Estimation of greenhouse gas emissions from rice and annual upland crops in Red River Delta of Vietnam using the denitri ﬁ cation – decomposition model

: This study used the denitri ﬁ cation – decomposition (DNDC) model and ArcGIS 10.1 to calculate and quantify the greenhouse gas (GHG) potential from rice and annual upland crops in the Red River Delta of Vietnam. GHG emissions were monitored, analyzed, and calculated at experimental sites. The operating mechanism, sensitivity analysis of the parameters, calibration, and veri ﬁ cation of the DNDC model for the GHG emission calculation were studied and performed, and a set of parameters was built. A good correlation between actual and simulated values was shown. From the data set of meteorological stations in and around the Red River Delta, the current land-use map, the topographic and soil map, a complex map of meteorology – soil – land use was built. Each unit of this map contains complete information about climate, soil, and crops as input data for modeling GHG emissions from crop production. From the spatial analysis and collected input data, GHG emissions were measured and calculated for the cultivated ﬁ eld of the Red River Delta (annual rice and upland crops) using the DNDC model. The model's outputs were used to build thematic maps on the distribution of global warming potential (CH 4 , N 2 O) for each unit of the complex map of climate, soil, and crops.


Introduction
Greenhouse gas (GHG) emissions from agriculture production have become a global problem.Significantly, it can be mentioned that Vietnam is one of the countries that depend on agricultural production as a prominent citizen's livelihood.Vietnam's agricultural production is not only heavily affected by climate change but also causes climate change because it contributes to emissions.
Vietnam's National GHG inventory was implemented four times in 1994, 2000, and 2014, and most of the calculations of GHG inventory used the emission factors of IPCC Tier 1 [1] but not the country-specific factors (Tier 2 and higher).However, precise quantification of GHG emissions from rice and other crops is complex because of spatial variations in climate and soil, crops, and farming practices.
Besides using the closed-chamber method of direct field measurement or inventory emissions according to IPCC guidelines, countries, such as China, India, etc., have used the denitrification-decomposition (DNDC) model method combined with remote sensing to calculate the GHG emissions from rice farming.The application of a mathematical model quantifying GHG emissions is a possible solution to meet both technical requirements and emission calculation with high and stable accuracy [2].
DNDC model has modules to calculate CO 2 , N 2 O, and CH 4 emissions by day and simulate the nitrogen cycle in agricultural soil.DNDC has a structure to simulate relatively fully the biochemical and physicochemical processes in soil, as well as other environmental factors (temperature, precipitation, etc.) that affect the processes of formation and release of GHGs from the soil into the atmosphere.DNDC has proven its reliability in GHG calculations through studies in many countries around the world and is considered as one of the most comprehensive currently available tools to calculate and forecast GHG emissions from agricultural and forestry activities [3][4][5].
The DNDC model has been applied to estimate N 2 O emissions and SOC fluctuations in agricultural cropping systems in the US since the early 1990s in the studies of Li et al. [5,6].Bouwman et al. [7] used the DNDC model to calculate N 2 O emissions from small livestock production systems.Cai et al. tested the DNDC model in GHG emission calculations in East Asia [8].Then, Pathak et al. and Babu et al. [9] continued to perfect the DNDC model combined with remote sensing to calculate emissions from rice cultivation in India.Most of the changes focused on the addition of anaerobic biogeochemical equations, rice growth, and farming practice parameters [9,10].
Then, a series of studies of DNDC model applications to estimate the emission reduction potential of improved farming practices were done by Cai et al. and Li et al., [11][12][13].In 2008, Fumoto and his colleagues continued to improve the DNDC model into a version of DNDC-Rice specifically for rice land.The enhancements allow DNDC to improve the accuracy of quantification and CH 4 emissions from rice fields under scenarios of climate, soil properties, farming practices, etc. and are used to determine the potential for CH 4 mitigation in Japan [14].Smith et al. [15] studied the sensitivity of the parameters of the DNDC model and used the calibrated DNDC model to estimate the GHG emission factor in Canada.Zhang et al. used the DNDC model in combination with remote sensing to calculate emissions for 1.44 million hectares of rice cultivation in southern China [16].In 2016, Zhang and colleagues applied the DNDC model to estimate N 2 O emissions under different types of irrigation in vineyards in Ningxia, China [17].Zhao et al. conducted sensitivity analysis, calibration, and verification of the DNDC model and then applied the model to suggest potential measures to reduce GHG emissions from rice fields in Shanghai, China [18].
Thus, it can be affirmed that the DNDC model has been continuously tested by CH 4 and N 2 O monitoring data measured in the field and continues to be calibrated, perfected, and upgraded by researchers around the world.The application of the DNDC model to different rice farming systems in the United States, China, Thailand, India, Japan, and the Philippines has given very positive results.These studies have demonstrated that it is possible to use the DNDC model to estimate GHG emissions from rice fields at a large scale.

