Biomass burning induced surface darkening and its impact on regional meteorology in eastern China

. Biomass burning has attracted great concerns for the emission of particular matters and trace gases, which substantially impact air quality, human health and climate change. Meanwhile, large areas of dark char, carbon residue produced in incomplete combustion, can stick to the surface over fire prone areas after open burning, leading to a sharp drop in surface albedo, so-called ‘surface darkening’. However, exploration into such decrease in albedo and its radiative and 10 meteorological effects is still fairly limited. As one of the most high-yield agricultural areas, eastern China features straw burning every early summer, the harvest season for winter wheat, which was particularly strong in 2012. Satellite retrievals shows that surface albedo was significantly decreased (up to -0.16 decrease in the near-infrared broadband). Observational evidences of abnormal surface warming were found by comparing radiosonde and reanalysis data. Most sites around intensive burned scars, especially in the downwind direction shows a positive deviation. Comparisons between ‘pre-fire’ and 15 ‘post-fire’ from 2007 to 2015 indicated a larger temperature bias of the forecast during ‘post-fire’ stage. The signal is more obvious in the later afternoon. WRF-Chem simulations suggest that including ‘surface darkening’ can decrease model bias and well captured temperature variation after burning at sites in fire areas and its adjacent area. This work highlights the importance of biomass burning induced albedo change in weather forecast and regional climate.

variations and distinct regional characteristics (Duncan B. N.;Martin, 2003;Hao and Liu, 1994;Streets et al., 2003;Kaiser et 30 al., 2012;Wiedinmyer et al., 2011;Uhl and Kauffman, 1990;Bowman and Panton, 1994), and bring about much human health loss, premature mortality and morbidity, especially in developing countries, which is an alarming situation (Pöschl, 2005;Qin et al., 2019;Wang et al., 2015;Wang et al., 2017). Though supervision has been strengthened, threats from global fire activities are still increasing dramatically with climatic and anthropogenic drivers (Marlon et al., 2009;Van Der Werf et al., 2006;Pechony and Shindell, 2010;Marshall et al., 1996;Penner et al., 1992). 35 Since open BB tends to occur directly on land surface, land cover and underlying surface properties would be affected straightway (Myhre et al., 2005;Certini, 2005). Surface albedo was revealed to be highly sensitive to BB, a sharp drop down to abnormally low values of which can be observed over fire prone areas due to the surface darkening caused by black charcoal produced in incomplete combustion, and will maintain for a period of time before char materials are removed by weathering and vegetation starts to regenerate (Jin and Roy, 2005;Govaerts et al., 2002;Myhre et al., 2005;Veraverbeke et al., 40 2012;Beringer et al., 2003;Bremer, 1999;Amiro et al., 2006;Govaerts et al., 2003;Tsuyuzaki et al., 2008;Lyons et al., 2008).
The magnitude of such a decrease in surface albedo depends on fire intensity and can be up to half of pre-fire values (Veraverbeke et al., 2012). Reduction in surface albedo enhances the capacity of fire-affected surface to absorb solar radiation under clear sky condition, perturbing the surface energy budget and the process of land-atmosphere interaction, and then modifying local or even regional circulations (Govaerts et al., 2003;Gatebe et al., 2014). These changes in 45 meteorological parameters will be important for weather and air quality forecast Ding et al., 2013a;Yang et al., 2019).
