Impact of Aerosols on NPP in Basins: Case Study of WRF−Solar in the Jinghe River Basin

Abstract

Aerosols impact vegetation productivity by increasing diffuse radiation and changing temperature and humidity conditions. In this study, climate simulations of the Jinghe River Basin in 2020 based on aerosol and aerosol−free scenarios were carried out using the control variable method and the aerosol optical depth parameter as the external input data of Weather Report Forecast (WRF)−solar. These two output results were used as input data for the Carnegie Ames Stanford Approach (CASA) model to calculate the impact of aerosols on vegetation productivity. The results showed that WRF−solar accurately simulated changes in meteorological factors such as temperature, rainfall, solar radiation, and relative humidity in the Jinghe River Basin, with a correlation coefficient above 0.85. Aerosols significantly change the ratio of diffuse to direct radiation, act as a cooling function to reduce temperature, and affect rainfall by interacting with clouds. The scenario simulation results showed that under the influence of aerosols, the total solar radiation was reduced by 224.98 MJ/m², accounting for 3.44% of the total annual radiation. Correspondingly, the average net primary productivity of vegetation in the Jinghe River Basin in 2020 decreased by 26.64 gC/m², which was not conducive to vegetation photosynthesis and carbon fixation in the basin.

1. Introduction

Solar radiation is the fundamental source of energy on Earth. Atmospheric aerosols affect the amount of surface radiation and radiation composition through absorption and scattering and affect the radiation characteristics of clouds [1]. The presence of aerosols can enhance the diffuse radiation of solar radiation, and the proportion of effective photosynthetic radiation can be increased by appropriately increasing the diffuse radiation [2], thereby improving the vegetation productivity of terrestrial ecosystems, which is called diffuse radiation fertilization. At the same time, atmospheric aerosol particles, as a basic element to change the energy budget of earth–air radiation and the formation of cloud droplets [3], affect temperature and humidity changes by absorbing and scattering solar radiation, and indirectly have environmental effects on vegetation photosynthesis. The impact of aerosols on vegetation productivity in terrestrial ecosystems cannot be ignored; however, the responses of ecosystems to aerosols differ under complex environmental conditions. Scholars have studied the impact of aerosol optical depth (AOD) on vegetation productivity [4,5]; however, most elaborated on the relationship between the two through correlation analysis without quantifying the change in vegetation productivity caused by aerosols. Williams et al. [6] proved the importance of controlling phenology when observing the changes in winter wheat gross primary productivity (GPP) caused by diffuse radiation through experiments. Environmental factors can modulate diffuse fertilization effects, so other environmental factors must be taken into account when quantifying the influence of aerosols [7]. Ref. [8] showed that although total radiation is reduced, the rate of photosynthesis is increased by increased diffuse reflection. However, Wolffe et al. [9] found that the reduction in ground radiation caused by particulate matter (PM) resulted in a reduction in production effect far exceeding the positive effect brought by the increase in the ratio of diffuse reflection. Kalina et al. [10] used the Weather Report Forecast (WRF) model to study the influence of aerosol concentration change on rainfall. These studies have shown that both direct and indirect radiation effects of aerosols can affect vegetation photosynthesis. However, it is often impossible to account for both responses simultaneously in these studies.

WRF−solar is an evolving mesoscale numerical model designed to meet the demand of irradiance forecasting [11]. WRF−solar can directly output direct normal irradiance (DNI) and indirect irradiance (DIF) components. Many scholars have applied WRF−solar to research solar energy prediction [12,13]. Ji et al. [14] proposed a multi−step short−term solar radiation prediction method based on the WRF−solar model, deep fully convolutional network, and chaotic Aquila optimization algorithm. WRF−solar offers different aerosol options that allow users to use homemade and local aerosol data to improve simulation accuracy [15]. Climate simulations with and without aerosols can be performed by turning the aerosol option on or off. Cheng et al. [16] used satellite AOD data and the WRF−solar model to verify the attenuation effect of aerosol on surface solar radiation. WRF−solar has a complete aerosol–cloud–radiation feedback mechanism [11]. WRF−solar uses a simplified representation of the interaction between aerosols and clouds in Thompson’s microphysics scheme [17] to enhance the interaction between aerosols and clouds, which changes the evolution of cloud and aerosol properties, making the simulation results more accurate. Liu et al. [18] found that the sensitivity of solar radiation to aerosol–cloud parameters simulated by WRF−solar was affected by AOD and cloud cover. The WRF−solar model has been shown to have good accuracy in solar radiation simulation.