Research framework
The research steps are shown in Figure 1.
Regarding the implementation process, first of all, the study has collected all data on bio-physical and socio-economic conditions of the study area, meteorology, soil, land use, crop, and farming activities in both spatial and nonspatial data of the Red River Delta.
The field measurements (to measure CH 4 and N 2 O emissions) were performed in four provinces (Nam Dinh, Thai Binh, Hai Duong provinces, and Hanoi city) on Fluvisols soil, gray soil, saline soil, and thionic soil and planted some rice varieties of Huong Viet 3, BT7, TX111, DS1, LVN17, and BC15.
The spatial data of the whole Red River Delta were processed and edited for georeferences.Then, the layers of land map information and land-use status map are superimposed and combined with the information of meteorological stations in the region to create a combination map of meteorology-land-land use representation.Therefore, each polygon contains information on meteorology, soil, and cultivation in the format of the DNDC model as inputs for further GHG emission estimation process.
In the next step, the study uses the modified DNDC model to calculate CH 4 and N 2 O emissions from rice and upland crops in the Red River Delta with different meteorological and soil conditions.The global warming potential (GWP) was estimated from CH 4 and N 2 O emissions in terms of CO 2 equivalents, guided by IPCC (2006), with a ratio of 28 for CH 4 and 265 for N 2 O.The polygons containing emissions of CH 4 , N 2 O, and CO 2 eq are then presented on maps as thematic maps (using ArcGIS 10.0) of GHG emissions (Figure 1).

Data collection
The spatial data include land-use map of the Red River Delta in 2015 received from the Vietnam General Department of Land Management and the Ministry of Natural Resources and Environment, and the soil map of the Red River Delta in 2016 received from the Soils and Fertilizers Research Institute, Vietnam Academy of Agricultural Sciences.
The soil data information includes soil types, soil layer thickness, and sand oil physical and biochemical properties received from the Soils and Fertilizers Research Institute.The crop data include crop varieties, phenology, seasonal crop calendar, farming techniques, fertilizers types, nutrient content, and application procedures.
The crop data include rice varieties and other annual crops; physiological and biochemical characteristics of rice and maize varieties; seasonal calendar; farming techniques (soil preparation, irrigation, fertilization, weeding, plant protection spraying); and types and characteristics of fertilizers (from documents, books, and scientific articles; information about varieties and results of variety testing, December 2017 statistical report of the Ministry of Agriculture and Rural Development [MARD]) [19].
Meteorological data were collected during the period 2010-2020 at 28 principal meteorological stations in the network of national monitoring of Viet Nam Meteorological and Hydrological Administration.These stations ensure measurements on a unified basis and serve for basic investigation.The information includes station coordinates, minimum and maximum temperature, sunshine hours, wind direction and speed, rainfall, and humidity in a daily resolution.The meteorological station's locations are shown in Figure 2, and their information is presented in Tables S1 and S3.

Rice varieties
The rice varieties at the experimental sites were regionalized by MARD and were Huong Viet 3, BT7, TX111, DS1, and BC15.All these five varieties have durations from 110 to 138 days and can be grown in two seasons in a year; spring rice has a longer duration than summer season rice.All five varieties have a high yield potential range from 5.5 to 7.5 tons•ha −1 season −1 (Table S2).Estimation of GHG from rice and annual upland crops using the DNDC model  3