With the highest crop production in the world, about 122 Tg of crop residue are burned on spot annually in China in estimation (Streets et al., 2003). Eastern China, especially the middle and lower reaches of Yellow River and Yangtze River, features large plain area and dense population with high agricultural production. Owing to a distinctive rotation cycle of 50 "winter wheat, summer maize", eastern China has been characterized by intensive straw burning every summer, especially June, the harvest season for winter wheat (Yin et al., 2019;Li et al., 2016;He et al., 2007;Huang et al., 2012a;Huang et al., 2018;Yang et al., 2013). After revolution in rural fuel structure, farmers who are eager to deal with tons of wheat straw and continue the cycle always resort to burning on spot rather than taking them as fuel. Hence, the wheat straw burning is exceedingly dense in both time and space, posing alarming threats to air quality and human health (Huang et al., 2012b;Nie 55 et al., 2015;Ding et al., 2016a;Ding et al., 2016b;Ding et al., 2013a). The combination of BB emissions and fossil fuel emissions in eastern China also brings great challenges to weather and climate forecast (Xu et al., 2018;Zhou et al., 2018;Ding et al., 2013b;Wang et al., 2018).
An extremely typical pollution episode was triggered in eastern China in June 2012, when the most severe straw burning happened based on statistics (Huang et al., 2012b). A yellow haze blanketed the Yellow River Delta (YRD), resulting from 60 intensive BB emissions from northern agricultural area mixed with local fossil fuel emissions Ding et al., 2013a). During the episode, a series of meteorological anomalies were observed, such as air temperature changes and precipitation redistribution, which have been attributed to aerosols, particularly absorption aerosols like black carbon (BC)  Ding et al., 2013a;Gao et al., 2018). However, the impact of local and region energy budget, like surface darkening (Jin and Roy, 2005;Myhre et al., 2005) have not been depicted yet. The striking spectral contrast between dark 65 charcoal deposit over burned surface and bright day ripening wheat could arouse distinct physical signals (Govaerts et al., 2003;Pereira et al., 1999). What's more, the effect of fire on albedo is complex and depends on the pre-fire structure, underlying soil reflectance, combustion completeness of fire and vegetation recovery after fire (Roy and Landmann, 2005).
It also announces the urgent need of investigation in radiative and meteorological effects of fire-induced albedo change in eastern China combined with the high intensity of BB, especially in summer when solar radiation is extremely strong. 70 In this study, magnitude and distribution of surface albedo decrease induced by crop residue burning in eastern China, so called 'surface darkening' was investigated combining fire detections by MODIS based on straw burning in June 2012.
Temperature anomalies after BB in eastern China were also explored using long-term site observations from 2007 to 2015, and comparisons were made between the two status, 'pre-fire' and 'post-fire'. Moreover, with the aid of the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem model), decrease in surface albedo over fire prone 75 areas were taken into consideration in the simulation targeting at temperature anomalies in June 2012. The further impact of 'surface darkening' in radiation budget, energy reallocation in Earth-atmosphere system and meteorology were discussed on a local and regional scale, which is of assignable climatological significance and should not be neglected.

Fire counts and surface albedo 80
The intensity of fire counts and magnitude of surface albedo decrease induced by crop residue burning were analyzed based on standard Moderate Resolution Imaging Spectroradiometer (MODIS) data products provided by NASA (Vermote, 2015;Giglio and Justice, 2015a, 2015b. Surface Reflectance product MOD09A1 provides an estimate of the surface spectral reflectance of MODIS Band 1 through 7 at a horizontal resolution of 500 m and use Bidirectional Reflectance Distribution Function (BRDF) database to better 85 constrain the different threshold used (Schaaf et al., 2002). The seven-narrowband albedo can be converted to broadband like shortwave and near-infrared through valid algorithms (Liang, 2001;Liang et al., 2003). It has been corrected for atmospheric conditions such as gases, aerosols and Rayleigh scattering. The dataset is updated every 8 d, for selecting one value from all acquisitions in an 8-day composite period for each pixel unit. The criteria for grid cell value selection includes cloud conditions and solar zenith angle. 90 Thermal Anomalies/Fire Daily L3 Global product (MOD/MYD14A1) combined with burned area product (MCD64A1) were adopted to determine the locations and time of burning and mark the burned scar. MOD14A1 and MYD141 are daily surface thermal anomaly data obtained by Terra and Aqua satellites with spatial resolution of 1000m, and fire counts with low confidence were counted out for quality control (Hawbaker et al., 2008). MCD64A1 Burned Area data product is monthly upgraded and contains burned date and quality information for each pixel, using burn sensitive vegetation index (VI) derived from atmospherically corrected surface albedo information of MODIS short-wave infrared band to create dynamic thresholds for composite data (Giglio et al., 2016;Giglio et al., 2009). The "fire count" described in the following part indicates the central location of MODIS pixels with surface thermal anomalies. The actual geographic size of the pixel varies depending on the angle at which satellite scans. When counting the number of pixels in target region, the pixels are abstracted into points on a two-dimensional plane. 100

Temperature observation and forecast
Meteorological sites observations from National Centers for Environmental Information (NCEI) offer the hourly air temperature at 2 m height, while atmospheric sounding observation at Xuzhou station is provided by University of Wyoming (http://weather.uwyo.edu/upperair/sounding.html).The temperature observations were used to compare with National Centers for Environmental Prediction (NCEP) global final analysis data (FNL) with a 1°× 1° spatial resolution which is 105 updated every 6 h, and were also applied for verification and analysis of numerical simulations.

