LUO Chong ,LlU Huan-jun ,FU Qiang,GUAN Hai-xiangYE QiangZHANG Xin-leKONG Fanchang

1 School of Economics and Management,Northeast Agricultural University,Harbin 150030,P.R.China

2 School of Pubilc Adminstration and Law,Northeast Agricultural University,Harbin 150030,P.R.China

3 Northeast Institute of Geography and Agroecology,Chinese Academy of Sciences,Changchun 130102,P.R.China

Abstract Rice growth requires a large amount of water,and planting rice will increase the contradiction between supply and demand of water resources.Paddy field fallowing is important for the sustainable development of an agricultural region,but it remains a great challenge to accurately and quickly monitor the extent and area of fallowed paddy fields.Paddy fields have unique physical features associated with paddy rice during the flooding and transplanting phases.By comparing the differences in phenology before and after paddy field fallowing,we proposed a phenology-based fallowed paddy field mapping algorithm.We used the Google Earth Engine (GEE) cloud computing platform and Landsat 8 images to extract the fallowed paddy field area on Sanjiang Plain of China in 2018.The results indicated that the Landsat8,GEE,and phenology-based fallowed paddy field mapping algorithm can effectively support the mapping of fallowed paddy fields on Sanjiang Plain of China.Based on remote sensing monitoring,the total fallowed paddy field area of Sanjiang Plain is 91 543 ha.The resultant fallowed paddy field map is of high accuracy,with a producer (user) accuracy of 83% (81%),based on validation using ground-truth samples.The Landsat-based map also exhibits high consistency with the agricultural statistical data.We estimated that paddy field fallowing reduced irrigation water by 384-521 million cubic meters on Sanjiang Plain in 2018.The research results can support subsidization grants for fallowed paddy fields,the evaluation of fallowed paddy field effects and improvement in subsequent fallowed paddy field policy in the future.

Keywords:fallowed paddy fields,Landsat 8,Sanjiang Plain,Google Earth Engine,water security

1.lntroduction

Fallowing is a process by which farmland does not grow crops during the season when crops can be planted (Franzel 1999).Farmland fallowing can reduce water and nutrient consumption,allowing farmland to recuperate over a short time,promote the transformation of potential nutrients in soil by accumulating rainwater,and create a good soil environment and conditions for crop growth in the future(Curry and Stucki 1997).To control grain yield and protect the ecological environment,land fallowing policies in developed countries and regions such as the United States,Japan,and the European Union have been implemented for decades (Yamashita 2006；Britzet al.2011；Larket al.2015).China's reoccurring shortage of agricultural resources and food supply pressure resulting from the large population translate to intensive use of water resources on arable land,and the pressure on all resources and the environment is enormous (Godfray and Tara 2014).Thus,in 2016,the Chinese government announced a pilot scheme to explore the implementation of cropland rotation and a fallowing system (Wu and Xie 2017).Paddy field fallowing is the product of China's fallowing policy,which takes three years for a full cycle and is not allowed to be abandoned during the fallow period.

Sanjiang Plain in Heilongjiang Province is the main commodity grain base of China (Chenet al.2014).Because of the country's reform and opening up to international trade,paddy rice planting in Sanjiang Plain has rapidly expanded.By 2015,the paddy rice planting area in Sanjiang Plain had reached 2 604 946 ha (Yanet al.2018).During this period,a large number of wetlands on Sanjiang Plain were converted to paddy fields；Sanjiang Plain has been converted from a“great northern wilderness” to a “great northern granary”(Yanet al.2016).At the same time,this has also caused many ecological problems,such as a decrease in the groundwater level,land salinization,etc (Werneret al.2013).Therefore,in 2018,Heilongjiang Province focused on fallowing paddy fields in the third and fourth cumulative temperate irrigation areas of Sanjiang Plain.This policy was implemented to curb the decline in groundwater and improve the surplus ofjaponicarice to ensure the sustainable development of agriculture.Therefore,the development of high-resolution annual maps of fallowed paddy fields distribution is important for assessing the effects of fallowing policies and their impact on socioeconomic development(Renwicket al.2013).

