## 1. Introduction

*ζ*:

*κ*is the von Kármán constant;

*z*is the height above the ground;

*u*

_{*}and

*θ*

_{*}are, respectively, the friction velocity and temperature scales; and

*φ*

_{m}and

*φ*

_{h}are, respectively, the nondimensional similarity functions. Here,

*ζ*is defined as

*g*is the acceleration due to gravity.

*ζ*→ 0), both

*φ*

_{m}and

*φ*

_{h}must be equal to a constant that is equal to unity for

*φ*

_{m}, while the corresponding value for

*φ*

_{h}is determined from field observations. In the case of strongly unstable conditions, when the value of −

*ζ*becoming large, it is assumed that buoyancy dominates the local turbulence process (Högström 1996). In such a case, friction velocity

*u*

_{*}ceases to be the correct scaling parameter and should not be used for a nondimensional process, which suggests that

*φ*

_{m}and

*φ*

_{h}should vary as power-law functions of

*ζ*in local free-convective conditions (Grachev et al. 2000). The dimensional analysis suggests that the exponent should be −1/3 for both

*φ*

_{m}and

*φ*

_{h}, while Businger (1966) suggested that the exponent should be −1/4 for

*φ*

_{m}and −1/2 for

*φ*

_{h}in free-convection limit. Various authors (Obukhov 1971; Kazansky and Monin 1956; Ellison 1957; Yamamoto 1959; Panofsky 1961; Sellers 1962) have proposed a formula for

*φ*

_{m}, commonly referred to as the Kazansky and Monin–Ellison–Yamamoto–Panofsky–Sellers (KEYPS) or Obukhov KEYPS (O’KEYPS) formula in the literature as

*γ*is a constant. The KEYPS formula captures the theoretical scaling for near-neutral to moderately unstable conditions and the −1/3 power law in free-convection limit as predicted by MOST. Based on the “direction dimensional analysis,” Kader (1988), Kader and Perepelkin (1989), and Kader and Yaglom (1990) presented a different conceptual structure of the atmospheric surface layer (ASL) under unstable conditions by dividing the unstable ABL into three different sublayers, namely, dynamic sublayer, dynamic–convective sublayer, and free-convective sublayer. The three sublayer model suggested by Kader and Yaglom (1990) predicts that the function

*φ*

_{m}, in theory, follows a +1/3 power law in free-convective conditions, which is neither consistent with MOST prediction nor the historical KEYPS formula. Based on the above three types of theoretical frameworks, various forms of

*φ*

_{m}and

*φ*

_{h}functions have been developed over the years. Most of these functions are derived from turbulence observations whose characteristics depend upon the accuracy of the measurements as well as the theoretical applicability of the similarity theory in that range. Making accurate measurements in strongly stratified conditions is a challenging task, which prevents us from determining the exact functional form of similarity functions in very unstable conditions.

The parameterized transfer coefficients of momentum (*C*_{D}) and heat (*C*_{H}) and corresponding surface fluxes in the atmospheric models depend on the behavior of *φ*_{m} and *φ*_{h} (Yang et al. 2001; Srivastava and Sharan 2017). In an earlier study, Srivastava and Sharan (2015) have pointed out the existence of a nonmonotonic behavior of *C*_{D} with atmospheric stability, which is not consistent with the widely expected monotonically increasing characteristic of drag coefficient with increasing instability of the surface layer. They argue that while MOST-based parameterized form of *C*_{D} depends upon the functional behavior of similarity function *φ*_{m}, various operational weather and climate models use one or another form of *φ*_{m} and *φ*_{h} for parameterization of fluxes using bulk flux algorithm without quantifying the uncertainties involved in the estimated values of fluxes associated with the form of those functions. For example, the way functional forms of *φ*_{m} and *φ*_{h} are prescribed in the bulk algorithm is quite different for global and regional scale models such as the Community Earth System Model (CESM; Zeng et al. 1998), Regional Climate Model (RegCM; Giorgi et al. 2012), and the Weather Research and Forecast (WRF) Model (Skamarock et al. 2008). The CESM utilizes functional form based on theoretical prediction of Kader and Yaglom (1990) in free-convective conditions, while RegCM and WRF use either classical Businger–Dyer functions (Businger et al. 1971) or an interpolation function suggested by Grachev et al. (2000) depending upon which surface layer module is selected for simulations (Jimenez et al. 2012). The level of disparity in the utilization of similarity functions among different weather and climate models points out that uncertainties in the estimated transfer coefficients and surface fluxes due to different similarity functions might be larger than the uncertainties in the other input parameters such as prescribed surface temperature and roughness lengths of momentum and heat in the bulk flux algorithm. The motivation behind this study is to find out the level of uncertainty one may expect in the estimated fluxes and transfer coefficients associated with the functional forms of similarity functions used for parameterization under unstable conditions. Thus, an attempt has been made here to classify various functional forms of *φ*_{m} and *φ*_{h} in different categories and then quantify the possible uncertainties in the estimation of surface fluxes for different classes of functions used in the weather forecast and climate models.

## 2. Bulk algorithm for parameterization of transfer coefficients and surface fluxes

*z*

_{0}and

*z*

_{h}are, respectively, roughness lengths of momentum and heat;

*θ*

_{0}is the potential temperature at the height

*z*

_{h};

*ψ*

_{m}and

*ψ*

_{h}are, respectively, the integrated similarity functions corresponding to similarity functions

*φ*

_{m}and

*φ*

_{h}. The stability parameter

*ζ*is defined as in Eq. (3). The integrated similarity functions, commonly known as stability correction functions

*ψ*

_{m}and

*ψ*

_{h}are given by

*ζ*′ is a dummy variable of integration corresponding to

*ζ*. For land surfaces, where

*z*

_{0}and

*z*

_{h}are taken as constants, Eqs. (3), (5), snf (6) are solved iteratively to estimate

*ζ*and bulk transfer coefficients

*C*

_{D}and

*C*

_{H}as

*C*

_{D}and

*C*

_{H}are used to estimate the turbulent fluxes of heat (

*H*) and momentum (−

*τ*) from

*ρ*is the dry air density and

*C*

_{p}is the specific heat capacity of dry air at constant pressure.