Experiment design and model configuration
WRF-Chem version 3.6.1, an online-coupled chemical transport model, was adopted in this study, in which multiple physical and chemical processes are depicted including emission, transport, mixing, and chemical transformation of trace gases and aerosols simultaneously with meteorology (Grell et al., 2005). 110 In order to take fire-induced albedo change into consideration and figure impact of 'surface darkening' in radiation budget and meteorology in eastern China, a pair of parallel experiments in WRF-Chem with surface albedo as variable was designed: In CTL run, default value of monthly surface albedo in model was adopted, while in ABD run, surface albedo was set by subtracting a certain value from the default model value over fire-prone areas. The fire-prone areas were extracted out based on density of fire counts in June 2012, and the certain value characterizing the magnitude of albedo decrease were defined 115 based on satellite data analysis in Section 3.1.
The initial and lateral boundary conditions of meteorological variables are obtained from NCEP FNL that update every 6 h.
MIX, a mosaic Asian anthropogenic emission inventory developed by Tsinghua University (Li et al., 2017)and the Fire Inventory from NCAR (FINN) providing high resolution global emission estimates from open burning (Wiedinmyer et al., 2011) were applied in WRF-Chem as anthropogenic and BB emissions, respectively. The simulations were conducted for the 120 time period 2-18 June 2012, and was split into eight independent 72-h runs for the meteorological fields, while the chemical outputs from the preceding run were used as the initial conditions for the following run. Only the last 48 h results of each run were kept for analysis, and the first 7 d were regarded as the model spin-up period for atmospheric chemistry, so as to allow the model to reach a state of statistical equilibrium under the applied forcing (Lo et al., 2008;Berge et al., 2001). scheme was adopted in CTL and ABD to keep consistency. The specific settings of domain and parameterization schemes in both CTL and ABD are listed in Table 1.

Fire-induced surface albedo change 130
In 2012, the total mass of crop residue burned was 160 Tg, and 38,053 fires were recorded on farmland accounting for 36% of the total and most occurred in June, the harvest season for wheat . Yellow-Huai River areas (YHR), located in the vast flat plain in eastern China, is the major wheat producing area in China. Figure 1a shows the daily agricultural MODIS fire counts in YHR, where cropland especially wheat fields dominates vegetation cover (Gong et al., 2019;Van Der Werf et al., 2010;Li et al., 2016). The limited duration of agricultural fire combined with the low temporal 135 resolution of MODIS may increase the difficulty of signal capture and lead to underestimation of fire counts (Randerson et al., 2012;Li et al., 2016;Giglio et al., 2009;Song, 2010, 2009). As illustrated in Fig. 1a, severe and intensive BB events happened in mid-early June, especially 7-13 June, a tense and busy period in the crop cycle for harvest of wheat and sowing of maize. Among them, burning on 13 June is the most serious. Black charcoal deposit produced in high-intensive open straw burning would adhere to exposed soil and bound surface albedo over a period of time before removal of char by 140 natural process or regeneration of vegetation (Veraverbeke et al., 2012). The strong spectral contrast between dark burned surface and original bright dry ripening wheat cover can arouse distinct physical signals (Govaerts et al., 2003;Pereira et al., 1999).