Mapping of fallow land has been conducted in some locations.For example,Estelet al.(2015) mapped active and fallow land using a Random Forest classifier and MODIS-NDVI time series in Europe.Gummaet al.(2016)used MODIS time-series data to identify rice fallow cropland areas in South Asia.Wallaceet al.(2017) used MODIS data and the Neighborhood and Temporal Anomalies(FANTA) model to map the fallowed land in California,USA.Teluguntlaet al.(2017) used spectral matching techniques(SMTs) and automated cropland classification algorithms(ACCAs) to capture the extent and vigor of the Australian croplands versus fallow croplands accurately.In addition,Gummaet al.(2018) mapped fallowed land in Myanmar using the MODIS-NDVI and land surface water index.In conclusion,the previous studies on fallow land mapping have two characteristics:1) using MODIS products；2) using the time series normalized differenced vegetation index(NDVI).The resolution of MODIS images cannot identify the fallow land on Sanjiang Plain because the plots of fallow land are scattered and small.China's fallowing policy does not allow for abandoned farming；thus,the NDVI change in fallow land may not be much different from that of other cultivated land (Wu and Xie 2017).

Recently,many studies have attempted to map crop distribution using the crop's phenological characteristics(Wanget al.2018；Haoet al.2020).Donget al.(2015)demonstrated the use of temperature to define the time window of the rice transplanting phase in Northeast China；the study indicated that Landsat data availability,especially during the flooding and transplanting phase,is critical in guaranteeing a high-quality map of the paddy rice planting area (Donget al.2015).The Google Earth Engine (GEE)stores Petabyte-level processed data,enabling researchers to quickly process millions of images in parallel processing,greatly improving image analysis efficiency (Gorelicket al.2017).Some studies have used GEE to analyze spectral time-series data from individual pixels to generate annual rice,forest,and surface water distribution maps (Donchytset al.2016；Donget al.2016；Chenet al.2017).

In this study,time-series Landsat 8 and GEE were used to map fallowed paddy fields on Sanjiang Plain of China in 2018.The specific objectives of this paper are as follows:1)develop a phenology-based algorithm to generate an annual fallowed paddy field map of Sanjiang Plain；2) evaluate the applicability of Landsat 8 images to the fallowed paddy field algorithm；3) estimate the amount by which irrigation water was reduced in Sanjiang Plain in 2018 due to paddy field fallowing.This simple and rapid fallowed paddy field mapping algorithm will help monitor fallowed paddy in Heilongjiang Province,and provid more accurate up-to-date data for future research on food security,water resourcecarrying capacity and fallow effect evaluation.

2.Materials and methods

2.1.Study area

Sanjiang Plain is in the northeast of China,ranging from the southeastern end of Xiaoxing'anling in the west to Wusuli River in the east,and from the Heilongjiang River in the north to Xingkai Lake in the south (Fig.1).It covers the area of 45°01′-48°27′56′′N and 130°13′-135°05′26′′E.The total area is approximately 108 900 ha,with a total population of 8.625 million people and a population density of approximately 79 people per square kilometer.The administrative area includes 21 counties of Jiamusi,Hegang,Shuangyashan,Qitaihe and Jixi,and Yilan counties in Harbin.

Fig.1 Elevation and ground-truth sample locations of Sanjiang Plain,China.

Sanjiang Plain is China's main grain production and commodity grain base.The main food crops are corn,soybean and rice.The area percentage of paddy fields dramatically increased from 0.54% in 1954 to 23.92% in 2015 (Fig.2).The resulting decline in groundwater level is particularly obvious (Qiet al.2018)；thus,it is imperative that the appropriate size of fallowed paddy fields be implemented as quickly as possible (Yanet al.2018).

2.2.Data preparation

MODlS data and preprocessingRice plant transplanting is conducted when a stable temperature threshold is reached,such that the plant will not be damaged by low temperature(Zhanget al.2015).Some studies have proved that temperature is the main limiting factor for crop growth and planting in cold regions (Wanet al.2004；Panet al.2015).Thus,this study used MYD11A2 (Aqua satellite),for which the data are closer to the daily minimum temperature in the 8-day composite MODIS surface temperature product covering Sanjiang Plain from 2017 to 2018.The MYD11A2 V6 is an 8-day composite surface-temperature product with a spatial resolution of 1 km and a temporal resolution of 8 days (https://developers.google.com).

Fig.2 Percentages of paddy fields on Sanjiang Plain,China,from 1954 to 2015.