Net primary productivity (NPP) is the total quantity of organic matter generated by photosynthesis per unit time per unit area of plants after deducting autotrophic respiration, that is, the net increase in biomass obtained [19]. As an important factor in determining the quality of ecosystems and carbon sinks [20], NPP reflects the productive capacity and ecological processes of vegetation communities, which are of great significance for regulating the global carbon balance and enhancing ecological service functions. The main NPP estimation models can be divided into statistical, process, and parameter models [21]. CASA is a light energy utilization process model that combines the regulatory factors of an NPP in a relatively simple manner. This model is simple and practical and can obtain an important parameter, the photosynthetically active radiation absorption ratio (FPAR), from remote sensing data. Thus, it is widely used [22]. Jia et al. [23] used CASA to quantify the vegetation NPP of the Ordos area from 2000 to 2019 and verified it using measured data. The NPP data derived from the inversion of the CASA model and the observed data were reasonably close to one another.

Industrial development has resulted in large amounts of anthropogenic aerosol emissions [24]. Human activities not only dominate urban and rural development but are also closely related to the vegetation ecological area which is the background. Therefore, there is an urgent need to understand the effects of aerosols on direct solar radiation, diffuse radiation, temperature, rainfall, and other meteorological factors to further clarify the process and mechanism of aerosol impact on terrestrial ecosystem productivity and the carbon budget. The Jinghe River Basin is located in the middle of the Loess Plateau, which is in a transitional zone from a sub−humid to a semi−arid climate with high vegetation coverage and rapid urban development. Therefore, it is a suitable research area to study the impact of AOD on vegetation productivity. Taking the Jinghe River Basin in 2020 as an example, this study uses AOD as the control variable to analyze the changes in diffuse radiation, temperature, and humidity environmental conditions caused by the direct and indirect radiation effects of aerosols and quantifies the impact of aerosols on the productivity of terrestrial ecosystems through the WRF−solar numerical model and the CASA model. This study aims to offer a reference for accurately simulating the productivity of terrestrial ecosystems and formulating countermeasures for vegetation adaptation to climate change.

2. Study Area and Data

2.1. Study Area

The Jinghe River is a secondary tributary of the Yellow River, which originates in Ningxia. The latitude and longitude range of Jinghe River Basin is between 106°14′E ~108°42′E and 34°46′N~37°19′N, and the total drainage area is 45,421 km². A diagram of the location of the Jinghe River Basin is shown in Figure 1a. The highest point in the Jinghe River Basin is located west of the Liupan Mountain cradle. The terrain gradually decreases in altitude from northwest to southeast and finally extends into the gentle Guanzhong Plain. The Jinghe River Basin has a typical temperate continental climate and is located in a transitional zone from a sub−humid to a semi−arid climate. The average maximum temperature can reach 24 °C in summer, and the average minimum temperature is about −8 °C in winter. The temperature in the basin gradually decreased from south to north under the influence of latitude. There are clear seasonal differences in the distribution of rainfall throughout the year, with more rain in summer and less rain in winter. Rainfall from June to September accounts for 70% of annual rainfall. As shown in Figure 1b, cultivated plants occupy the largest proportion of the vegetation area in the Jinghe River Basin, followed by grassland and broadleaf forest. The main land use in the Jinghe River Basin was cropland, accounting for 45.4%, followed by grassland and forestland, accounting for 32.33% and 19.91%, respectively. Forestland is mainly distributed in the Liupan Mountain area in the west and the Ziwuling forest zone in the east, as shown in Figure 1c.

2.2. Data Source and Preprocessing

2.2.1. Reanalysis Data

The climate model in this work was driven by NCEP FNL (Final) Operational Global Analysis data from the National Center for Atmospheric Research (https://rda.ucar.edu/ accessed on 29 March 2023). FNL data are provided in 1° × 1° grids that are updated on a six−hourly basis. This product was obtained from the Global Data Assimilation System (GDAS) and includes surface pressure, sea level pressure, geopotential height, temperature, sea surface temperature, soil value, ice cover, relative humidity, wind speed, and many other parameters. A number of studies [25,26] have applied FNL reanalysis data to climate simulation and achieved good simulation results.

2.2.2. Moderate−Resolution Imaging Spectroradiometer (MODIS) Data

MODIS is an important sensor mounted on the TERRA and AQUA satellites. The global observation data provided by the MODIS sensor has been widely used in remote sensing research, and the reliability of the data has been recognized by scholars at home and abroad [27,28]. The normalized difference vegetation index (NDVI), NPP, and AOD550 nm data used in this study were obtained from the NASA MODIS website (https://ladsweb.modaps.esodis.nasa.gov/, accessed on 29 March 2023). The pre−processing of the three MODIS datasets included data format conversion, projection conversion, and clipping. The NDVI uses MOD13Q1 data with a resolution of 250 m for 16 days and synthesizes the multi−scene MOD13Q1 data of the year into the monthly value NDVI of 12 months. To calculate the photosynthetically active radiation (APAR) in the CASA model, NDVI was resampled to a resolution of 1 km. NPP uses MOD17A3H data with a spatial resolution of 500 m and a temporal resolution of one year. The data were resampled to a resolution of 1 km for comparison with the vegetation NPP estimation results in this study. AOD550 nm uses MCD19A2 data, which is daily data with a resolution of 1 km. The AOD data of MCD19A2 are multiband data with both Terra and Aqua stars. There are many voids in these data due to clouds and snow. We selected quality−assured level 2.0 data from AERONET (https://aeronet.gsfc.nasa.gov/cgi-bin/webtool_aod_v3, accessed on 29 March 2023) as ground−truth data to validate the satellite data. The results showed that the correlation coefficient between MCD19A2 AOD data and AERONET AOD is 0.84. The effective values of the morning and afternoon stars were averaged and the effective values for each month were averaged to convert the AOD data into usable monthly averages. Meanwhile, according to the setting of the nested pattern in WRF−solar, the AOD data were resampled to 0.05° and 0.25° to create the WRF−solar aerosol external input files.