Experimental layout
Field measurements were set up on 10 selected sites (Table 1).The farming activities were practiced following the recommendations of local extension with adjustments for local conditions of weather, soil, and farming levels.In general, Spring rice started on February 5th and ended on June 15th, and Summer rice started on 24th June and ended in October.
In spring, land preparation was done from February 5 to 18, 2018; transplantation was done from February 08 to 20, 2018, with a rice population of 30-35 hills•m −2 , and harvest from June 02 to 11, 2018.In summer, land preparation was done from June 24 to 25, 2018, transplantation was done from 30 June 30 to July 02, 2018, with a rice population of 30-35 hills•m −2 and harvest from October 17 to 27, 2018.Water management was similar to farming practices, in which the field was flooded 10 cm until the rice was matured, and then the field was drained (in spring, it was drained from May 20 to 30, 2018; in summer, it was drained from October 01 to 19, 2018 depending on each experimental site).After harvesting, the straw was removed, but the stubble was plow-incorporated into the soil.
Fertilizer was applied three times per season.Basal fertilizer uses 100% phosphorus fertilizer, 30% nitrogen fertilizer, and 30% potassium fertilizer.The first split at tillering was with 40% nitrogen fertilizer, and the second split was at panicle initiation; the second additional fertilizer was 30% nitrogen fertilizer, and the remaining 70% was potash fertilizer.The rate of fertilizer application was recommended by the provincial Departments of Agriculture and Rural Development and trained by the Agricultural Extension Centers of the provinces.This is also the fertilizer level we obtained from the survey results of 720 farmer households.

Gas sampling
The CH 4 and N 2 O fluxes were determined using the techniques of static flux chamber and gas chromatography, following the methods of Rochette and Eriksen-Hamel (2008) [20].The chamber consists of a permanently installed base unit (open bottom) and a removable top.A stainless steel base unit (45 cm length, 9 40 cm width, 9 40 cm height) with a water groove (5 cm in depth) on the top was placed 10 cm deep in the soil for 3 days before transplanting to avoid lateral gas diffusion.The removable top (45 cm length, 9 40 cm width, 9 9 cm height) covered six hills of rice, and the plant density inside the chamber was the same as that outside of the chamber (see www.climaviet.org).Floodwater was used to seal the plexiglass top to the base unit during gas sampling.A rubber septum, thermometer, and two minifans (12 V) were installed at the top of the chamber [21].Pressure control (a plastic tube with 7.6 m length and 1.5 mm diameter) was also installed to maintain an equilibrium gas pressure between the inside and outside of the chamber and to minimize mixing of the internal chamber gases with the exterior atmosphere [22].Removable wooden boardwalks were set up in the early stages of the rice season to avoid soil disturbances during gas sampling.Samples for analysis of CH 4 and N 2 O were taken at the stages of transplant, tillering stage, stem elongation, panicle initiation, booting stage, flowering stage, milk stage, and dough stage.

Analysis of gas samples
Gas samples were analyzed by gas chromatography: CH 4 gas was determined by a flame ionization detector at 300°C and N 2 O was determined by an electron capture detector at 350°C.
For determining the GHG emission, the intensity of CH 4 or N 2 O emission (mg•m −2 •h −1 ) was calculated using the equation of Smith et al. [15].

Soil analysis
Soil samples were taken in the survey field at the rooting layer before the experiment.Soil sampling was carried out following the Vietnam standards: soil texture by a pipette (TCVN 8567:2010); soil pH by using a pH meter (TCVN 5979-2007); total OC by Walkley -Black (TCVN 9294:2012); total N by the Kjeldahl procedure (TCVN 7373:2004); total P by the colorimetric method (TCVN 8940:2011); total K by atomic absorption spectroscopy method (TCVN 8660:2011); cation exchange capacity (CEC) by the ammonium acetate extraction method; and available K 2 O (TCVN8662:2011) and available P 2 O 5 by Olsen (TCVN8661:2011).

Model description
The DNDC model was used to simulate GHG emissions from rice cultivation and some other annual crops.The DNDC model is a process-oriented computer simulation model of carbon and nitrogen biogeochemistry in agroecosystems.The model consists of two components: the first component, consisting of the soil climate, crop growth, and decomposition sub-models, predicts the soil temperature, moisture, pH, redox potential (Eh), and substrate concentration profiles driven by ecological drivers (climate, soil, vegetation, and anthropogenic activity).The second component, consisting of nitrification, denitrification, and fermentation sub-models, predicts emissions of carbon dioxide (CO 2 ), methane (CH 4 ), ammonia (NH 3 ), nitric oxide (NO), nitrous oxide (N 2 O), and dinitrogen (N 2 ) from plant-soil systems.Classical laws of physics, chemistry, and biology, as well as empirical equations generated from laboratory studies, have been incorporated into the model to parameterize each specific geochemical or biochemical reaction.The entire model forms a bridge between the C and N biogeochemical cycles and the primary ecological drivers [23].
The input data of the model included meteorology (temperature, precipitation, wind speed, solar radiation, humidity); cultivation (seeds, time of sowing, harvesting, fertilizers, watering, crop management, weeds, etc.); and soil (soil type, pH, bulk density, hydraulic conductivity, clay content, OC content, etc.) The output data of the model involved CH 4 and N 2 O emissions per unit of cultivated area and other indicators related to OC, Eh, etc.