The digital number (DN) values from MYD11A2 were converted to land surface temperature (LST) with centigrade unit values based on the following equation:

LST (°C)=DN×0.02-273.15

The LST data with missing observations in a time series were also gap-filled using the linear interpolation approach(Panet al.2015).Based on previous studies,we determined that the likely starting date of flooding and transplanting(SOF) is when the nighttime LST remains above 5°C.We defined the length of the transplanting phase to open canopy as 60 days in this study (Zhouet al.2016).The starting date of the stable temperatures above 5°C was calculated,as shown in Fig.3.

Landsat 8 data and preprocessingTo map the fallowed paddy field area on Sanjiang Plain of China during 2018,this study used all the Landsat 8 Satellite images covering Sanjiang Plain during 2017 and 2018.All the atmosphericcorrected surface reflectance (SR) products from Landsat 8 were used,which were archived in the GEE as the image collection provided by the United States Geological Survey(USGS).The Landsat 8 SR data have a spatial resolution of 30 m and a temporal resolution of 16 m,which is very suitable for agricultural research (https://developers.google.com).Detailed statistical information regarding the Landsat 8 data used in this study is provided in Table 1.

Fig.4-A and C show the spatial distribution of the total observatios of Landsat 8 on Sanjiang Plain during 2017 and 2018,respectively.Cloud coverage has an important impact on image interpretation；thus,it is necessary to remove the effects of clouds as much as possible (Wilson and Jetz 2016).To mitigate the limitation that arises because of cloud cover,we applied a selection criterion to the cloud percentage (＜20%) when producing our cloud-free composite.Next,the pixel_qa bitmask band (a quality flag band) provided in the metadata was used to detect and mask clouds and shadows (Mahdianpariet al.2019).The remaining pixels were defined as effective observations.Fig.4-B and D show the spatial distribution of good observations of Landsat 8 on Sanjiang Plain during 2017 and 2018,respectively.Among them,pixels in good observation images account for 35% of the study area of the total observation images during 2017.Pixels in good observation images account for 38% of the study area of the total observation images during 2018.

Fig.3 Spatial distributions of the starting dates of the nighttime land surface temperature remaining above 5°C on Sanjiang Plain,China,based on MODIS LST data from 2017 and 2018.

Table 1 Properties of image collection selected for this study

The time series of Landsat 8 SR image collection was used to calculate three vegetation indices (VIs),including the NDVI,enhanced vegetation index (EVI) and land surface water index (LSWI) (Tucker 1979；Hueteet al.1997；Xiaoet al.2005).The spectral indices were calculated using the following equations:

wherepBlue,pRed,pNIRandpSWIRare the surface reflectance values of Band 2 (452-512 nm),Band 4 (636-673 nm),Band 5 (851-879 nm) and Band 6 (1 566-1 651 nm) in the Landsat 8 SR,respectively.

All MODIS and Landsat data were processed using the GEE Cloud Computing Data Platform (https://code.earthengine.google.com),which has been proven to work with large-area computing and big data processing.After eliminating poor observations,the obtained time-series data were used for the phenology-based fallowed paddy field mapping algorithm.

2.3.Algorithm for identifying fallowed paddy field area

The traditional idea of extracting fallow land is to consider the difference in the VI (such as NDVI and EVI) values before and after fallowing (Estelet al.2015；Gummaet al.2016,2018；Wallaceet al.2017).Generally,the value of the NDVI or EVI during the crop growth period after land fallowing is less than it was before fallowing (Prishchepovet al.2012).However,China's fallowing policy does not allow for abandoned land,and the fallowed paddy field area may be planted with some other crops,which means that the NDVI or EVI value during the crop growth period before and after fallow does not change much(Fig.5-B).

Farmers do not irrigate paddy fields after fallowing during the rice transplanting phase (Fig.6)；thus,we considered using this feature to identify the fallowed paddy field area.Gao (1996) developed the normalized difference water index (NDWI),also known as the LSWI or normalized difference moisture index (NDMI),which quantifies moisture levels in vegetation and shows a positive correlation with surface water and soil moisture(Zhaoet al.2009).Thus,we analyzed the changes in the LSWI before and after fallowing in paddy fields during the rice transplanting phase.Fig.5 shows that the change in the LSWI during the rice transplanting phase before and after fallowing in paddy fields is obvious.The LSWI value after fallowing is always less than that before,because after fallowing the paddy fields,farmers do not irrigate the paddy fields during the rice transplanting phase.Therefore,we can design parameters based on the LSWI during specific phenological periods.