2.2.3. Other Data

The vegetation cover type was obtained from the Center for Resources and Environmental Sciences, Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 29 March 2023), with a resolution of 1 km. According to the data description, vegetation types were divided into seven categories: coniferous forests, broadleaf forests, cultivated plants, shrubs, grasslands, meadows, and tussocks. Land−use data were obtained from Global Land 30 (http://www.globallandcover.com/GLC30Download/index.aspx, accessed on 29 March 2023). The meteorological data used to verify the model results were obtained from the National Meteorological Center (https://data.cma.cn/, accessed on 29 March 2023), which contains daily site data, including temperature, rainfall, relative humidity, and sunshine duration. Meteorological stations in and around the study area in 2020 were selected, and those with missing values for more than 10 days per month were excluded. The pre−processing of meteorological data included removing outliers and synthesizing monthly data from daily data. See Figure 1 for the spatial locations of the meteorological stations. Downward shortwave radiation data from the Global Land Surface Satellite (GLASS, http://www.glass.umd.edu/Download.html, accessed on 29 March 2023) was also used to verify the model result. A DEM was obtained from a geographic information cloud (https://www.gscloud.cn/, accessed on 29 March 2023) with a resolution of 30 m. The DEM data processing includes projection transformation, mosaics, and clipping.

3. Methods

3.1. WRF−Solar Model

WRF is a mesoscale weather forecast model developed by the National Center for Atmospheric Research of the United States. This model has been successfully applied to data assimilation research [29,30], air quality modeling [31,32], and regional climate simulations, such as short−term rainfall forecasts [33] and typhoon simulation [34]. The WRF was dynamically scaled down using region nesting to improve the simulation resolution. The WRF−solar model is an updated version of the WRF model with improvements related to the solar radiation component. Unlike the WRF mode, which outputs only short−wave radiation, WRF−solar outputs global horizontal irradiance (GHI), DNI, and DIF at each time step. The AOD reflects the extinction effect of aerosol particles on solar radiation, which is an important optical parameter of aerosols. Studies have shown that the aerosol optical depth can significantly change the radiation composition and temperature and humidity meteorological factors [35]. At the same time, WRF−solar can input long−term satellite observation AOD550 nm data to replace the climate aerosol data for calculation.

In this study, WRF−solar version 4.2.2 was used for two sensitivity experiments. AOD was not added for the first time as a control test, which was called the WRF−solar group. Second, the monthly average AOD was added as external input data, forming the WRF−solar−AOD group. To avoid a large simulation error at the start time of the experiment, the simulation date was advanced by one month as the warm−up period. The two experiments adopted the parameter settings listed in

Table 1. The parameter settings of WRF−solar.

Double−Layer Nesting Scheme
Initial and lateral boundary conditions	FNL Data
Projection system	Lat−Lon
Regional center location	36.0°N, 107.4°E
Time step	120 s
Number of nested grids	d01: 38 × 38	d02: 71 × 76
Spatial resolution	d01: 0.25°	d02: 0.05°
Time resolution of output data	d01: 3 h	d02: 1 h
Physical Schemes
Longwave and shortwave radiation scheme	RRTMG
Surface layer scheme	Revised MM5 Monin–Obukhov
Cloud microphysical scheme	Thompson
Land surface process scheme	Noah LSM
Atmospheric boundary layer scheme	MYNN

. The output data were post−processed and analyzed using Python software.

3.2. CASA Model

The CASA model is based on light efficiency and is driven by NDVI, temperature, rainfall, solar radiation, and vegetation type to calculate the net primary vegetation productivity [36]. The CASA model has high precision, and the vegetation productivity simulation results obtained by using this model are similar to the measured results [37]. The NPP was calculated using the following formula [38]:

$$NPP\left(x,t\right)=APAR\left(x,t\right)\times \epsilon \left(x,t\right)$$

(1)

where $APAR\left(x,t\right)$ is photosynthetically active radiation and $\epsilon \left(x,t\right)$ is the actual light energy utilisation rate.