_{B}, which is defined as

_{B}becomes

_{B},

*ζ*is estimated by calculating the root of least magnitude of Eq. (13).

## 3. Commonly used functional forms of *φ*_{m} and *φ*_{h} under unstable conditions

*φ*

_{m}and

*φ*

_{h}was proposed by Businger (1966) and A. J. Dyer [1965, unpublished material; see Businger (1988) for details] for unstable stratification as

*γ*

_{m}= 15 and

*γ*

_{h}= 9 and

*φ*

_{m}valid for the stability range −10 ≤

*ζ*≤ 0 as

*β*

_{m}and

*β*

_{h}can be found in the literature; for example,

*β*

_{m}=

*β*

_{h}= 15 (Carl et al. 1973),

*β*

_{m}and

*β*

_{h}= 40 (Delage and Girard 1992), and

*β*

_{m}=

*β*

_{h}= 12.87 (Fairall et al. 1996). Högström (1988) reformulated the universal similarity functions of Businger et al. (1971) of the form

*ζ*≤ 0 (Foken 2006). Fairall et al. (1996, 2003) suggested an interpolation function valid for the full range of instability as

*γ*

_{m}=

*γ*

_{h}= 16), and

*φ*

_{m,h}functions rather than the corresponding integrated functions

*ψ*

_{m,h}suggested by Fairall et al. (1996, 2003) [Eq. (19)].

*φ*

_{m}with

*ζ*in the three sublayers. In the dynamic sublayer, the theory suggests that turbulence production occurs predominately in the along wind direction, (

*x*component) and the pressure redistribution term feeds turbulence energy into vertical

*z*component (Högström 1996). The sublayer is valid for

*ζ*lying in the range −1/40 <

*ζ*< 0, which corresponds to near-neutral conditions. In this condition,

*φ*

_{m}(

*ζ*) = 1 while

*φ*

_{h}(

*ζ*) = Pr

_{t}. In the dynamic-convective sublayer, mechanical energy is produced in the

*x*component, whereas buoyancy produces energy in the

*z*component independently. This sublayer is valid for

*ζ*lying in the range −0.4 <

*ζ*< −1/40. For this sublayer, the dimension analysis suggests the functional form for similarity functions as

*A*

_{u}and

*A*

_{T}are constants. In this sublayer, the similarity functions behave similarly to that in the classical free-convective conditions. However, the range of values of

*ζ*for which this sublayer is valid is limited to

*ζ*= −0.4. In this sublayer, both the mechanical and buoyancy production terms of turbulence kinetic energy are equally important (Kader and Yaglom 1990). In the free-convective sublayer, energy production due to buoyancy dominates the mechanical production term, and the energy produced due to buoyancy in the vertical direction is fed into the horizontal direction by pressure redistribution term (Kader and Yaglom 1990; Högström 1996). Thus, in this case, the two length scales are again coupled, and there exists only one relevant velocity scale in the vertical direction. The dimensional analysis suggests

*B*

_{u}and

*B*

_{T}are constants.

The remarkable prediction of this sublayer is that unlike the case of classical free-convection limit, in which both *φ*_{m} and *φ*_{h} follow a −1/3 power law, here *φ*_{m} varies as the +1/3 power of *ζ* for sufficiently large values of −*ζ*. Thus, the theory of Kader and Yaglom (1990) suggests that *φ*_{m} has a nonmonotonic variation with respect to −*ζ* under unstable conditions. Based on the three sublayer model of Kader and Yaglom (1990), various expressions for *φ*_{m} and *φ*_{h} have been suggested and one of those expressions of *φ*_{m} and *φ*_{h}, which is being utilized in CESM (Zeng et al. 1998), is taken here for the analysis.

To quantify the impact of functional forms of *φ*_{m} and *φ*_{h} on the values of estimated fluxes, we have identified four classes of *φ*_{m} and *φ*_{h} functions satisfying the following:

−1/4 and −1/2 power law for

*φ*_{m}and*φ*_{h}, respectively (Businger et al. 1971; Högström 1996), for the full range of instability [Eqs. (14) and (15); class I functions].−1/3 power law for both

*φ*_{m}and*φ*_{h}(Carl et al. 1973) for the full range of instability [Eq. (16); class II functions].−1/4 and −1/2 power law in near neutral for

*φ*_{m}and*φ*_{h}, respectively, and −1/3 power law in very unstable conditions for both*φ*_{m}and*φ*_{h}(Fairall et al. 1996; Grachev et al. 2000; Fairall et al. 2003) [Eq. (19); class III functions].−1/4 and −1/2 power law in near-neutral for

*φ*_{m}and*φ*_{h}, respectively, with a −1/3 power law in very unstable conditions for*φ*_{h}and +1/3 power law in very unstable conditions for*φ*_{m}(Kader and Yaglom 1990; Zeng et al. 1998; class IV functions).

_{B}, the value of

*ζ*is estimated by calculating the root of least magnitude of the transcendental equation in

*ζ*and Ri

_{B}[Eq. (13)]. For computation purposes, the range of values of Ri

_{B},

*z*/

*z*

_{0}, and

*z*

_{0}/

*z*

_{h}is taken as −2 ≤ Ri

_{B}< 0, −0.5 ≤ log(

*z*

_{0}/

*z*

_{h}) < 29.0, and 10 ≤

*z*/

*z*

_{0}< 10

^{5}. The estimated value of

*ζ*is then utilized to compute the transfer coefficients of momentum

*C*

_{D}and heat

*C*

_{H}[Eqs. (8) and (9)].