Sensitivity assessment of the DNDC model for the calculation of emissions
The input sensitivity analysis scenarios are prepared based on the climate, soil, and actual farming practices in the fluviol farming system in Thai Binh.Sensitivity analysis of the inputs is performed by changing a single input parameter within an observable range while keeping all other inputs at the original parameter (baseline).Input factors for sensitivity assessment include climate factors (temperature, precipitation, humidity), soil properties (organic carbon content in soil, clay fraction, pH, bulk density, and porosity), or farming practices (the amount of nitrogen fertilizer and manure used).The selected initial value is rice grown on fluviol soil in Nguyen Xa commune, Vü Thu district, Thai Binh province.The rice variety used in the experiment is the one that is widely grown in the area.The soil pH ranged from 4.8 to 5.0; soil texture values (%) were 19.96% (<0.002 mm), 49.38% (0.002-0.02 mm), and 30.66% (0.02-0.2 mm); the soil organic content was 0.18%; plant available phosphorus was 182.09 mg/100 g; and CEC was 26.75 cmol•kg −1 .Fertilizer was applied three times/season without application of manure, and the plow depth was about 20 cm.The irrigation mode was still mainly regular flooding.The average daily temperature and daily rainfall data were obtained from the Thai Binh meteorological station in 2018.

Calibration of the DNDC model for emission calculation
The estimated results of CH 4 and N 2 O emissions by the DNDC model were compared with the measured ones at the experimental site in two rice seasons.The correlation between the two results was determined using the R 2 [24] and Nash -Sutcliffe efficiency index (NSI) [25], as follows: where i is the total number of tests, O i is the actual measured value i, Ō is the mean of the actual measured value, P i is the simulated value i that corresponds to O i , and n is the total number of tests.The R 2 value ranges from 0 to 1, representing the correlation between the actual measured value and the simulated value.The NSI value ranges from −∞ to 1, representing the match between the actual measured value and the simulated value on a 1:1 straight line.If R 2 and NSI are less than or close to zero, then the results are considered unacceptable or of poor reliability.On the contrary, if this value is equal to 1, the simulation results of the model are perfect.The model is accepted when the R 2 coefficient and NSI index are greater than 0.5.The model parameters that can be calibrated are the bulk density, clay fraction, drainage efficiency, porosity, wilting point, field capacity, conductivity, rainwater collection index, soil salinity index, initial nitrate concentration at the surface soil, initial ammonium concentration at the surface soil, pH, and microbial activity index.

Model verification
The model was verified by comparing the GHG emission calculation results of the model with the experimental data set from the study of Mai et al. [26], which measured the field emissions at Thinh Long, Hai Hau, and Nam Dinh.
Similar to the content of model calibration, this research has calculated the value of the coefficient of determination R 2 and the NSI to compare the fit between the calculated emission values by the model (after calibration) and observed values.

Spatial analysis
The DNDC model estimated GHG emissions for each polygon on the map unit.Then, ArcGIS 10.1 software was used to present all these results on thematic maps, the so-called maps of CH 4 , N 2 O, and CO 2 eq emission.

Statistical analysis
The data and results of the conducted experiments after applying the DNDC model were processed using Microsoft Excel.GHGs were converted to CO 2 eq with a factor of 28 for CH 4 and 265 for N 2 O, according to IPCC 2014 [1].