Fig.4 Spatial distribution of total observation and good observation over the Sanjiang Plain,China,during the study periods of Landsat 8.A and C,total observations during 2017 and 2018,respectively.B and D,good observations during 2017 and 2018,respectively.

Based on the difference in the LSWI before and after fallowing,the fallowed paddy field index (PFI) was developed.The index mainly considers the slope of the LSWI change before and after fallowing in paddy fields during the rice transplanting phase.Because the LSWI change in the fallowed paddy fields was much greater than that of other land types before and after fallow,the equation for PFI is as follows:

where LSWImedian(before) is the median LSWI during the rice transplanting phase before fallowing and LSWImedian(after) is the median LSWI during the rice transplanting phase after fallowing.We used the median of the LSWI time series to reduce the impact of clouds and shadows.Because the PFI represents the possibility that the pixel is a fallowed paddy field,it is necessary to determine the optimal threshold of the fallowed paddy field area.Through experimental trials,we determined that 0.8 is the best threshold for PFI,because the highest overall accuracy is obtained when the threshold is 0.8 (Fig.7).When the PFI value is greater than 0.8,the pixel can be considered as representing a fallowed paddy field area.

A flowchart for mapping the fallowed rice in this study is shown in Fig.8.

2.4.lmplementation of the fallowed paddy field mapping algorithm

Although the LSWI change is a distinct feature of the fallowed rice land during the rice transplanting phase,it may also be affected by water bodies such as rivers and lakes,natural wetlands,natural vegetation,and construction land (Vleeshouweret al.2015).Masking these nonfarm fields can improve the accuracy of fallowed rice land extraction,a digital “cultivated land” mask at a 30-m spatial resolution (unpublished data) was superimposed onto the crop map to help select crop sampling locations,it is a 30-m-resolution map layer based on object-oriented and human interpretation.

Fig.5 Temporal profile analysis of Landsat 8 vegetation indices at two different fallowed paddy fields in Fuyuan (A；134.24°N,47.90°E) and Hulin (B；132.69°N,45.95°E) in China.The solid line boxes are the rice transplanting phase in the region during 2017,and the dotted line boxes are the rice transplanting phase in the region during 2018.LSWI,land surface water index；EVI,enhanced vegetation index；NDVI,normalized difference vegetation index.

2.5.Using ground-truth samples for accuracy assessment

To collect ground-truth samples,we conducted field investigations in the study area from July to August 2018.Because the focus of this study was fallowed paddy fields,the reference samples consisted of two categories (fallowed and non-fallowed).The observations recorded during the surveys in 2018 contained 278 non-fallowed points and 155 fallowed points.

2.6.Validation using agricultural statistical data

We collected agricultural statistical data published on local government websites for comparison and assessment of the accuracy of the fallowed paddy field map.Because no data on fallow tillage have been published in many counties,we only collected data on fallowed paddy fields in three counties (Fuyuan,Tongjiang,and Suibin,which include 22 townships).

3.Results

3.1.Fallowed paddy field map of Sanjiang Plain from Landsat 8 of 2018

The spatial distribution of Landsat-based fallowed paddy fields on Sanjiang Plain in 2018 is shown in Fig.9.Thecounties with a fallow area of more than 10 000 ha in Sanjiang Plain are Fuyuan,Hulin and Fujin.Among them,the fallowed paddy field area in Fuyuan is the largest,reaching 17 333 ha.Through satellite monitoring,the total fallowed paddy fields area of Sanjiang Plain is 91 543 ha.

Fig.6 The median composite value of Landsat 8 during the rice transplanting phase in two fallowed paddy fields.A,fallow sample in Fuyuan (134.24°N,47.90°E).B,fallow sample in Hulin (132.69°N,45.95°E).

3.2.Assessment of the accuracy of the fallowed paddy field map using ground-truth sample

The fallowed paddy field map based on Landsat 8 data was validated using ground-truth samples.Table 2 presents the error matrix of the accuracy evaluation of the fallowed paddy field map of Sanjiang Plain during 2018.The producer accuracy of the fallow land is 83%,the user accuracy is 81%,the rate of missing data is 17% and the misclassification is 19%.The overall classification accuracy is 87%,and the Kappa coefficient is 0.71.The results showed that the fallowed paddy field map based on the Landsat 8 data was of high accuracy.