The formula for calculating photosynthetic active radiation absorbed by vegetation is as follows:

$$APAR\left(x,t\right)=SOL\left(x,t\right)\times FPAR\left(x,t\right)\times 0.5$$

(2)

where $SOL\left(x,t\right)$ is the global horizontal irradiation (MJ·m⁻²) including direct and indirect solar radiation and $FPAR\left(x,t\right)$ is the proportion of incident photosynthetically active radiation absorbed by vegetation canopy.

The formula for calculating the proportion of incident photosynthetic active radiation absorption (FPAR) by vegetation canopy is as follows:

$$FPAR\left(x,t\right)=min\left[\frac{SR-S{R}_{min}}{S{R}_{max}-S{R}_{min}},0.95\right]$$

(3)

where SR is the ratio vegetation index, which can be obtained according to the normalized vegetation index (NDVI), SR_min is 1.05, and the size of SR_max is related to the vegetation type.

The utilization rate of light energy is mainly affected by temperature and moisture, and is estimated as follows:

$$\epsilon \left(x,t\right)=T_{\epsilon 1}\left(x,t\right)\times T_{\epsilon 2}\left(x,t\right)\times W_{\epsilon }\left(x,t\right)\times {\epsilon }_{max}$$

(4)

where $T_{\epsilon 1}\left(x,t\right)$ and $T_{\epsilon 2}\left(x,t\right)$ are the influence coefficients of low−temperature and high−temperature stress, respectively $, W_{\epsilon }\left(x,t\right)$ is the influence coefficient of water stress, and ${\epsilon }_{max}$ is the maximum utilization rate of light energy under ideal condition (%).

3.3. Solar Radiation Calculation

Solar radiation is the basic energy source of the Earth, and it is an indispensable variable in ecosystem process simulations, hydrological simulations, and biophysical models; however, it is more difficult to obtain solar radiation variables than other meteorological elements due to the few numbers of radiation stations. Therefore, in the absence of optical radiation observation data, the classical method of optical radiation and sunshine duration is usually to carry out daily solar radiation simulation calculations by taking longitude, latitude, altitude, and sunshine duration data as inputs to provide the necessary parameters for the light utilization model. In this study, the daily solar radiation at the sites was calculated using the Angstrom–Prescott method [39]. The formula is as follows and more details can be found in references [40]:

$$H=H_L\times \left(a+b\times \frac{S}{S_L}\right)$$

(5)

where H is the total measured daily radiation, H_L is the total daily radiation under sunny conditions, S and S_L are sunshine duration and day length, respectively, and a and b are empirical parameters. The FAO suggestion was adopted in this study, and the values of a and b were 0.25 to 0.50.

According to existing studies, the transparency coefficient of the total radiation in the atmosphere is approximately 0.8 under normal circumstances, and the transparency coefficient varies under specific environmental conditions. The transparency coefficient of the total radiation in the atmosphere was set to 0.8:

$$H_L=H_0\times 0.8$$

(6)

$$H_0=\frac{1}{\pi }\times G_{sc}\times E_0\times \left(\left(\cos \Phi \times \cos \delta \times \cos W_s\right)+\frac{\pi }{180}\times \sin \Phi \times \sin \delta \times W_s\right)$$

(7)

where $G_{sc}$ is the solar constant, and its value is generally 1367 W·m⁻² (equivalent to 118∙108 MJ·m⁻²·d⁻¹);$E_0$ is the eccentricity correction factor of earth orbit; Φ is the latitude; δ is solar declination; and $W_s$ is time angle:

$$W_s={\cos }^{-1}\left(-\tan \Phi \times \tan \delta \right)$$

(8)

The time interval between sunrise and sunset was defined as $S_L$. Assuming that the sun’s altitude angle between sunrise and sunset is zero, the day length can be calculated as

$$S_L=\frac{2}{15}\times W_s$$

(9)

3.4. Model Evaluation

To determine the performance of the model, the results of the simulations were assessed using performance statistics such as root mean square error (RMSE), mean absolute error (MAE), correlation coefficient (R), and index of agreement (IA) with Equations (10)–(13), respectively.

$$RMSE=\sqrt{\frac{1}{n}{\displaystyle \sum }_{i=1}^n{\left(x_i-y_i\right)}^2}$$

(10)

$$MAE=\frac{1}{n}{\displaystyle \sum }_{i=1}^n\left|x_i-y_i\right|$$

(11)

$$R=\frac{{{\displaystyle \sum }}_{i=1}^n\left(x_i-\overline{x}\right)\left(y_i-\overline{y}\right)}{\sqrt{{{\displaystyle \sum }}_{i=1}^n{\left(x_i-\overline{x}\right)}^2{{\displaystyle \sum }}_{i=1}^n{\left(y_i-\overline{y}\right)}^2}}$$

(12)

$$IA=1-\frac{{{\displaystyle \sum }}_{i=1}^n{\left(x_i-y_i\right)}^2}{{{\displaystyle \sum }}_{i=1}^n{\left(\left|x_i-\overline{y}\right|+\left|y_i-\overline{y}\right|\right)}^2}$$

(13)

where $x_i$ is the simulated value, $y_i$ is the observation value, $\overline{x}$ is the average simulated values, and $\overline{y}$ is the average observation values.