Physical and chemical properties of soils at experimental sites
Most of the measurement sites have organic and total N contents at high levels.For the organic content, because fluvisols outside the dyke are regularly accreted, the organic C was determined to be low of less than 1%; in contrast, due to fluvisols in low-lying areas being poorly drained and poorly mineralized, the organic C content was higher than 2.2%.Fluvisols with OC ranged from 0.9% to 2.61%; saline soils ranged from 0.4% to 2.29%; thionic soils were 3.3%, and gray soils were 1.23%.Thus, the organic content in thionic soils was much higher than that in the remaining three soil types.Additionally, most of the total N fluctuated in the range of 0.12%-2.7%.At most of the study sites, phosphorus and available potassium were high, especially available phosphorus, which was very high.The soils' CEC was medium to high, ranging from 12.6 to 26.7 cmol•kg −1 .Mainly, thionic soils with total phosphorus were in the middle and poor levels.
The soil texture was classified according to three levels: clay, silt, and sand.Fluvisols mainly had silt particles: clay from 21.4% to 31.4%, silt from 54.2% to 57.2%, and sand from 14.4% to 21.4%.Soils varied from sandy to silt and clay, depending on the topographical conditions, the distribution distance of the soils from the river, and the distribution upstream, midstream, or downstream of the river.Among rice cultivation practices, the amount of nitrogen fertilizer (urea) and the amount of manure are the two main farming activities that have a significant impact on seasonal CH 4 emissions.When the amount of nitrogen fertilizer is reduced by 25%, 50%, and 75%, CH 4 emissions decrease in the range of 0.12-14.25%.When the amount of nitrogen fertilizer is increased by 25%, 50%, and 75%, CH 4 emissions increase in the range of 0.13-0.47%.When the amount of manure is increased from no fertilizer to 1, 1.5, and 2 tons•ha −1 , the CH 4 emission rate increases from 10.55% to 15.37%.Cai et al. showed that the form of nitrogen fertilizer directly or indirectly affects CH 4 and N 2 O emissions.The application of the ammonium sulfate nitrogen fertilizer helps reduce CH 4 emissions but increases N 2 O emissions by 25-50% [11].
The fourth sensitive factor for CH 4 emissions is the microbial activity index.When the microbial activity index decreased in the range of 25-75% compared to the initial value, CH 4 emissions decreased in the range of 34.79-86.65%.In contrast, when the microbial activity index is increased within the range of 25-75% compared to the initial value, CH 4 emissions are decreased by 33.98-89.91%,respectively.Soil microorganisms play an important role in the production of CO 2 , CH 4 , and N 2 O in most terrestrial ecosystems.However, the level of GHG emissions between the soil and atmosphere depends on various factors that affect the development of microorganisms, such as soil oxygen content, underground water content, soil temperature, soil mineral N, organic matter and pH, or farming practices such as manure and N fertilization, plowing of plant residues, and tillage methods.Tillage can also affect the activities of microorganisms and indirectly affect CH 4 and N 2 O emissions from soil.
Organic matter content in soil is a relatively sensitive factor for CH 4 emissions due to the influence of methanogenic bacteria density.Many studies have shown that organic matter plays an extremely important role in the formation and emission of CH 4 as the starting material of CH 4 formation in soil.The stronger the decomposition process of organic matter in the soil, the lower the oxidationreduction potential, creating favorable conditions for the formation of CH 4 .The richer the soil is in organic matter, the more CH4 emissions there are [31,32].Sensitivity analysis results show that when the initial OC content is increased from 25% to 75%, CH 4 emissions increase by 18.22% and 52.99%, respectively.On the contrary, when the initial OC content decreased to 25-75%, the amount of CH 4 decreased by 18.89% and 62.69%, respectively.
Porosity and conductivity of water are also factors that affect CH 4 emissions.CH 4 emissions decrease with water conductivity or porosity.CH 4 emissions increase with increasing water conductivity or porosity.The conductivity of water is an index that describes the total dissolved ions in a solution.The movement of these ions creates an electromagnetic current, also known as ion conduction.The conductivity of water is proportional to the temperature of the water.
Compared to the analyzed factors, pH, drainage efficiency, depth of water retention layer, field capacity, and initial nitrate concentration at surface soil might have little impact on the CH 4 emissions.
Density and clay rates were also sensitive factors.Some parameters that did not affect CH 4 emissions were the wilting point in moisture soils and the salinity index.These results were similar to those reported by Li et al. (2004), Wassmann et al. (2000), and Yagi et al.Among rice cultivation methods, nitrogen (urea) and manure were the two primary farming practices that significantly impacted seasonal CH 4 emissions [5,30,34,35].

