Cloud seeding operations for precipitation enhancement have been performed in many countries. China invests the most resources and carries out such operations every year for its vast agriculture needs ( Zheng et al., 2003). It is not possible to analyze the characteristics of seeded and unseeded clouds directly; numerical cloud models have been found to be the best tools for theoretical studies of cloud seeding for rain enhancement. Among these models, one-dimensional (1D) cloud models with parameterized or detailed microphysics schemes are convenient and efficient for evaluating rain enhancement effect of cloud seeding ( Hu et al., 1983; Orville and kopp, 1990; Guo et al., 1999; Liu and Niu, 2010). Traditional 1D cloud models, which use a prescribed vertical airflow and a constant sounding profile as the environment conditions, are not suitable for reproducing realistic precipitation processes; thus, their use is limited in the quantitative estimation of cloud seeding effects.

Over the past several decades, many three-dimensional (3D) cloud models and mesoscale numerical weather prediction (NWP) models with sophisticated cloud seeding parameterization schemes have been developed ( Levin et al., 1997; Chen, 2002; Ćurić et al., 2006; Guo et al., 2006; Fang and Guo, 2007; Straka et al., 2010). However, a mesoscale cloud-resolving model, such as the Weather Research and Forecasting (WRF) model, with a sophisticated and cloud seeding microphysical scheme, is too computationally expensive to be of any practical use in cloud seedability (cloud seeding potential; Lou et al., 2012) forecasting and in aiding the decision-making process in cloud seeding operations.

In a cloud seeding operation, determining the optimal seeding location, agent amount, and seeding time is critical to achieving the maximum rain enhancement rate. Combining a dynamically sophisticated and computing- intensive NWP model such as the WRF model with a computationally efficient 1D Stratiform Cold cloud model (1DSC), having a detailed microphysics scheme, to develop a probabilistic-based operational dual-model cloud- seeding system can be an innovative approach to provide a reliable model prediction of cloud seeding potential.

A precipitating stratiform cloud system was observed on 4-5 July 2004 (case 040705 hereafter), associated with a Northeast China cyclone. The detailed features of the precipitation system were described by Hu et al. (2007) in their work. In this case study, WRF-1DSC was integrated for 15 h, starting from 2000 UTC 4 July to 1100 UTC 5 July 2004, overlapping with the observed precipitation process at the Changchun station.

As shown in Fig. 1, the WRF-1DSC simulation resulted in realistic time evolution of surface precipitation rate in appropriate quantities and at the right time, which is essential for calculating rain enhancement rate in cloud seeding experiments. Due to the lack of data on vertical velocity observations, updraft profile in the traditional 1DSC is often determined by trial and error from the comparison between simulated results and observed radar echo structures. In contrast, in WRF-1DSC, as illustrated in Fig. 2, updrafts became dominant in mid and lower troposphere from 1800 UTC 4 July 2004 to 0500 UTC 5 July 2004, with the maximum vertical velocities surpass ing 12 cm s^-1 between 2 and 8 km above ground level (AGL) after 2000 UTC 4 July 2004, which was the starting point of precipitation.

As shown in Figs. 3a and 3b, ice particles and supercooled water droplets coexisted above the 0°C layer in the cold cloud. The -5°C layer was located at about 5 km AGL and the -10°C layer at around 6 km AGL. In addition, the liquid water content reached a maximum value of 0.5 g m^-3 at around 4.5-6 km AGL between 2200 and 2300 UTC 4 July 2004 (Fig. 3b). In our study, seeding height was chosen as 5.5 km AGL on the basis of the following criteria established in previous researches (e.g., Reisin et al., 1996; Mather et al., 1997; Yin et al., 2000; Ćurić et al., 2006; Zhao and Lei, 2010): (1) the activation temperature for AgI should be between -5 and -10°C; (2) the seeding region should be dominated by stronger updrafts; and (3) the cloud layer should have maximum content of supercooled liquid water; in our study, we considered seeding heights in the range of 5.5±0.5 km for comparison. According to Reisin et al (1996), the optimum time for increasing the total rain amounts coincides with the time of the natural ice formation. As shown in Fig. 3b, the number concentration of cloud ice attained a maximum of 8 L^-1 after 2300 UTC. Given these facts, seeding was applied between 5.0 and 6.0 km at 2300 UTC 4 July 2004. The sensitivity tests performed for AgI seeding effects are summarized in Table 1. We investigated the seeding effects mainly by varying the seeding amount (from 1 to 100 g); we also observed the effects at different seeding heights (5-6 km) and seeding times. In all experiments, the cloud was seeded at the cold parts, and X_s represented the maximum AgI mixing ratio, which decreased as the seeding agent moved.