## 4. Results and discussion

### a. Uncertainty in the estimation of transfer coefficients

Figures 1a–c show the variation of the stability parameter with the bulk Richardson number for the functions *φ*_{m} and *φ*_{h} of Dyer (1974), Carl et al. (1973), Fairall et al. (1996, 2003), and Zeng et al. (1998) for different values of roughness length of momentum. For the smaller magnitude of Ri_{B} the values of *ζ* obtained from different functional forms of *φ*_{m} and *φ*_{h} are found to be nearly identical. However, the deviations in the predicted values of *ζ* for a given Ri_{B} from different formulations are found to increase with decreasing values of Ri_{B}. The −1/4 and −1/2 power law for *φ*_{m} and *φ*_{h}, respectively (class I functions), and −1/3 power law for both *φ*_{m} and *φ*_{h} (class II functions) and the interpolation functions suggested by Fairall et al. (1996, 2003) (class III functions) predict relatively smaller absolute values of *ζ* for the values of Ri_{B} larger than 0.2. However, the absolute values of *ζ* predicted by the functional forms of *φ*_{m} and *φ*_{h} following +1/3 and −1/3 power laws, respectively (class IV; Kader and Yaglom 1990), are found to be very large as compared to those predicted by class I, II, and III functional forms of *φ*_{m} and *φ*_{h}. This behavior is consistent for all the values of *z*/*z*_{0} (Figs. 1a–c) representative of smooth, transitional, and rough surfaces. The relatively higher magnitude of *ζ* for a given value of Ri_{B} suggests that both the momentum and heat fluxes predicted with the class IV functional forms of *φ*_{m} and *φ*_{h} would be relatively small as compared to those predicted by class I, II, and III functional forms. The overall impact of the class IV functional form of *φ*_{m} and *φ*_{h} is a significant reduction in the estimated magnitude of surface fluxes in the moderately to strongly unstable conditions as compared to those predicted by other commonly used functional forms of *φ*_{m} and *φ*_{h}. Figures 1d–i show the variation of *C*_{D} and *C*_{H} with Ri_{B} for all the four types of similarity functions *φ*_{m} and *φ*_{h} for different values of roughness length of momentum. The class I, II, and III functional forms of *φ*_{m} and *φ*_{h} suggest relatively higher values of *C*_{D} for a given value of Ri_{B} in moderately to strongly unstable conditions. In contrast, the values of *C*_{D} predicted by class IV functional form of *φ*_{m} and *φ*_{h} are significantly smaller as compared to those predicted by the other three functional forms. One striking feature regarding the behavior of *C*_{D} observed for class IV functions is that *C*_{D} shows a nonmonotonic characteristic, which is in contradiction to the prediction of other forms of *φ*_{m} and *φ*_{h} (Figs. 1d–e). The values of *C*_{H} are found to increase with increasing instability for all four types of similarity functions. However, the rate of increase in the values of *C*_{H} with Ri_{B} is considerably small in the case of class IV functions (Figs. 1g–i). The difference in the estimated values of *C*_{H} for a given Ri_{B} is relatively more pronounced in the case of higher values roughness length *z*_{0} (Fig. 1i). Notice that in the case of the rough regime, even *C*_{H} shows a nonmonotonic variation with respect to Ri_{B} in case of class IV functions, which is in contradiction to that predicted by the other three classes of similarity functions. One notable feature with the class IV functions based on the three sublayer model of Kader and Yaglom (1990) is that the range of variation of values of *C*_{D} and *C*_{H} with stability is relatively narrow as compared to that shown by the other three classes of correction functions. The values of *C*_{D} and *C*_{H} are found to be bounded by twice of their near-neutral values with the class IV stability correction functions while the other classes of functions show continuously increasing values of *C*_{D} and *C*_{H} with increasing instability.

_{B}< 0, −0.5 ≤ log(

*z*

_{0}/

*z*

_{h}) < 29.0, 10 ≤

*z*/

*z*

_{0}< 10

^{5}using the formula

*ζ*

_{p},

*ζ*

_{KY},

_{B}. Figure 2a shows the maximum possible difference in the estimated values of stability parameter

*ζ*with respect to the ratio

*z*/

*z*

_{0}when class IV functions are utilized in place of class I (gray line), class II (dashed line), and class III (dotted line). Notice that the differences in the estimated values of

*ζ*,

*C*

_{D}, and

*C*

_{H}depend upon three parameters: Ri

_{B},

*z*

_{0}/

*z*

_{h}, and

*z*/

*z*

_{0}. In Fig. 2a, the maximum possible differences are calculated with respect to

*z*/

*z*

_{0}for the values of Ri

_{B}and

*z*

_{0}/

*z*

_{h}lying in the range −2 ≤ Ri

_{B}< 0 and −0.5 ≤ log(

*z*

_{0}/

*z*

_{h}) < 29.0. A large difference in the estimated values of

*ζ*is found to occur if class IV functions are utilized in the bulk parameterization of surface fluxes based on MOST. The difference is found to increase with the increasing value of

*z*

_{0}. Similar deviations are observed for momentum and transfer coefficients (Figs. 2b,c), suggesting that the differences in the estimated values of fluxes with different classes of stability correction functions are quite large as compared to possible uncertainties in the prescription of input parameters such as wind speed and temperature required for estimation of surface fluxes from bulk flux algorithm. Since the class IV functions are found to reduce the magnitude of the estimated fluxes considerably in moderately to strongly unstable conditions, they can be utilized as an alternative to overcome the problem of overestimation of surface fluxes by the weather forecast models under unstable conditions.