Sensitivity of parameters for N 2 O emissions
N 2 O emissions were not much influenced by meteorological and precipitation factors.while temperatures oscillated by 25%, 50%, and 75% of initial values, N 2 O emissions, respectively, increased by 3.59%, 12.92%, and 24.52% and decreased by 2.02%, 13.12%, and 24.01%.The reason is that microorganisms involved in the nitrification process decreased when the temperature decreased and increased significantly when the temperature increased.This result is consistent with the studies of Li et al. [3].
In addition, N 2 O emissions were impacted significantly by the clay rate and the microbial activity index.The parameters drainage capacity, water movement speed, and initial ammonium content in the topsoil at the study sites experienced a lower influence on N 2 O emissions from rice soils.The wilting point in moisture soil and salinity index were the unaffected factors.
The initial nitrate concentration at the surface soil is also a highly sensitive factor for calculating N 2 O emissions.When the initial nitrate concentration increased at the surface soil (0.3 mg•kg −1 ) to 25%, 50%, and 75%, the N 2 O emission level increased by 29.56%, 59.52%, and 89.74%, respectively.On the contrary, when the initial nitrate concentration decreased at the surface soil by 25%, 50%, and 75% compared to the initial value, the N 2 O emission level decreased by 28.94%, 56.72%, and 81.23%, respectively.
Similar to CH 4 emissions, bulk density is a very sensitive factor for N 2 O emissions.Bulk density positively affects the increase in N 2 O emissions.
Organic matter content in soil is the next most sensitive factor.Organic carbon plays an important role in the formation of N 2 O and N 2 in the soil by affecting the density of bacteria participating in nitrification and denitrification processes.Carbon can stimulate the growth and activity of microorganisms and provide the necessary organic carbon for reducing agents.The development of soil microorganisms increases O 2 consumption, accelerating the formation of anaerobic conditions necessary for denitrification.Zou et al. ( 2004) also showed that when the organic C content in the soil increased, N 2 O emissions were also increased.When the initial OC content increased from 25% to 75%, N 2 O emission levels increased by 75.2% and 78.3%, respectively.On the contrary, when the initial OC content is decreased, the N 2 O emission level decreases [33].
Some factors that also have a significant influence on N 2 O emissions are field capacity, porosity, and depth of the water retention layer.Field capacity and porosity are directly proportional to N 2 O emission levels, while the depth of the water retention layer is inversely proportional to N 2 O emission levels.Hua et al. showed that wet rice soil has both an aerobic surface soil layer and a deep anaerobic soil layer.It is the structure of two distinct soil layers in close proximity that creates conditions for the simultaneous occurrence of nitrification and denitrification reactions participating in the release of N 2 O.However, the nitrification process only dominates in the topsoil layer, while the denitrification process dominates in the deep soil layer during periods of high soil moisture.N 2 O produced from denitrification is often greater than that from nitrification [34].
Regarding fertilizers, nitrogen levels were positively linearly correlated with N 2 O emissions.The oscillation of the amount of N fertilizer applied to the soil by 25%, 50%, and 75% of the initial values increased or decreased N 2 O emissions by 2.29% and 10.72%, respectively.When the manure was changed from 0 to 1-2 ton•ha

Model calibration
The model coefficients were adjusted according to the measurement results at the experimental sites.After that, the CH 4 and N 2 O emissions calculated by the DNDC model were compared with the measured data in the field.Based on the CH 4 and N 2 O emission values measured in the field and calculated using the model to show the point distribution, the GHG emissions values were distributed close to the 1:1 line.Figures 4 and 5 show a good correlation between the actual and simulated values: R 2 in spring and summer crops reached 0.86 and 0.79, NSI reached0.82and 0.77 (for CH 4 ); R 2 in spring and winter crops reached 0.62 and 0.69, and NSI reached 0.69 and 0.76 (for N 2 O) (Figure 6).S3.

Current land-use map
Current land types in the study area are rice and other annual, urban areas, rural areas, and others, which are shown in Figure S1.The planted area of rice and annual upland crops in the Red River Delta are shown in Table S4.

