob and letting qob 5 0 in the selected (D2) and adjacent (D1) domains respectively. Addition of the field obtained through PV inversion in the complementary case should then yield the total (observed) field if PPVI is valid. As we have shown, by including the complementary cases, the formalism of piecewise PV inversion indeed remains valid in the cases presented in E2008. MARCH 2012 R Ø S T I N G A N D K R I S T J Á N S S O N The examples presented in E2008 do not include interaction between PV anomalies, a shortcoming also pointed out by Methven and de Vries (2008). Such interactions constitute an important part of assessing the dynamics through PPVI. The results from E2008 are also model dependent, which is illustrated by the example given by Eq. (23). Inversion of the QPV anomaly given by Eq. (23) yields a solution resembling that obtained through PPVI for a real case presented in Fig. 4. The solution of Eq. (23) is very different from those of the cases presented in E2008, illustrating the dependence of the results from PPVI on the mathematical and dynamical formulation of the problem. Based on the above examples and case study (Fig. 4), it appears that PPVI is a highly useful tool for diagnosing dynamical processes in the atmosphere. In particular, PPVI provides a method for assessing which regions of the atmospheric flow are important for specific dynamical developments. Although the various contributions from individual PV anomalies, obtained through PPVI, cannot be observed in the real atmosphere, nonetheless the addition of the contributions from all the PV anomalies, including those from the residual and mean PV, yields the observed streamfunction field. This suggests that the various contributions from different PV anomalies are conceptually meaningful (e.g., Thorsteinsson et al. 1999). This is also confirmed by case studies of real situations in which frontogenesis and low-level developments can be understood through the impact from specific PV anomalies as shown in section 4 (e.g., Davis and Emanuel 1991; Stoelinga 1996; Røsting et al. 2003; Bracegirdle and Gray 2009). An interesting example of application of PPVI was recently provided by Hinssen et al. (2011). They studied the impact of the stratospheric PV anomaly associated with the polar winter vortex on the tropospheric flow. Given the large horizontal dimension of the stratospheric polar vortex, an associated large-scale PV anomaly was defined, and its effect on the tropospheric flow turned out to be noticeable. During sudden stratospheric warmings (SSWs) a breakup or substantial weakening of the stratospheric PV anomaly and the associated polar vortex takes place. PPVI shows that the westerly mean flow in the troposphere weakens and may even become easterly in some regions after an SSW. REFERENCES Bracegirdle, T. J., and S. L. Gray, 2009: The dynamics of a polar low assessed using potential vorticity inversion. Quart. J. Roy. Meteor. Soc., 135, 880–893. Charney, J. G., 1947: The dynamics of long waves in a baroclinic westerly current. J. Meteor., 4, 135–163. ——, 1955: The use of the primitive equations of motions in numerical prediction. Tellus, 7, 22–26. 941 Davis, C. A., 1992: Piecewise potential vorticity inversion. J. Atmos. Sci., 49, 1397–1411. ——, and K. A. Emanuel, 1991: Potential vorticity diagnostics of cyclogenesis. Mon. Wea. Rev., 119, 1929–1953. Demirtas, M., and A. J. Thorpe, 1999: Sensitivity of short range weather forecasts to local potential vorticity modifications. Mon. Wea. Rev., 127, 922–939. Eady, E. T., 1949: Long waves and cyclone waves. Tellus, 1, 33–52. Egger, J., 2008: Piecewise potential vorticity inversion: Elementary tests. J. Atmos. Sci., 65, 2015–2024. Ertel, H., 1942: Ein neuer hydrodynamischer Wirbelsatz. Meteor. Z., 59, 277–281. Hinssen, Y., A. van Delden, and T. Opsteegh, 2011: Influence of sudden stratospheric warmings on tropospheric winds. Meteor. Z., 20, 259–266. Holton, J. R., 2004: An Introduction to Dynamic Meteorology. 4th ed. Elsevier Academic, 535 pp. Hoskins, B. J., M. E. McIntyre, and A. W. Robertson, 1985: On the use and significance of isentropic potential vorticity maps. Quart. J. Roy. Meteor. Soc., 111, 877–946. Kristjánsson, J. E., S. Thorsteinsson, and G. F. Ulfarsson, 1999: Potential vorticity–based interpretation of ‘‘the Greenhouse Low’’, 2–3 February 1991. Tellus, 51A, 233–248. Manders, A. M. M., W. T. M. Verkley, J. J. Diepeveen, and A. R. Moene, 2007: Application of a potential vorticity modification method to a case of rapid cyclogenesis over the Atlantic Ocean. Quart. J. Roy. Meteor. Soc., 133, 1755–1770. Mansfield, D., 1996: The use of potential vorticity as an operational forecasting tool. Meteor. Appl., 3, 195–210. McInnes, H., J. E. Kristjánsson, H. Schyberg, and B. Røsting, 2009: An assessment of a Greenland lee cyclone during the Greenland Flow Distortion experiment: An observational approach. Quart. J. Roy. Meteor. Soc., 135, 1968–1985. Methven, J., and H. de Vries, 2008: Comments on ‘‘Piecewise potential vorticity inversion: Elementary tests.’’ J. Atmos. Sci., 65, 3003–3008. Rossby, C. G., 1940: Planetary flow patterns in the atmosphere. Quart. J. Roy. Meteor. Soc., 66, 68–87. Røsting, B., and J. E. Kristjánsson, 2006: Improving simulations of severe winter storms by initial modification of potential vorticity in sensitive regions. Quart. J. Roy. Meteor. Soc., 132, 2625–2652. ——, and ——, 2008: A successful resimulation of the 7–8 January 2005 winter storm through initial potential vorticity modification in sensitive regions. Tellus, 60A, 604–619. ——, ——, and J. Sunde, 2003: The sensitivity of numerical simulations to initial modifications of potential vorticity—A casestudy. Quart. J. Roy. Meteor. Soc., 129, 2697–2718. Santurette, P., and C. G. Georgiev, 2005: Weather Analysis and Forecasting: Applying Satellite Water Vapor Imagery and Potential Vorticity Analysis. Elsevier Academic, 179 pp. Stoelinga, M. T., 1996: A potential vorticity–based study of the role of diabatic heating and friction in a numerically simulated baroclinic cyclone. Mon. Wea. Rev., 124, 849–874. Thorpe, A. J., and C. H. Bishop, 1995: Potential vorticity and the electrostatics analogy: Ertel–Rossby formulation. Quart. J. Roy. Meteor. Soc., 121, 1477–1495. Thorsteinsson, S. T., J. E. Kristjánsson, B. Røsting, V. Erlingsson, and G. F. Ulfarsson, 1999: A diagnostic study of the Flateyri avalanche cyclone, 24–26 October 1995, using potential vorticity inversion. Mon. Wea. Rev., 127, 1072–1088. Verkley, W. T. M., P. W. C. Vosbeek, and A. R. Moene, 2005: Manually adjusting a numerical weather analysis in terms of potential vorticity using three-dimensional variational dataassimilation. Quart. J. Roy. Meteor. Soc., 131, 1713–1736.