Maximum percentage difference between the values of *ζ*, *C*_{D}, and *C*_{H} computed with class IV functions and the functions lying in other three classes.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

Maximum percentage difference between the values of *ζ*, *C*_{D}, and *C*_{H} computed with class IV functions and the functions lying in other three classes.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

Maximum percentage difference between the values of *ζ*, *C*_{D}, and *C*_{H} computed with class IV functions and the functions lying in other three classes.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

### b. Nonmonotonic behavior of drag coefficient C_{D}

Srivastava and Sharan (2015) utilized the turbulent data over a tropical region to analyze the observational behavior of *C*_{D} with respect to wind speed *U* and stability parameter *ζ* under unstable conditions. They found that the observed variation of *C*_{D} with *ζ* is nonmonotonic and bounded by a curve that shows increasing behavior with −*ζ*, until it reaches a peak and then, decreases further with increasing instability (Srivastava and Sharan 2015). They have argued that MOST is unable to explain the observed decreasing behavior of *C*_{D} with *ζ* in moderate to strong unstable conditions over tropics within the framework of commonly used similarity functions belonging to classes I, II, and III. The predicted qualitative behavior of *C*_{D} with *ζ* is found to be consistent with those observed by Srivastava and Sharan (2015) over tropics if the class IV functions are utilized for estimation of surface fluxes. In appendix B of Srivastava and Sharan (2015), a nonmonotonic form of similarity function *φ*_{m} [appendix B, Eq. (B1), Srivastava and Sharan 2015] was considered, and based on the mathematical analysis, it was concluded that the parameterized drag coefficient *C*_{D} increases with increasing value of −*ζ* until it attains a maximum value at *ζ* = *ζ*_{c} and then starts decreasing with further increasing value of −*ζ*. Depending on the sign of the expression [Eq. (A9), appendix A, Srivastava and Sharan (2015)], it was argued that the value of *ζ* at which *C*_{D} changes its behavior is same at which function *φ*_{m} changes its behavior, that is, *ζ* = *ζ*_{c}. However, in general, both the values need not be same. This error was incurred due to wrong interpretation of the expression [Eq. (A9), appendix A, Srivastava and Sharan (2015)] and here we present a correction to the analysis given in the Srivastava and Sharan (2015).

*φ*

_{m}is taken as

*φ*

_{1}(

*ζ*) is a continuously decreasing function of −

*ζ*in the range

*ζ*

_{c}≤

*ζ*≤ 0 and

*φ*

_{2}(

*ζ*) is an increasing function for

*ζ*≤

*ζ*

_{c}.

*C*

_{D}can be written as

*ψ*

_{m}is the integrated similarity function corresponding to the similarity function

*φ*

_{m}. Notice that traditionally the drag coefficient is defined as

*C*

_{D}= (

*u*

_{*}/

*U*)

^{2}, where

*u*

_{*}is the friction velocity and

*U*is the wind speed at reference height. In case of low wind conditions when

*U*→ 0,

*C*

_{D}tend to have very large values. However, low wind convective conditions are associated with the effect of large-scale coherent structures within the whole boundary layer (Grachev et al. 1998), which generates random gusts. Godfrey and Beljaars (1991) suggested that the gustiness velocity associated with these random gusts should be added to the vector mean wind speed

*U*. Thus, in low wind convective boundary layer an effective wind speed

*U*

_{eff}should be used in place of the usual vector averaged wind speed

*U*, which is the sum of

*U*and the gustiness velocity

*U*

_{gs}. The gustiness velocity is parameterized in terms of Deardorff convective velocity scale

*w*

_{*}(Deardorff 1970) as

*U*

_{gs}=

*βw*

_{*}, where

*β*is an empirical constant. Thus, in the case of low wind convective conditions, the updated drag coefficient can be expressed as

*C*

_{D}= (

*u*

_{*}/

*U*

_{eff})

^{2}, where

*U*

_{eff}=

*U*+

*βw*

_{*}.

*η*= −

*ζ*in Eq. (27) implying that

*η >*0 in unstable conditions and differentiating

*η*, one gets

*Y*

_{m}with respect to

*η*, given as

*η*.

*φ*

_{m}(

*ζ*),

*C*

_{D}depends on the sign of

*φ*

_{1}(

*η*) and

*φ*

_{2}(

*η*) as well as the value of the ratio

*z*/

*z*

_{0}. For clarity, a functional form suggested by Kader and Yaglom (1990) may be taken as a representative of expression (25) defined as (Fig. 3)

*φ*

_{1}= (1 − 16

*ζ*)

^{−1/4},

*φ*

_{2}= 0.7

*κ*

^{2/3}(−

*ζ*)

^{1/3}, and

*ζ*

_{c}= −1.574.

Variation of drag coefficient (*C*_{D}) and similarity function *φ*_{m} suggested by Kader and Yaglom (1990) with stability parameter *ζ*. For the computation of values of *C*_{D}, *z*/*z*_{0} is assumed to be 10^{−5}. The vertical dashed lines are plotted at the values of *ζ* at which *C*_{D} and *φ*_{m} change their monotonic behavior.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

Variation of drag coefficient (*C*_{D}) and similarity function *φ*_{m} suggested by Kader and Yaglom (1990) with stability parameter *ζ*. For the computation of values of *C*_{D}, *z*/*z*_{0} is assumed to be 10^{−5}. The vertical dashed lines are plotted at the values of *ζ* at which *C*_{D} and *φ*_{m} change their monotonic behavior.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

Variation of drag coefficient (*C*_{D}) and similarity function *φ*_{m} suggested by Kader and Yaglom (1990) with stability parameter *ζ*. For the computation of values of *C*_{D}, *z*/*z*_{0} is assumed to be 10^{−5}. The vertical dashed lines are plotted at the values of *ζ* at which *C*_{D} and *φ*_{m} change their monotonic behavior.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

*φ*

_{m}, an expression for

*η*that attains its maximum value 1 as

*η*→ 0. The other function

*φ*

_{2}= 0.7

*κ*

^{2/3}(

*η*)

^{1/3}is a monotonic increasing that attains its minimum value 0 as

*η*→ 0 and

*φ*

_{2}> 1 for

*η*

_{m}, such that

*η*≤

*η*

_{m}but

*η*≥

*η*

_{m}. The exact value of

*η*

_{m}depends upon the value of the ratio

*z*/

*z*

_{0}.

Thus, the value of *C*_{D} increases until it attains a maximum value at *η* = *η*_{m} and then starts decreasing with increasing value of *η*. However, the points *ζ*_{c} and *ζ*_{m} = −*η*_{m} are different (Fig. 3).