4. Results

4.1. Spatial and Temporal Characteristics of Aerosol Optical Depth

Figure 2 shows the spatial and temporal distributions of the monthly mean AOD (AOD in Section 4.1 refers to AOD550 nm) in the Jinghe River Basin in 2020. The natural discontinuity method was used to grade the spatial distribution of the monthly mean AOD550 nm in the Jinghe River Basin in 2020. The spatial distribution of the AOD in the Jinghe River Basin was significantly correlated with human activity, and the value of the AOD decreased with the radius of human activity. The AOD in areas with dense populations and frequent human activities, such as cities and towns, was significantly higher than that in areas covered by vegetation, such as mountainous and forested areas. For example, the average monthly AOD of the Guanzhong Plain in the southern part of the Jinghe River Basin was as high as 0.5, whereas the AOD550 nm of Helan Mountain in the northern part of the river basin, the Ziwuling Forest area in the eastern part of the river basin, and Liupan Mountain in the western part of the river basin were significantly lower than those of the surrounding areas at below 0.26. The AOD of 60.8% in the middle of the basin ranged from 0.29 to 0.38, and the overall average AOD of the basin in 2020 was 0.297. The trend of AOD for 2020 was Increasing, then decreasing, and increasing again. The AOD value was highest in March and April and then decreased rapidly, with the lowest value appearing in August. The AOD in the Jinghe River Basin was significantly higher in winter than in summer. In winter, with lower temperatures, lower wind speeds, less rainfall, and a lower boundary layer, it is easier for the atmosphere to form a stable structure and pollutants are not easily diffused [41].

4.2. Examination of the Results of the WRF−Solar Model

By comparing the simulated values with the observed values at the site, the performance of the WRF model in the simulation of temperature, rainfall, solar radiation, and relative humidity was tested. Figure 3 shows the monthly variation in the simulated results and observed values of the WRF−solar group.

Table 2. Performance statistics of R, RMSE, MAE, and IA for WRF−solar and WRF−solar−AOD.

	R	RMSE	MAE	IA
Tem	0.9986	2.52 °C	2.41 °C	0.9807
Pre	0.9871	14.75 mm	8.50 mm	0.9859
Rad	0.8514/0.9962	113.78/130.19 MJ/m2	103.40/127.10 MJ/m2	0.7870/0.8581
Rhu	0.9093	15.31	14.43	0.6786
Tem_AOD	0.9987	2.30 °C	2.13 °C	0.9842
Pre_AOD	0.9867	10.86 mm	8.33 mm	0.9917
Rad_AOD	0.8387/0.9966	120.92/96.94 MJ/m2	105.58/93.95 MJ/m2	0.7632/0.9126
Rhu_AOD	0.8896	14.73	13.64	0.6972

shows the evaluation indexes of the two groups of experimental and observed values. The WRF−solar model simulated the temperature change well, and the correlation between the simulated and observed data was very high, reaching more than 0.9. However, overestimation occurred throughout the year and was more pronounced during the warmer months of May to September. The model simulated rainfall well and accurately simulated the time trend of rainfall, with a correlation coefficient of 0.98. However, in the peak period of rainfall in August, there was an obvious overestimation phenomenon. Since there were no observation data of solar radiation, solar radiation calculated by day length and GLASS downward shortwave radiation are chosen to verify solar radiation. Compared with the site solar radiation, the trend of the solar radiation curve simulated by WRF−solar is closer to the GLASS solar radiation data, with a correlation coefficient of 0.99. For solar radiation, the model showed a trend similar to that of temperature; however, the peak solar radiation appeared in May, earlier than the temperature peak in July. Moreover, compared to the solar radiation calculated using formulas, the models showed more obvious seasonal differences. In months with lower temperatures, solar radiation was lower, and in months with higher temperatures, solar radiation was higher. The relative humidity simulated by the model had a consistent trend with the observed data, but was slightly lower than the observed value in general, with a correlation coefficient of 0.91. In conclusion, the model accurately described the annual variation trends of meteorological elements in the basin and reflected the energy budget of the study area.