Fig.7 Change in the overall accuracy with the fallowed paddy field index (PFI) threshold.

3.3.Validation results using agricultural statistical data

We compared the fallowed paddy field area in this study with the statistical data published by the Heilongjiang provincial governments at the township level (Fig.10).The results showed that theR2was 0.7808 between the Landsat-based and statistical data；the correlation was significant at a level ofP＜0.001 (n=22).The root mean square error (RMSE)between them was 228.74 ha.The Landsat-based data exhibit high consistency with the agricultural statistical data.

3.4.Assessment of irrigation water reduction in 2018

The current paddy rice irrigation quota in Sanjiang Plain of Heilongjiang Province is 4 200-5 700 cubic meters per ha(DB 23/T 727-2017 2017).We monitored the fallow area of paddy fields in Sanjiang Plain as 91 543 ha.We estimated that Sanjiang Plain reduced irrigation water by 384-521 million cubic meters in 2018 due to paddy field fallowing.

4.Discussion

4.1.Effectiveness of China's fallowing policy implementation on Sanjiang Plain

Fig.8 Workflow for the phenology-based fallowed paddy field mapping in this study.LST,land surface temperature；SOF,starting date of flooding and transplanting；EOF,ending date of flooding and transplanting；SR,surface reflectance；EVI,enhanced vegetation index；NDVI,normalized difference vegetation index；LSWI,land surface water index；PFI,fallowed paddy field index.

Fig.9 Distribution of fallowed paddy fields on Sanjiang Plain,China,in 2018.A,spatial distribution of fallowed paddy fields.B,area with counties of fallowed paddy fields.

Table 2 Error matrix of evaluation results of the fallowed paddy field results on the Sanjiang Plain,China,during 2018

Fig.10 Township-level area comparison of the fallowed paddy field area on Sanjiang Plain,China,between the Landsat-based and statistical data.

Japan,the United States,and the European Union are countries or regions where fallowing policies have been relatively successful (Yamashita 2006；Britzet al.2011；Larket al.2015),they established their own fallowing policies in 1971,1985,and 1992,respectively.These fallowing policies have enabled these countries to successfully control food production while also protecting the ecological environment.To maintain the sustainable use of cultivated land,the Chinese government has also introduced a fallowing policy.However,China is still in the primary stage of exploration compared with countries that have achieved good results in fallow (Yuet al.2019).

Taking this study as an example,the government originally planned to fallow 73 333 ha of paddy fields on Sanjiang Plain.According to remote sensing monitoring,the actual fallowed paddy field area on Sanjiang Plain is 91 543 ha,18 210 ha more than the planned area,which shows that the fallowing policy has been well implemented on Sanjiang Plain.From our monitoring results,China's fallow policy for Sanjiang Plain has been effective.

4.2.Advantage of mapping fallowed paddy fields using Landsat 8 and GEE

Fallowing is an important means to ensure the sustainable development of agriculture.The two goals of fallowing are to control food production and ease the pressure on the ecological environment on Sanjiang Plain (Wu and Xie 2017).Obtaining the spatial distribution of fallowing is the key to achieving these two goals.

In the past,the government relied on reports from farmers to obtain the location of fallowed land.This method takes considerable time and is also affected by the accuracy of the farmers' reporting (Wallaceet al.2017).The result is not sufficiently objective.In contrast,remote sensing technology can provide a more objective allocation of fallowed paddy filed (Fig.11).However,the traditional remote sensing fallowing monitoring method is limited,and the influence of computing power and image resolution is not suitable for monitoring the spatial distribution of fallowed paddy fields on Sanjiang Plain.The phenological-based paddy field fallow mapping algorithm proposed in this study can obtain the distribution of fallowed paddy fields before August.This method mainly considers the change in surface moisture before and after fallowing paddy fields.The advantage of this method is that it does not require field validation data for training,but it instead relies on temporal and spatial differences,i.e.,comparing current cells to their historical cells and their neighbors.Although this method has only been tested on Sanjiang Plain,it is expected to be easily extended to other fallowed paddy field areas.

4.3.Uncertainties in the algorithm to map fallowed paddy fields

Although the fallowed paddy field map in this study is consistent with the statistical data and the accuracy is quite high,for some reason,mapping the fallowed paddy fields on Sanjiang Plain of China remains an arduous task.