### c. Observational evidence and recommendation for the form of φ_{m} over the Indian region

*φ*

_{m}, which appears to capture the qualitative nature of drag coefficient with near-surface atmospheric stability over the Indian region (Srivastava and Sharan 2015). We have further carried out data analysis to support the findings of Srivastava and Sharan (2015). The dataset used in the present study is obtained from a fast response sensor (CSAT3 sonic anemometer) installed at 10 m height in the Birla Institute of Technology Mesra, Ranchi (23.412°N, 85.440°E), India, with an average elevation 609 m above sea level (Dwivedi et al. 2015; Sharan and Srivastava 2016; Srivastava and Sharan 2017). Turbulence measurements at 10 Hz frequency for the year 2009 are used to calculate the hourly turbulent fluxes using the eddy covariance technique. Friction velocity

*u*

_{*}is calculated from the expression

*u*′,

*υ*′, and

*w*′ are, respectively, the fluctuations in longitudinal, lateral, and vertical wind components.

*ζ*is calculated from the expression (3) and the drag coefficient

*C*

_{D}is estimated using

^{−1}are excluded to minimize the uncertainty in values of drag coefficient in low wind conditions. Similar to Srivastava and Sharan (2015), the dataset is further classified into five unstable sublayers (Kader and Yaglom 1990; Bernardes and Dias 2010), namely, dynamic (DNS), dynamic–dynamic-convective transition (DNS–DCS transition), dynamic-convective (DCS), dynamic-convective–free-convective transition (DCS–FCS transition), and free-convective (FCS) sublayers. Notice that Srivastava and Sharan (2015) have analyzed turbulent measurements of three months (March–May 2009) corresponding to premonsoon season, while in the present study we have extended the analysis using data for the whole year. Figure 4 shows the variation of

*C*

_{D}with

*ζ*along with the MOST predicted forms of

*C*

_{D}for different classes of similarity functions

*φ*

_{m}. The bin-averaged values of

*C*

_{D}are found to increase with increasing instability in DNS and DNS–DCS transition layers and attain a peak at

*ζ*≈ −0.12 (Fig. 4). A decreasing trend of

*C*

_{D}is observed from its peak value with increasing instability in DNS–FCS transition to FCS sublayers similar to that found by Srivastava and Sharan (2015). The average values of

*C*

_{D}in different sublayers are also found to be consistent to those reported by Srivastava and Sharan (2015) for limited dataset of three months. According to MOST,

*C*

_{D}is parameterized using Eq. (8). Equivalently one can write Eq. (9) in the form

*C*

_{DN}is the neutral drag coefficient. In Fig. 4, the continuous lines, showing the predicted variation of

*C*

_{D}for different forms of similarity functions, are plotted using

*C*

_{DN}obtained from the observational data. Figure 4 suggests that all the formulations agree well for the values of

*C*

_{D}in the range −0.1 <

*ζ*< 0, that is, in near-neutral to weakly unstable condition. With class I, II, and III functions,

*C*

_{D}shows monotonically increasing behavior in the full range of instability, while with class IV function,

*C*

_{D}shows a decreasing behavior after attaining a peak at around

*ζ*= −2. Notice that the observational data also suggest the existence of nonmonotonic behavior of

*C*

_{D}with

*ζ*. However, the point at which

*C*

_{D}attains a peak is

*ζ*= −0.1, which lies in DCS sublayer rather than in FCS sublayer as predicted by class IV function in its original form. To address this issue, we have empirically updated the class IV

*φ*

_{m}function in such a way that point at which

*C*

_{D}attains its maximum value lies in the DCS sublayer as found in the observational data. The modified functional form for

*φ*

_{m}is given by

*φ*

_{m}, the observed nature of

*C*

_{D}with

*ζ*in the unstable atmospheric conditions is well captured by MOST (Fig. 4). However, the suitability of the new form of

*φ*

_{m}for estimating both momentum and heat fluxes in the models by bulk flux algorithm requires additional analysis of neutral transfer coefficients for heat and moisture and corresponding stability correction functions. Due to the limitation of data, at present, we are unable to evaluate the impact of the new form of

*φ*

_{m}in estimating the transfer coefficients of heat and moisture from the bulk algorithm.

Variation of drag coefficient *C*_{D} with stability parameter *ζ* observed for the whole year 2009 over Ranchi (India) is shown with red markers. The mean values of *C*_{D} in each unstable sublayers are shown with black markers along with standard deviations in the form of error bars. Depending upon the data availability, two to three bins of equal width are chosen in each sublayer. The background color scheme corresponds to different unstable sublayers (Kader and Yaglom 1990; Bernardes and Dias 2010), i.e., dynamic sublayer (light gray) to free-convective sublayer (dark gray). The continuous curves represent the predicted variation of *C*_{D} using different classes (I–IV) of stability correction functions. The continuous black curve shows the predicted variation of *C*_{D} with the updated functional form of *φ*_{m} based on Kader and Yaglom (1990). For plotting each continuous curve, the value of the neutral transfer coefficient (*C*_{DN}) is taken as an average value *C*_{DN} in the dynamic sublayer. The dotted vertical line shows *ζ* = *ζ*_{m} = −0.1.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

Variation of drag coefficient *C*_{D} with stability parameter *ζ* observed for the whole year 2009 over Ranchi (India) is shown with red markers. The mean values of *C*_{D} in each unstable sublayers are shown with black markers along with standard deviations in the form of error bars. Depending upon the data availability, two to three bins of equal width are chosen in each sublayer. The background color scheme corresponds to different unstable sublayers (Kader and Yaglom 1990; Bernardes and Dias 2010), i.e., dynamic sublayer (light gray) to free-convective sublayer (dark gray). The continuous curves represent the predicted variation of *C*_{D} using different classes (I–IV) of stability correction functions. The continuous black curve shows the predicted variation of *C*_{D} with the updated functional form of *φ*_{m} based on Kader and Yaglom (1990). For plotting each continuous curve, the value of the neutral transfer coefficient (*C*_{DN}) is taken as an average value *C*_{DN} in the dynamic sublayer. The dotted vertical line shows *ζ* = *ζ*_{m} = −0.1.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

Variation of drag coefficient *C*_{D} with stability parameter *ζ* observed for the whole year 2009 over Ranchi (India) is shown with red markers. The mean values of *C*_{D} in each unstable sublayers are shown with black markers along with standard deviations in the form of error bars. Depending upon the data availability, two to three bins of equal width are chosen in each sublayer. The background color scheme corresponds to different unstable sublayers (Kader and Yaglom 1990; Bernardes and Dias 2010), i.e., dynamic sublayer (light gray) to free-convective sublayer (dark gray). The continuous curves represent the predicted variation of *C*_{D} using different classes (I–IV) of stability correction functions. The continuous black curve shows the predicted variation of *C*_{D} with the updated functional form of *φ*_{m} based on Kader and Yaglom (1990). For plotting each continuous curve, the value of the neutral transfer coefficient (*C*_{DN}) is taken as an average value *C*_{DN} in the dynamic sublayer. The dotted vertical line shows *ζ* = *ζ*_{m} = −0.1.

Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0350.1

## 5. Issues and limitations

From the application point of view, two major points of concern are 1) “equivalent” applicability of three sublayer model for heat exchange properties and 2) the treatment of both the roughness length of heat (*z*_{h}) and momentum (*z*_{0}) (or neutral transfer coefficients *C*_{D} and *C*_{H}), which has not been analyzed in the present study due to limitation of data. Here, we have assumed that apart from the uncertainty due to the form of similarity function used in the bulk flux algorithm, all the other uncertainties have the same magnitude for each of the four classes considered here for the analysis. However, while establishing the superiority of one bulk flux algorithm over the other in terms of comparison of estimated fluxes with those observed, the uncertainty associated with the way the roughness lengths are treated additionally might play a crucial role in the overall uncertainty in the estimated fluxes. Generally, the values of *z*_{0} and *z*_{h} are inversely derived from observations in field experiments or empirically estimated for practical applications (Yang et al. 2008) and these values deviate from each other considerably (Beljaars and Holtslag 1991). Numerical models have rather simplified approaches to parameterize *z*_{0} and *z*_{h} over different underlying surfaces. For example, depending upon the surface characteristics fixed values of *z*_{0} and *z*_{h} are assigned in the numerical models (Stull 1988), while over the open water *z*_{0} is parameterized using the Charnock relation (Charnock 1955). When the surface consists of a mix of sea ice and open water, a classical mosaic approach (Vihma 1995) is utilized to estimate effective values of *z*_{0} and *z*_{h}. Although significant progress has been made, accurate prescription of “neutral” transfer coefficient of heat, momentum, and moisture (equivalently roughness lengths for heat, momentum, and moisture) over various types of surfaces still presents challenges for numerical models leading to uncertainties in the estimated fluxes. In addition, uncertainty in the estimated fluxes might be due to the errors in the input parameters as well as the limitation of the bulk algorithm itself. For example, Blanc (1985) presented an alarming review of the large scheme-to-scheme differences in estimated fluxes of heat, momentum, and moisture from more than 10 bulk flux algorithms. Blanc (1983) also performed an analytical error analysis of estimated fluxes due to errors in the input parameters and found typical errors of 100% and 60% in sensible heat and momentum fluxes, respectively. Also, the current approach of analyzing the variation of *C*_{D} with respect to stability parameter *ζ* or Ri_{B} in free-convection limit has theoretical limitation. Based on the viscous generalization of the O’KEYPS equation, Grachev (1990) has suggested that in the free-convective conditions developed over the aerodynamically smooth surface, the drag coefficient *C*_{D} depends upon three additional parameters—the molecular coefficient of kinematic viscosity, the thermal diffusivity, and sensible heat flux apart from stability parameter *ζ* and roughness length*z*_{0}. However, from an application point of view, *C*_{D} is still parameterized as a function of *ζ* in the numerical models.

## 6. Conclusions

In the present study, a theoretical analysis is carried out to highlight the uncertainty in the estimation of surface fluxes depending upon the functional form of similarity functions *φ*_{m} and *φ*_{h} being utilized in the bulk parameterization of surface fluxes under unstable conditions. The bulk flux algorithm, in its lowest level of complexity, is selected for estimation of transfer coefficients of heat and momentum as a representative of heat and momentum fluxes using four different classes of functions *φ*_{m} and *φ*_{h}. These classes are identified based on the functional behavior of *φ*_{m} and *φ*_{h} following different power laws. The differences in the estimated values of fluxes from the first three classes of functions are found to be reasonably small as compared to the possible uncertainty in the prescription of input parameters in the bulk flux algorithm. However, the corresponding differences in the estimated fluxes with the class IV functions are very large as compared to those obtained with the other three classes of functions. The theoretical analysis presented here suggests a large deviation in the values of estimated fluxes if different forms of stability correction functions are utilized for the estimation of surface fluxes from the bulk flux algorithm.

Out of these four classes, functions lying in the first three classes satisfy the general assumption that under unstable conditions, similarity functions *φ*_{m} and *φ*_{h} are monotonically decreasing functions of the Monin–Obukhov stability parameter *ζ* in the whole range of values *ζ*. Consequently, these three functional forms predict a monotonically increasing behavior of transfer coefficients *C*_{D} and *C*_{H} with *ζ*. However, the functional form derived from the sublayer model suggested by Kader and Yaglom (1990) shows a nonmonotonic nature of similarity function for momentum and resulting in a nonmonotonic variation of drag coefficient *C*_{D} with *ζ*. This nonmonotonic nature of *C*_{D} was first reported by Srivastava and Sharan (2015) who argued that MOST, with commonly used similarity functions, is unable to capture the observed behavior of drag coefficient in moderately to very unstable conditions.

The existence of unusual nonmonotonic variation of *C*_{D} versus *ζ* found in case of functions suggested by Kader and Yaglom (1990) was largely remained unnoticed before Srivastava and Sharan (2015). This is possibly due to the fact that term *ψ*_{m}(*ζ*_{0}), where *ζ*_{0} = (*z*_{0}/*z*)*ζ* in MOST equations, is mostly neglected in the bulk formulation of surface fluxes (Zeng et al. 1998), which leads to shifting of point of inflection (the point at which the nature of the *C*_{D} vs *ζ* curve changes from increasing to decreasing) toward the higher magnitude of *ζ* beyond *ζ* ~ −10 for the surfaces with lower roughness lengths. However, this term becomes important in convective conditions and over the surface with high roughness lengths where the point of inflection tends to shift toward the lower magnitude of *ζ* with increasing surface roughness length *z*_{0}.