4.3. Comparison of WRF−Solar Model Results and WRF−Solar−AOD Results

To show the impact of aerosols on meteorological elements more intuitively, the difference between the WRF−solar simulation results and WRF−solar−AOD simulation results was determined, as shown in Figure 4. From Table 2, it is clear that the simulation of all other indicators becomes better under the influence of AOD, except for the correlation coefficient of rainfall and relative humidity. Because meteorological factors are mainly controlled by the weather and climate background, the feedback strength is closely related to the weather and climate background and the impact of aerosols on the environment has obvious seasonality [42]. AOD had little effect on air temperature during the high−temperature period from April to September, while it reduced the temperature by more than 0.2℃ during the low−temperature periods from January to February and November to December. The effects of aerosols on rainfall are complex. In this study, the influence of aerosols on rainfall was observed during the rainfall concentration period from June to September. Figure 4e shows the annual cloud–water mixing ratio, which was high from June to September. In August, the peak rainfall month of the year, aerosols decreased the total rainfall by more than 20 mm. The AOD had a significant impact on solar radiation, with a total reduction of 224.98 MJ/m² in 2020, accounting for 3.44% of the total annual radiation. The relative humidity in the study area increased after the addition of AOD. Relative humidity and temperature were more affected in winter than in summer.

According to the spatial distribution shown in Figure 5, the solar radiation received by the Jinghe River Basin decreases from northeast to southwest, and the site receiving the least solar radiation is the Liupan Mountain area, which is mainly affected by the terrain. Aerosols reduce the amount of solar radiation received across the basins. The south was more affected, with a greater reduction in total solar radiation. Aerosols significantly affected the changes in direct and diffuse radiation, and solar diffuse radiation increased significantly in the middle of the basin and the Guanzhong region. Changes in direct and diffuse solar radiation complement each other in space. The southern region decreased the direct solar radiation but gained more diffuse solar radiation, whereas the northern region showed the opposite trend.

The mean annual temperature in the Jinghe River Basin ranged from 4 to 18 °C, and its spatial distribution showed a flame shape. However, the temperature change was opposite to that of the flame. The temperature was the highest in the Guanzhong Plain, the center of the firework, and gradually decreased from the inside to the outside, as shown in Figure 6. The effect of aerosols on temperature showed an evident spatial trend, and the temperature decrease was more obvious where the optical thickness of the aerosol was large. Due to the presence of aerosols, the southern Jinghe Basin receives less direct solar radiation, resulting in a greater temperature drop in the southern part of the basin than in the northern part. The annual rainfall in the Jinghe River Basin decreased gradually from southwest to northeast, and the maximum annual rainfall occurred in the Liupan Mountain area in the west. The influence of aerosols on rainfall was complex, and variations in precipitation in the region were clustered and distributed across patches. Relative humidity and total annual rainfall had similar spatial distributions. The influence of aerosols on relative humidity was opposite to that of temperature, showing an increase in the south.

4.4. Effects of Aerosols on Vegetation Productivity

To keep consistency with the resolution of vegetation type, the monthly mean temperature, monthly total rainfall, monthly total solar radiation obtained from the WRF−solar simulation results, and monthly NDVI synthesized from MODIS were resampled to 1 km and input into the CASA model to obtain the annual NPP. A comparison of these three results is presented in

Table 3. Comparison of NPP results from different sources (unit: gC/m²).

Source	Min	Max	Mean
MODIS_NPP	97.40	1035.70	465.30
WRF−solar_NPP	85.26	1244.75	563.17
WRF−solar−AOD_NPP	77.93	1184.45	536.53

and Figure 7. The estimation results of the WRF−solar and WRF−solar−AOD models were in good spatial agreement with those of the MODIS NPP model, showing a banded increasing trend from north to south. The spatial correlation coefficient between the WRF−solar model and the MODIS NPP was 0.78, whereas the correlation between the WRF−Solar−AOD model and the MODIS NPP was more than 0.80, indicating that the estimation results of the WRF−Solar−AOD model had high accuracy.

In the maximum and minimum estimation, the NPP estimated by the CASA model was less different from the value of the MODIS NPP. However, CASA produced significantly higher NPP estimates than MODIS in the average estimation due to the overestimated solar radiation and temperature. High NPP values were distributed in densely vegetated areas, such as the Liupan Mountains and the Ziwuling forest area. The NPP calculated using the CASA model has a clear direction for the river system in the basin, as shown in Figure 7b,c. In Guanzhong City in the southern part of the basin, the NPP calculated using the CASA model showed a low value, which was more consistent with the actual situation than the MODIS NPP result.

Aerosols enhance vegetation photosynthesis by increasing diffuse radiation. However, a decrease in direct radiation can inhibit or even outweigh the effect of diffuse radiation fertilization. The presence of aerosols resulted in a decrease of 26.64 gC/m² in the total mean NPP in the Jinghe River Basin, which is not conducive to vegetation photosynthesis and carbon fixation in the basin. Spatially, NPP increased in some local areas, but decreased overall. The decrease degree of NPP was the highest in the Ziwuling area in the eastern part of the basin, whereas the decrease degree of NPP was low in the Helan Mountain area in the northern part of the basin.