In WRF-1DSC, rainwater was initiated by autoconversion of cloud droplets into raindrops, melting of precipitating ice crystals (e.g., ice, snow, and graupel), or collisions among ice crystals at temperatures higher than 0°C. For this case, seeding rate was fixed at 0.6 g s^-1. Results of these sensitivity tests indicated that the amount of AgI required for seeding was a critical factor for obtaining positive results (rain increase). A change in the amount of seeding agent altered the quantity of accumulated surface rain, and the greatest augmentation in precipitation amount (5.61%) was achieved when the cloud was seeded at 2300 UTC, with the maximum AgI mixing ratio X_s = 15 ng kg^-1. The seeding effect increased with the amount of seeding agent until it reached a maximum (15 ng kg^-1), and then it decreased. On the other hand, seeding with a maximum AgI mixing ratio of over 75 ng kg^-1 was found to reduce the total precipitation amount. When the modeled cloud was seeded with a maximum X_s of 100 ng kg^-1, the accumulated precipitation at the surface decreased by as much as -12.42%. The time of seeding is also very important. The finding that seeding during the time (2300 UTC) when updraft attained its maximum value (23.7 cm s^-1) provided the best seeding effect (resulting in an increase in the precipitation amount by 5.61%) might imply that vertical air motions are highly correlated to the quantity of precipitation that a seeded cloud may produce. Seeding at an earlier stage (2200 UTC 4 July 2004) reduced the total rain (-2.47%), and that at a later stage (0100 UTC 5 July 2004) increased the total rain by only 3.92%. As shown in Table 1, although seeding below or above 5.5 km AGL also resulted in rain enhancement, seeding at 5.5 km was found to be more effective. This height corresponded to the maximum value of supercooled liquid water content, as shown in Fig. 3a; this finding was consistent with those of previous studies, which proposed that the maximum liquid water content plays an important role in cloud seeding ( Ćurić et al., 2006).

In order to evaluate seeding effects produced by different amounts of seeding agent, we compared the results of seeding experiments for maximum X_s = 15 ng kg^-1 and X_s = 100 ng kg^-1, keeping all other conditions same as in previous experiments. Figure 4 shows the total accumulated precipitation on the ground as a function of time. The total rainfall on the ground increased for maximum X_s= 15 ng kg^-1, but decreased for its other value. This might be because ice crystals need available water vapor to grow large enough to fall through the cloud and collide with supercooled liquid water that freeze onto them as a result of the impact, creating larger crystals. Adding excess number of ice particles would lead to water vapor competition. The lack of sufficient supercooled liquid water would repress the growing process of ice particles. On the other hand, proper seeding can assist in the conversion of the supercooled liquid water into precipitating ice when natural ice nuclei are very scarce for an effective conversion. In short, an appropriate seeding amount would enhance the total accumulated rain while overseeding would reduce it.

These results demonstrate WRF-1DSC’s capability to simulate the effects of seeding on rain formation. Ignoring downdrafts and with an unchanging environment, the clouds modeled by the traditional 1DSC model ( Hu et al., 2007) always reach a steady state after a spin-up period, and their life spans are infinite. It is obvious that the addition of WRF forcing produces a solution for lacking of observed atmosphere. With realistic soundings and vertical velocities, WRF-1DSC was able to simulate quantitatively the surface rain enhancement and reduction rate, respectively. Compared to the traditional 1DSC, WRF- 1DSC could capture the distribution of precipitation in a more realistic way. The modeling results presented here also provide evidence that environmental variables such as vertical velocities and liquid water content play important roles in determining the seeding effect, and that better modeling of these environmental variables is instrumental in assessment of seeding effects.