The turbulence observations presented here suggest that functions lying in class IV are relatively more consistent with the observational behavior of drag coefficient in moderately to very unstable conditions over tropics. However, to the authors’ knowledge, the applicability of class IV functions is yet not tested in any weather forecast models. Further, a careful analysis of turbulence data in free-convective conditions is required for determining the consistent functional form of similarity functions to avoid the large uncertainties in parameterized fluxes in the atmospheric models.

## Acknowledgments

We thank the anonymous reviewers for their helpful comments. This work is partially supported by J C Bose Fellowship to M.S. from Department of Science and Technology (DST)-Science and Education Research Board (SERB), Government of India (SB/S2/JCB-79/2014). P.S. was supported by Inspire Faculty Fellowship from DST, Government of India (DST/INSPIRE/04/2019/003125). The authors declare no competing interests.

## Data availability statement

The raw turbulence data for Ranchi (India) site used in this study can be obtained from the Indian National Centre for Ocean Information Service (http://www.incois.gov.in/portal/datainfo/ctczdata.jsp) upon request.

## REFERENCES

Akylas, E., and M. Tombrou, 2005: Interpolation between Businger–Dyer formulae and free convection forms: A revised approach.

,*Bound.-Layer Meteor.***115**, 381–398, https://doi.org/10.1007/s10546-004-1426-3.Beljaars, A. C. M., and A. A. M. Holtslag, 1991: Flux parameterization over land surfaces for atmospheric models.

,*J. Appl. Meteor.***30**, 327–341, https://doi.org/10.1175/1520-0450(1991)030<0327:FPOLSF>2.0.CO;2.Bernardes, M., and N. L. Dias, 2010: The alignment of the mean wind and stress vectors in the unstable surface layer.

,*Bound.-Layer Meteor.***134**, 41–59, https://doi.org/10.1007/s10546-009-9429-8.Blanc, T. V., 1983: An error analysis of profile flux, stability, and roughness length measurements made in the marine atmospheric surface layer.

,*Bound.-Layer Meteor.***26**, 243–267, https://doi.org/10.1007/BF00121401.Blanc, T. V., 1985: Variation of bulk-derived surface flux, stability, and roughness results due to the use of different transfer coefficient schemes.

,*J. Phys. Oceanogr.***15**, 650–669, https://doi.org/10.1175/1520-0485(1985)015<0650:VOBDSF>2.0.CO;2.Businger, J. A., 1966: Transfer of heat and momentum in the atmospheric boundary layer.

*Proceedings of the Symposium on the Arctic Heat Budget and Atmospheric Circulation*, J.O. Fletcher, Ed., The Rand Corporation, 305–332.Businger, J. A., 1988: A note on the Businger-Dyer profiles.

,*Bound.-Layer Meteor.***42**, 145–151, https://doi.org/10.1007/BF00119880.Businger, J. A., J. C. Wyngaard, Y. Izumi, and E. F. Bradley, 1971: Flux profile relationships in the atmospheric surface layer.

,*J. Atmos. Sci.***28**, 181–189, https://doi.org/10.1175/1520-0469(1971)028<0181:FPRITA>2.0.CO;2.Carl, D. M., T. C. Tarbell, and H. A. Panofsky, 1973: Profiles of wind and temperature from towers over homogeneous terrain.

,*J. Atmos. Sci.***30**, 788–794, https://doi.org/10.1175/1520-0469(1973)030<0788:POWATF>2.0.CO;2.Charnock, H., 1955: Wind stress over a water surface.

,*Quart. J. Roy. Meteor. Soc.***81**, 639–640, https://doi.org/10.1002/qj.49708135027.Deardorff, J. W., 1970: Convective velocity and temperature scales for the unstable planetary boundary layer and for Rayleigh convection.

,*J. Atmos. Sci.***27**, 1211–1213, https://doi.org/10.1175/1520-0469(1970)027<1211:CVATSF>2.0.CO;2.Delage, Y., and C. Girard, 1992: Stability functions correct at the free convection limit and consistent for both the surface and Ekman layers.

,*Bound.-Layer Meteor.***58**, 19–31, https://doi.org/10.1007/BF00120749.Dwivedi, A. K., S. Chandra, M. Kumar, S. Kumar, and N. V. P. K. Kumar, 2015: Spectral analysis of wind and temperature components during lightning in pre-monsoon season over Ranchi.

,*Meteor. Atmos. Phys.***127**, 95–105, https://doi.org/10.1007/s00703-014-0346-0.Dyer, A. J., 1974: A review of flux-profile relationships.

,*Bound.-Layer Meteor.***7**, 363–372, https://doi.org/10.1007/BF00240838.Ellison, A. B., 1957: Turbulent transport of heat and momentum from an infinite rough plane.

,*J. Fluid Mech.***2**, 456–466, https://doi.org/10.1017/S0022112057000269.Fairall, C. W., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young, 1996: Bulk parameterization of air-sea fluxes for Tropical Ocean-Global Atmosphere Coupled-Ocean Atmosphere Response Experiment.

,*J. Geophys. Res.***101**, 3747–3764, https://doi.org/10.1029/95JC03205.Fairall, C. W., E. F. Bradley, J. E. Hare, A. A. Grachev, and J. B. Edson, 2003: Bulk parameterization of air–sea fluxes: Updates and verification for the COARE algorithm.

,*J. Climate***16**, 571–591, https://doi.org/10.1175/1520-0442(2003)016<0571:BPOASF>2.0.CO;2.Foken, T., 2006: 50 years of the Monin–Obukhov similarity theory.