Understanding the saliency and persistence of signals is important for estimation of albedo change amplitude and further analysis of its radiative and meteorological effects. The signals of 'surface darkening' was a relatively short duration for 145 about one week and much more sensitive in shortwave near-infrared bands (Trigg and Flasse, 2000). Considering higher resolution and better depiction of narrow-band satellite observations, and the outstanding capacity of near-infrared to separate the signals between vegetation and charcoal (Jin and Roy, 2005;Trigg and Flasse, 2000), MODIS Narrow-band2 (shortwave near-infrared, from 0.841 µm to 0.876 µm) was chosen. Figure 1b and 1c show the surface albedo distribution before and after the severe burning season separately, based on MOD09A1 dataset which updates every 8 d and has ruled out 150 the interference of clouds. Distribution on 1 June has not been adopted for high missing data rate. For MODIS narrow band2, decrease induced by fire is about 0.30 to 0.14 in zone YHR, almost half of original values. The immediate surface albedo decrease after fire can be attributed to the large-scale replacement of ripening wheat with black charcoal over fire prone areas, so-called 'burned scar'. Char materials strongly absorb the incoming sunlight and as such and cause a significant reduction of the reflection-to-incoming sunlight ratio (Veraverbeke et al., 2012), so burning like this could disturb surface radiation 155 balance at a level that cannot be ignored, which needs further exploration.
The distribution accumulated high-confidence fire counts during the highly intensive straw-burning in June 2012 is shown in The peaks are relatively flat for both near-infrared and shortwave, and peak values for shortwave distribute between -0.02 and -0.06, for near-infrared distribute between -0.06 and -0.10. Since nearly half of the solar energy that reaches the surface is at wavelengths longer than 0.7 µm, albedo change in near-infrared is rather significant for the energy budget of the surface (Hartmann, 1994). According to the above analysis, in Section 3.3, the certain value of surface albedo decrease adopted in ABD run was -0.1 to examine the disturbance on surface energy budget and its potential regional meteorology impacts. 175

Observational evidences on surface warming in 'post-fire' period
Surface albedo is defined as the fraction of the downward solar flux density that is reflected by the surface, directly determining the absorptivity of the surface (Hartmann, 1994). Under the condition with same downward solar flux at the surface, decrease in albedo more solar radiation will be absorbed. Without any plants doing photosynthesis over fire prone areas, the immediate fire-induced increase in absorption of radiative energy are used to heat the ground mostly and then 180 warm the near surface atmosphere (Andrews, 2010;Wallace and Hobbs, 2006), which could lead to a phenomenon of surface warming. What's more, solar radiation is the strongest in summer when straw burning bursts in YHR, and the response of surface albedo can be the fiercest. Hence, the air temperature observations over fire prone areas after BB were investigated.
Near-surface air temperature observations at meteorological sites were compared with FNL data in 2012. Refer to Fig. 1a In fact, intensive straw burning occurred in the zone YHR every June during the harvest of wheat over the past twenty years before strict regulation launched in recent years (Yin et al., 2019;Li et al., 2016), which could announce significance in climatology to understand the signals of surface warming after BB.