Variations in climate conditions, aerosol characteristics, and vegetation cover contribute to the heterogeneous effects of aerosols on radiation and plant growth. The characteristics of the vegetation itself, such as plant function type, leaf area, and canopy structure, will also affect vegetation productivity [43]. To compare the impacts of aerosols on vegetation productivity more intuitively, the rate of change in vegetation productivity was calculated. The method was to obtain NPP by subtracting the NPP in the no−aerosol scenario from the NPP in the aerosol scenario and then dividing by the NPP in the no−aerosol scenario. The change rate of vegetation productivity in the Jinghe River Basin ranged from −0.13% to 0.03%, and the change rate of vegetation productivity was positive, indicating that vegetation productivity increases under the influence of aerosols. The AOD and vegetation productivity change rates were superimposed under different land use types, and the results are shown in Figure 8 and Table 3.

As shown in

Table 4. Overlay analysis results of AOD and vegetation productivity change rate of cropland, forestland, and grassland in Jinghe River Basin.

Crop Value	Crop Count	Crop Ratio	Forest Value	Forest Count	Forest Ratio	Glass Value	Glass Count	Glass Ratio
111	7	0.44%	112	87	3.67%	113	8	0.37%
121	153	9.65%	122	442	18.65%	123	163	7.61%
131	962	60.69%	132	1164	49.11%	133	1231	57.44%
141	457	28.83%	142	594	25.06%	143	731	34.11%
151	6	0.38%	152	83	3.50%	153	10	0.47%
211	70	2.28%	212	23	0.88%	213	75	2.29%
221	590	19.20%	222	581	22.16%	223	854	26.05%
231	1677	54.57%	232	1582	60.34%	233	1480	45.15%
241	730	23.76%	242	415	15.83%	243	866	26.42%
251	6	0.20%	252	21	0.80%	253	3	0.09%
311	512	6.80%	312	54	2.56%	313	352	6.35%
321	2660	35.33%	322	660	31.26%	323	2067	37.26%
331	3097	41.13%	332	936	44.34%	333	2286	41.21%
341	1249	16.59%	342	452	21.41%	343	834	15.04%
351	12	0.16%	352	9	0.43%	353	8	0.14%
411	1151	17.29%	412	163	15.44%	413	493	15.73%
421	2541	38.17%	422	349	33.05%	423	1171	37.36%
431	2041	30.66%	432	309	29.26%	433	1039	33.15%
441	882	13.25%	442	206	19.51%	443	414	13.21%
451	42	0.63%	452	29	2.75%	453	17	0.54%
511	523	\	512	5	\	513	47	\
521	499	\	522	0	\	523	1	\
531	9	\	532	0	\	533	0	\
541	0	\	542	0	\	543	0	\
551	0	\	552	0	\	553	0	\

, with an increase in the AOD, the proportion of productivity reduction levels 4 and 5 in cropland and forestland decreased continuously, while the change of forestland decreased first and then increased. This indicates that high AOD is conducive to forest photosynthesis to a certain extent. According to Strada et al. [44], environmental conditions with high AOD can improve the productivity of the deciduous forest, which is in agreement with the research results of this paper. When the AOD level was 2, that is, when the AOD was 0.26–0.29, it is most unfavorable to forest growth. As the numbers of forestland and grassland areas with AOD grade 5 were too small to be statistically reliable, they were ignored.

5. Discussion

5.1. Aerosol and Radiation

Direct and diffuse solar radiation are important solar radiation parameters; however, few models output these two variables. Aerosols significantly affect the amount and composition of solar radiation. Compared with the GHI, aerosols have a greater influence on the DNI and DIF [45], which was in line with the simulation results in this study. The WRF−solar model uses the radiative transfer model RRTMG [46] to consider aerosol radiation feedback and to output direct solar radiation and diffuse radiation. The total solar radiation decreased by 224.98 MJ/m², accounting for 3.44% of the total annual radiation. Aerosols reduce direct radiation and enhance diffused radiation across basins. Under the influence of aerosol, the direct radiation decreased by 1486.90 MJ/m², accounting for 19.64% of the total annual direct radiation, and the diffuse radiation increased by 496.91 MJ/m², accounting for 23.14% of the total annual diffuse radiation. The formula for calculating solar radiation based on day length is not localized, which may account for the large difference between the solar radiation of the site and the WRF−solar results. In this experiment, the RRTMG scheme and rural−type aerosols were selected, and WRF−solar determined the hygroscopic properties of the aerosol based on the input AOD data, selected aerosol type, and relative humidity of the environment. Therefore, numerous variables acting in concert caused radiation changes in space.

5.2. Aerosol and Temperature

By reflecting and diffusing solar radiation, aerosol reduces the solar radiation reaching the Earth’s surface, resulting in a cooling effect under the action of atmospheric circulation [47]. Therefore, the simulated temperature with aerosols was lower than that without aerosols. In this paper, the influence of aerosols on the temperature was seasonal, and the temperature difference curve was inverted U−shaped, as shown in Figure 4a. However, it should be noted that the monthly AOD curve had the same trend as the temperature difference curve, which was higher in winter and lower in summer. Therefore, the AOD concentration in the Jinghe River Basin should also be considered as a factor for temperature variation. Radiation changes coincided with temperature changes in space.