First,cloud pollution is the most important factor affecting fallowed paddy field mapping (Liaoet al.2018).Because the research period of this study is approximately two months of paddy rice transplanting every year,the image quality of rice transplanting directly affects fallowed paddy field data extraction.Secondly,a pixel-based model was used in this study.There is a considerable “salt and pepper”noise in the pixel-based classification map,because only spectral information is used (Zhanget al.2017).The large fallowed paddy field area in this study may have been caused by the “salt and pepper” phenomenon causing some non-fallowed land to be categorized as fallow land.Third,although our results have removed most other land types through the mask of cultivated land,some changes in land use types may still affect fallowed paddy field extraction.

4.4.Future developments in fallowed paddy field mapping

Sentinel 2 data have similar spatial resolution and bands to Landsat data,and can be obtained free of charge.With the Sentinel-2B satellite launch,the time resolution of the Sentinel-2 data increased to 5 days (Kutseret al.2018).The spatial resolution and spectral similarity of the Sentinel-2 and Landsat data indicate that these image data can be combined for mixed analysis (Fig.12).This can further improve the accuracy of fallowed paddy field mapping.

Fig.11 Spatial distribution of fallowed paddy fields on the two major counties of Hulin (A) and Fuyuan (B) in Heilongjiang Province,China.

Fig.12 Spatial distribution of total observations over Sanjiang Plain,China,during the transplanting phase of Sentinel-2.A and B,total observations during the transplanting phase in 2017 and 2018,respectively.

Differences in climate,temperature,and precipitation in different regions lead to phenological differences,and thus,large-scale agricultural monitoring has always been a great challenge.Mapping fallowed paddy field is necessary,but the use of common methods for monitoring large-scale fallowed rice requires enormous human and material resources.This work requires satellite imagery with good temporal and spatial resolution and consumes considerable computational power.GEE has almost unlimited processing power,which solves the computer processing power problem that puzzles researchers,and can provide enough computing power for monitoring of fallowed paddy fields.Future multisource spatiotemporal data and cloud processing platforms can play a greater role in large-scale agricultural monitoring.

4.5.lmpact of fallow on water security

Large amount of water is needed for rice planting in Heilongjiang Province.In 2017,the water consumption of paddy field irrigation in Heilongjiang Province was 30.192 billion cubic meters,accounting for 86% of total social water consumption,and agricultural water consumption was 31.644 billion cubic meters,accounting for 98% of farmland irrigation water (Heilongjiang Provincial Bureau of Statistics 2018).The irrigation water for rice in Sanjiang Plain is mainly groundwater and surface water.The cost of groundwater irrigation is less than that of surface water irrigation,which leads rice growers to prefer groundwater irrigation.

Through our research,it is found that in 2018,only the paddy fields in Sanjiang Plain of Heilongjiang Province saved more than 378 million cubic meters of irrigation water,accounting for more than 1.2% of the total agricultural water consumption in Heilongjiang Province in 2017.Therefore,the paddy field fallow in Sanjiang Plain has great significance for ensuring the safety of water resources in the region.Most of the paddy fields in Sanjiang Plain are irrigated by groundwater (Fig.13),and the problem of groundwater overexploitation will be alleviated by fallow.

5.Conclusion

Fig.13 Paddy field fallowing on Sanjiang Plain,China.A and B,paddy field fallowing in Sanjiang Plain.C,high-power pump for paddy field irrigation in Sanjiang Plain.D,paddy fields without fallow.

This study proposed a method for mapping fallowed paddy field based on phenological changes before and after fallow.The slope of the LSWI change before and after fallowing in paddy fields during the rice transplanting phase was applied as the primary metric to identify fallowed paddy field area.Its application on Sanjiang Plain of China showed its efficiency.The fallowed paddy field maps produced by this study provide an opportunity to study food security,water resource use,and other possible impacts after fallowing paddy fields.We studied the status of paddy field fallow in Sanjiang Plain.As China's fallow policy expands,more research on land fallow should be encouraged nationwide.Different regions should adopt different fallow policies according to the local ecological environment and food security level.Moreover,this method is only suitable for Northeast China,and the feasibility of this method for other regions requires further study.

Acknowledgements

This research was supported by the National Key Research and Development Program of China (2016YFD0300604-4),the Academic Backbone Project of Northeast Agricultural University,China,and the Jilin Scientific and Technological Development Program,China (20170301001NY).