Xuzhou station (marked in Fig. 4), the only sounding site located in the most intensive fire prone area in zone YHR. 195 Sounding observations at Xuzhou from 2007 to 2015 was compared with FNL data in the planet boundary layer. Sample of days was selected out according to burning condition in each year, and has been classified into specific status, 'pre-fire' and 'post-fire'. 'Post-fire' days must be in 5 days after a severe burning, while 'pre-fire' must be among the 5 days before fire but not in 5 days after previous fire. Only days under clear sky condition were selected. The statistical distribution of temperature bias on different isobaric levels at 20:00 LT in the two status was shown in Fig. 5a and 5b. Temperature bias 200 was defined as the value of sounding observation minus FNL data. As illustrated, temperature biases are mostly positive on the lower levels in both two status, and absolute values of 'post-fire' are larger than 'pre-fire'. Combined with Fig. 5c, it means observations at 'post-fire' have bigger positive deviation from forecast results, a signal of abnormal warming at 'postfire'. In addition, with regards to the vertical profile of bias at 'post-fire' (Fig. 5b), deviations on lower levels tend to be bigger and more positive than those on higher levels. Sounding data updates 2 times a day, 08:00 LT and 20:00 LT, and only 205 patterns at 20:00 LT was shown here since no significance difference exists between the two status at 08:00 LT. Under clear sky condition, the thermal states near the ground level have passed through radiation absorption and accumulation during daytime by 20:00 LT. The abnormal warming signal that tends to be more obvious near the surface, can be explained as the enhancement of surface heat absorptivity by decrease in surface albedo.
For the same sample of days in 2007-2015, the diurnal variation of 2-m temperature bias between site observations and FNL, 210 in the two status was shown in Fig. 6 by box plot. Negative values of deviation at 08:00 LT and 14:00 LT show that FNL tends to overestimate temperature during daytime and underestimate at night. It is noteworthy that the absolute deviation of 'post-fire' is always larger than 'pre-fire' no matter during the day or night, which means decrease in predictability of temperature at 'post-fire'. With regard to the median values, bias goes through the process of a gradually surpass over 'prefire', consistent with the warming signal revealed by sounding analysis at Xuzhou at 20:00 LT. The continuous rising 215 tendency during the daytime demonstrates a needed response time of heat accumulation, which may be attributed to the radiative effect of 'surface darkening' (Myhre et al., 2005;Andrews, 2010). So, the signal of abnormal warming is more obvious in the later afternoon and early evening.

Improved temperature simulation by considering surface albedo change
As aforementioned, fire-induced surface albedo change and observational evidences of abnormal surface warming were 220 substantial in eastern China. However, the possible radiative effects on local or regional meteorology and the behind physical images has not been figured out yet, which should resort to model simulation. Moreover, YHR is just adjacent to the Yangtze River Delta, an economic and cultural centre in Asia with worldwide concern, which emphasizing the knowledge of the regional impacts of albedo change. In existing studies, radiative effects of albedo changes over burned scars were investigated on different land covers, like boreal forest (Lyons et al., 2008) and savanna (Myhre et al., 2005;Jin 225 and Roy, 2005), but few of them was conducted on agricultural fire in production areas.
Hence, WRF-Chem experiments targeting at surface albedo change induced by agricultural BB in eastern China were conducted. The most severe episode happened in June 2012 was selected as the case. Distributions of the two parallel experiments, CTL and ABD, were shown in Fig. 7a and 7b. The surface albedo decrease over fire prone areas was set to 0.1.
Model results were compared with site observational air temperature, respectively in concentrated fire prone areas (Fig. 8a) 230 and its southern adjacent area, YRD (Fig. 8b). In Fig. 8, results of ABD show temperature increase and enjoy a better fit with observations at both sites in afternoon and early evening compared to CTL. Temperature anomalies can be simulated to some extent, after considering the decrease in surface albedo caused by BB.