Figure 1

Figure 1 Observed (black-colored) and simulated (gray-colored) 1-h surface precipitation histogram (units: mm h-1) and accumulated rainfall curve (units: mm) at Changchun station.

Figure 2

Figure 2 Hourly evolutions of the vertical velocity (units: cm s-1) derived from WRF for 1300 UTC 4 July-1200 UTC 5 July 2004.

Figure 3

Figure 3 Time-height cross-section of (a) mass content of liquid cloud water (units: g m-3) and (b) number concentration of cloud ice (units: L-1).

Table 1

Table 1

Table 1 Seeding effects estimated using different amounts of AgI at diverse times. Maximum AgI mixing ratio ( Xs: ng kg-1) Seeding time (UTC) Seeding height (km) Accumulated rain (mm) Enhancement rate (%) —— —— —— 6.012 —— 1.0 2300 (4 July 2004) 5.5 6.131 1.98 5 2300 5.5 6.181 2.82 9 2300 5.5 6.203 3.19 15 2300 5.5 6.348 5.61 30 2300 5.5 6.150 2.31 50 2300 5.5 6.108 1.12 75 2300 5.5 5.652 -5.99 100 2300 5.5 5.265 -12.42 15 2200 5.5 5.863 -2.47 15 0100 (5 July 2004) 5.5 6.247 3.92 15 2300 5.0 6.293 4.66 15 2300 6.0 6.219 3.44

Table 1 Seeding effects estimated using different amounts of AgI at diverse times.

Figure 4

Figure 4 Total accumulated precipitation (units: mm) as a function of time as estimated with the WRF-1DSC model.

In this study, a case of stratiform cloud over the Jilin Province in China was simulated with a dual-model system (WRF-1DSC) to examine the cloud seeding effects in terms of enhancing rainfall. Compared to the original version of the 1DSC model, WRF-1DSC was able to reproduce better the generation and dissipation of natural clouds. Forced by changing environment conditions, which were derived from the mesoscale WRF forecast outputs, WRF-1DSC was found suitable to be employed for evaluating the cloud seeding potential in realistic terms. In this case, we found that the optimal seeding strategy depended not only on the seeding amount, but also on the seeding height and seeding time, which were closely related to the environment conditions. The optimal seeding agent mixing ratio was observed to be 15 ng kg^-1, with which the modeled cloud was capable of producing 5.61% more precipitation. On the other hand, seeding the cloud with a maximum AgI mixing ratio exceeding 75 ng kg^-1 was found to decrease the total rainfall; the greatest reduction in the accumulated rain (-12.42%) occurred for an overseeded cloud when the maximum AgI mixing ratio attained a value of 100 ng kg^-1. The optimal seeding height was found to be the altitude of maximum cloud water content, and the ideal seeding time was 180 min of simulated time when the vertical speed reached its maximum value near the seeding height, implying that the vertical motion may play an important role in determining the optimal seeding time. Although a single case study cannot provide enough evidence to form a generalized conclusion in a statistical sense, our study was able to support the claim that WRF-1DSC has potential for determining the optimal AgI seeding strategy in the practical operations of precipitation enhancement.

This work was just the first step of a larger research project on applying WRF-1DSC to test the effects of seeding stratiform clouds, and further investigations need to be conducted in the associated physical processes. For example, we need to further examine the impact of the WRF-driven procedure on the seeding agent diffusion in 1DSC, as this is essential for understanding the influence of environmental fields on the seeding process. Moreover, the dynamic effects of the transformation from liquid saturation to ice saturation on the simulated clouds need to be considered in future studies, and more detailed physical processes, such as immersion freezing and sorption nucleation, need to be taken into account in numerical experiments for a better understanding of the seeding effects of AgI.