,*Bound.-Layer Meteor.***119**, 431–447, https://doi.org/10.1007/s10546-006-9048-6.Giorgi, E., and Coauthors, 2012: RegCM4: Model description and preliminary tests over multiple CORDEX domains.

,*Climate Res.***52**, 7–29, https://doi.org/10.3354/cr01018.Godfrey, J. S., and A. C. M. Beljaars, 1991: On the turbulent fluxes of buoyancy, heat, and moisture at the air-sea interface at low wind speeds.

,*J. Geophys. Res.***96**, 22 043–22 048, https://doi.org/10.1029/91JC02015.Grachev, A. A., 1990: Friction law in the free-convection limit.

*Izv. Akad. Nauk SSSR, Ser. Fiz. Atmos. Okeana*,**26**, 837–846.Grachev, A. A., C. W. Fairall, and S. E. Larsen, 1998: On the determination of the neutral drag coefficient in the convective boundary layer.

,*Bound.-Layer Meteor.***86**, 257–278, https://doi.org/10.1023/A:1000617300732.Grachev, A. A., C. W. Fairall, and E. F. Bradley, 2000: Convective profile constants revisited.

,*Bound.-Layer Meteor.***94**, 495–515, https://doi.org/10.1023/A:1002452529672.Högström, U., 1988: Non-dimensional wind and temperature profiles in the atmospheric surface layer: A re-evaluation.

,*Bound.-Layer Meteor.***42**, 55–58, https://doi.org/10.1007/BF00119875.Högström, U., 1996: Review of some basic characteristics of the atmospheric surface layer.

,*Bound.-Layer Meteor.***78**, 215–246, https://doi.org/10.1007/BF00120937.Jimenez, P. A., J. Dudhia, J. F. Gonzalez-Rouco, J. Navarro, J. P. Montavez, and E. Gracia-Bustamante, 2012: A revised scheme for the WRF surface layer formulation.

,*Mon. Wea. Rev.***140**, 898–918, https://doi.org/10.1175/MWR-D-11-00056.1.Kader, B. A., 1988: Three-layer structure of an unstably stratified atmospheric surface layer.

*Izv. Akad. Nauk SSSR, Ser. Fiz. Atmos. Okeana*,**24**, 1235–1250.Kader, B. A., and G. Perepelkin, 1989: Effect of the unstable stratification on wind and temperature profiles in the surface layer.

*Izv. Akad. Nauk SSSR, Ser. Fiz. Atmos. Okeana*,**25**, 787–795.Kader, B. A., and A. M. Yaglom, 1990: Mean fields and fluctuation moments in unstably stratified turbulent boundary layers.

,*J. Fluid Mech.***212**, 637–662, https://doi.org/10.1017/S0022112090002129.Kazansky, A. B., and A. S. Monin, 1956: Turbulence in the inversion layer near the surface.

,*Izv. Akad. Nauk SSSR, Ser. Geofiz.***1**, 79–86.Monin, A. S., and A. M. Obukhov, 1954: Osnovnye zakonomernosti turbulentnogo peremeshivanija v prizemnom sloe atmosfery (Basic laws of turbulent mixing in the atmosphere near the ground).

*Tr. Geofiz. Inst. Akad. Nauk SSSR*,**24**, 163–187.Obukhov, A. M., 1971: Turbulence in an atmosphere with a non-uniform temperature.

,*Bound.-Layer Meteor.***2**, 7–29, https://doi.org/10.1007/BF00718085.Panofsky, H. A., 1961: An alternative derivation of the diabatic wind profile.

,*Quart. J. Roy. Meteor. Soc.***87**, 109–110, https://doi.org/10.1002/qj.49708737113.Sellers, W. D., 1962: Simplified derivation of the diabatic wind profile.

,*J. Atmos. Sci.***19**, 180–181, https://doi.org/10.1175/1520-0469(1962)019<0180:ASDOTD>2.0.CO;2.Sharan, M., and P. Srivastava, 2016: Characteristics of heat flux in the unstable atmospheric surface layer.

,*J. Atmos. Sci.***73**, 4519–4529, https://doi.org/10.1175/JAS-D-15-0291.1.Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., https://doi.org/10.5065/D68S4MVH.

Srivastava, P., and M. Sharan, 2015: Characteristics of drag coefficient over a tropical environment in convective conditions.

,*J. Atmos. Sci.***72**, 4903–4913, https://doi.org/10.1175/JAS-D-14-0383.1.Srivastava, P., and M. Sharan, 2017: An analytical formulation of the Obukhov stability parameter in the atmospheric surface layer under unstable conditions.

,*Bound.-Layer Meteor.***165**, 371–384, https://doi.org/10.1007/s10546-017-0273-y.Stull, R., 1988:

*An Introduction to Boundary Layer Meteorology*. Kluwer Academic, 666 pp.Vihma, T., 1995: Subgrid parameterization of surface heat and momentum fluxes over polar oceans.

,*J. Geophys. Res.***100**, 22 625–22 646, https://doi.org/10.1029/95JC02498.Yamamoto, G., 1959: Theory of turbulent transfer in non-neutral conditions.

,*J. Meteor. Soc. Japan***37**, 60–67, https://doi.org/10.2151/jmsj1923.37.2_60.Yang, K., and Coauthors, 2008: Turbulent flux transfer over bare-soil surfaces: Characteristics and parameterization.

,*J. Appl. Meteor. Climatol.***40**, 276–290, https://doi.org/10.1175/2007JAMC1547.1.Yang, K., N. Tamai, and T. Koike, 2001: Analytical solution of surface layer similarity equations.

,*J. Appl. Meteor.***40**, 1647–1653, https://doi.org/10.1175/1520-0450(2001)040<1647:ASOSLS>2.0.CO;2.Zeng, X., M. Zhao, and R. E. Dickinson, 1998: Intercomparison of bulk aerodynamic algorithms for the computation of sea surface fluxes using TOGA COARE and TAO data.

,*J. Climate***11**, 2628–2644, https://doi.org/10.1175/1520-0442(1998)011<2628:IOBAAF>2.0.CO;2.