The local surface energy budget at Bengbu was shown in Fig. 9. Bengbu locates in fire affected area where adopted the decreased albedo. As shown in Fig. 9a, upward shortwave radiation at the surface reduces a lot, while only a little change 235 exists in downward shortwave radiation reaching surface. The smaller ratio of solar radiation reflected to the atmosphere by surface can be attributed surface albedo decrease. Accordingly, with more solar radiation absorbed by surface, the ground will be heated and the surface temperature will be higher in ABD (Fig. 9b). Higher surface temperature would give rise to larger upward heat flux, thereby heating the atmosphere near the surface (Stull, 1988). So, local radiation budget change induced by surface albedo change would directly influence the near-surface air temperature and arouse an abnormal warming 240 signal over fire prone areas in the afternoon and early evening with a certain lag.
Meanwhile, the warming effect of fire-induced albedo change is also obvious in adjacent areas like Lukou (Fig. 8b), located in south of Nanjing. Some other stations like Wuhu and Liyang in YRD also show similar patterns. Considering that local temperature change is affected by various factors such as local radiation budget change and advection transport caused by thermal disturbance. As illustrated in Fig. 10a, surface warming shows in the upwind area of Lukou station, which means 245 influence of warm advection. According to Fig. 10b and 10c, sharp decrease in low cloud water contents was found in the upwind area. Low clouds are very effective reflectors of solar radiation (Hartmann, 1994), giving rise to large perturbations to surface radiation budget. In Fig. 10d, downward shortwave reaching the surface in ABD reduces quit a lot due to reflection and occlusion of low clouds. Since surface over fire prone areas are additionally heated, surface temperature increases, and both sensible heat flux and latent heat flux will be changed which could arouse variation in boundary layer 250 like vertical velocity and disturb the cloud formation (Stull, 1988). Decrease in surface albedo over fire prone area can also has great impact on adjacent areas where surface albedo was barely changed by disturbance to cloud formation and advection transmission.
'Burned scar' can remain over the surface for about one week under rainy condition or even longer if dryer and calmer (Trigg and Flasse, 2000). The length of period allows the process of local radiative accumulation and disturbance in regional 255 circulation, and influence on temperature in both fire prone area and adjacent areas.

Conclusions
To figure out the change in surface albedo induced by biomass burning and its impact on regional meteorology in eastern China, an investigation into the relationship between surface albedo change and temperature anomalies was conducted based on meteorological observations and satellite retrievals, combining with WRF-Chem simulations. This study focuses on the 260 intensive wheat straw burning occurred in the YHR area, the major wheat producing area of eastern China, in early summer.
The most severe and classic burning episode in 2012 was chosen to be the case.
Fire-induced surface albedo change was verified in YHR area based on satellite retrievals. Large area of "surface darkening" shows over fire prone area. In the near-infrared broadband, the absolute albedo decreased ranges in -0.16 ~ -0.02, which can be an obvious signal in the regional scale and arouse large radiative disturbance. Peak values of reduction for shortwave 265 distributed between -0.06 and -0.02, for near-infrared distribute between -0.10 and -0.06. especially on lower levels, which is a signal of abnormal surface warming at 'post-fire' and decrease in predictability. The signal is more obvious in the later afternoon and early evening.
To examine the direct radiative effects and its potential regional meteorological impacts of surface albedo change, WRF-Chem experiments targeting at the surface albedo change induced by agricultural BB in eastern China were carried out.
Results of experiment ABD show temperature increase and enjoy a better fit with observations at both sites in concentrated 275 fire prone areas and its southern adjacent area compared to CTL, especially in afternoon and early evening. Surface albedo change over fire prone areas influence the surface temperature through direct local radiation budget change and indirect disturbance in cloud formation and advection transmission.
This study shows that either human or natural induced biomass burning will not only influence the weather and regional climate by emission of aerosols and trace gases. The change in surface albedo could also cause significant impacts at 280 regional scale during the post-fire periods, particular in regions with strong solar radiation. Such kind of short-term disturbance should be considered in future weather, climate and air quality forecast models.   In Fig. 4a and 4b, grey arrows mark the FNL wind arrow at 10 m height, and grey dots mark the 'burned scar' (defined as accumulated fire counts in the past 5 days) while orange dots mark fire counts on the day.