5.3. Aerosol–Cloud and Rainfall

Like aerosols, clouds can cause changes in the amount and composition of solar radiation reaching the ground and affect temperature and humidity conditions. Therefore, the shielding effects of clouds cannot be ignored [48]. The aerosol–cloud interaction changes the cloud albedo and rainfall efficiency by affecting the size and concentration of cloud water. As cloud condensation nuclei, soluble aerosols improve the cloud albedo and reduce rainfall efficiency. When the total cloud liquid water simulated by WRF−solar is 0, the situation is determined to be cloudless [49]. Cloud liquid water content is the product of the cloud water mixing ratio and atmospheric density. As shown in Figure 4, from June to September, the cloud–water mixing ratio was high, indicating a cloudy situation. Clouds interacted with aerosols, which had a significant influence on rainfall during this period. However, some studies have proven that the influence of the change in cloud condensation nucleus concentration on the physical process of cloud precipitation is non−monotonic [10] and highly dependent on environmental conditions. The influence of aerosols on temperature and radiation changed gradually with latitude; however, the spatial distribution of rainfall changes showed a patchy pattern, with a certain degree of aggregation. This suggests that other environmental factors have synergistic effects on rainfall and aerosols [50]. Therefore, clouds and other environmental conditions should be considered when assessing rainfall.

5.4. Aerosol and NPP

Different levels of AOD have different effects on vegetation. In other words, there is a threshold at which the AOD has opposing effects on vegetation photosynthesis [51]. It should be noted that this study did not divide the threshold value further, which should be addressed in future research. Compared with the absence of aerosols, diffuse radiation caused by aerosols can promote an increase in NPP in local areas. However, continuously high AOD conditions obstruct direct radiation, reduce the availability of total solar radiation, and are not beneficial for vegetation photosynthesis. Based on the above analysis, aerosols in the atmosphere can increase diffuse radiation; however, whether they can eventually lead to the improvement of vegetation productivity is restricted by a variety of factors. For example, the environmental conditions of vegetation (temperature, soil moisture, evapotranspiration, etc.) affect the regulatory ability of vegetation from time to time and have an inseparable impact on vegetation productivity. In addition to aerosols, other atmospheric components can also affect the carbon sequestration process in terrestrial ecosystems. Surface ozone damages leaf photosynthesis by oxidizing plant cells [52]. At the same time, the collaborative control of aerosols and ozone has become a major challenge for cities. In future studies, ozone should also be taken as an influencing factor to explore the changes in vegetation productivity under the combined action of aerosol and ozone.

5.5. Unit, Scale, and Uncertainty

As a unified and complete ecosystem, a river basin is not only a carrier of human activity but also a gathering place for vegetation growth. In this study, AOD and NPP could not be separated from human activities and vegetation growth; therefore, the basin was selected as the study unit. The Jinghe River Basin has the characteristics of high vegetation cover and rapid urban development, which made it suitable for the study area. In addition, the WRF−solar model adopts the dynamic downscaling mode, which requires a large amount of operational computing power. Considering factors such as operation time and operational computing power, the scale of the basin matched that of the climate model and was thus suitable for this study.

In this paper, multiple resolution data were used to calculate NPP. In order to unify the resolution, a resampling method was used, which increases the uncertainty of the results. AOD data and FNL data contain a great deal of uncertainty. Different data sources may lead to changes in the results. At the same time, the CASA model relies on empirical settings for key parameters, which also brings uncertainty to the simulation of terrestrial ecosystem productivity.

6. Conclusions

Based on the WRF−solar model and FNL data, this study simulated climate change in the Jinghe River Basin in 2020. Using the AOD as the external input variable in the experiment, the impact of aerosols on climatic elements such as solar radiation, temperature, rainfall, and relative humidity in the river basin was considered using a control variable method. The net primary production of vegetation in the Jinghe River Basin in 2020 was determined using these two results as input variables for the CASA model, with and without taking the influence of aerosols into account. The results showed that WRF−solar can accurately describe the annual variation trend of meteorological elements in the basin. The correlation coefficients between the simulated temperature, rainfall, solar radiation, and relative humidity and the observed values can exceed 0.85, which properly reflects the energy budget in the study area. Aerosols play a cooling role in the Jinghe River Basin, resulting in a decrease in temperature in the basin. Additionally, aerosols as cloud condensation nuclei affected the size and concentration of cloud droplets and then affected rainfall. Different AOD levels have different effects on different vegetation types. Aerosols result in lower productivity in a large area of vegetation in the basin, which is not conducive to basin vegetation photosynthesis and carbon fixation. Whether aerosols have a positive effect on vegetation productivity is limited by many factors, such as AOD concentration, meteorological environment, cloud cover, and vegetation characteristics.