Soil map in the Red River Delta
From the collected soil map, filtering and establishing the thematic map were carried out; thereby, a new soil map for the Red River Delta was formed with the main soil types being fluvisols, gray soil, saline soil, thionic soil, sand soil, sloping, peat soil, erosion, chalk soil, which is shown in Figure S2.The planted area of rice and annual upland crops by main soil types of the Red River Delta are shown in Table S5.

Complex map of meteorology-soil-land use in the Red River Delta region
From the coordinate data of meteorological stations, the current land-use map and the distribution map of the soil types, a complex map of meteorology-soil-land use (Figure 9) was built by using the overlay analysis method.
From the coordinate information of meteorological stations, the current land-use map and the soil map, a complex map of meteorology-soil-land use (Figure 8) was built using the overlay analysis method.Each unit of the complex map already contains information about the location of the land plot, meteorology, soil type, crop status, and area of the land parcel (cells).
The total area of annual rice and upland crops on four main types of land is 1,115.41thousand hectares, in which      Estimation of GHG from rice and annual upland crops using the DNDC model  13 the total area of four main types of land for rice cultivation is 727.87 thousand ha, and the total area of four main types of land for annual upland crops is 387.54 thousand ha.

GHG emissions from rice and upland
crops in the Red River Delta

GWP
N 2 O emission and GWP from upland crops were estimated using IPCC (2014) guidelines and were then presented on a map (Figure 11) with a unit of kgN 2 O•ha −1 •year −1 .N 2 O emissions from these range from 17,500 to 20,000 kg CO 2 eq•ha −1 •year −1 .In the coastal area, GWP ranges from 18,500 to 22,842.21kg CO 2 eq•ha −1 •year −1 (Figure 13).
For annual upland crops, the calculated results show that the area around the Thai Binh station has the lowest emission (average: 260.41 kg CO 2 eq•ha −1 •year −1 ), followed by the area around the Phu Lien station (average emission of 274.81 kg CO 2 eq•ha −1 •year −1 ).The area around the Nam Dinh station has the highest emission (average 346.09 kg CO 2 eq•ha −1 •year −1 ).Total CO 2 eq mostly fluctuates in the range of 310-355 kg CO 2 eq•ha −1 •year −1 due to high soil N content, light mechanical composition, and a large amount of N fertilizer.The result is shown on the emission map in Figure 14.The map of GWP (total GHG emissions in CO 2 eq) from rice and upland crop soil is presented in Figure 15.The map values show that the northern parts have a lower N 2 O emission (average: 4,586.33 kg CO 2 eq•ha −1 •year −1 ) than the Southern parts (average: 22,842.21kg CO 2 eq•ha −1 •year −1 ).The reason for that is the Southern parts have alluvial, saline, and alum soils, while the area around the Tam Dao station and Ha Nam station has only two types of soil: alluvial and gray.The monitoring and analysis results at the experimental sites show that gray soil has lower emissions than thionic soil, saline soil, and fluvisol soil.Besides, meteorological factors also affect the emission level, especially temperature and humidity.Lu et al. (2000) showed that a temperature increase of 15-25°C under 45-day anaerobic conditions significantly increased the decomposition of carbon in the soil [28].

Conclusion
This research monitored, analyzed, and calculated GHG emissions from rice crops at experimental sites in the Red River Delta.The DNDC model used to simulate GHG emission showed a good correlation between the observed and simulated values.The complex map of meteorologysoil-land use for the Red River Delta has been established from three single maps by overlaying operation.Each unit of this complex map contains information on climatic, soil, and crops, which uses input data for the DNDC model.Simulated results showed that CH 4 emissions ranged from 72.20 to 859.16 kg•ha −1 •year −1 , in which gray soils had the lowest CH 4 emission (average: 250.57kg•ha −1 •year −1 ) and thionic soils had the highest (average: 802.74 kg•ha −1 •year −1 ).Similarly, N 2 O emissions ranged from 0.306 to 2,247 kg•ha −1 •year −1 , and N 2 O emission was the lowest (average: 0.667 kg•ha −1 •year −1 ) in gray soils and highest (average 1,389 kg•ha −1 •year −1 ) in saline soils.GWP in CO 2 eq values are in the range of 17,500-22,842 kg CO 2 eq•ha −1 •year −1 , which tends to be higher in the coastal area.For annual crops, N 2 O emission ranged from 0.716 to 2,728 kg N 2 O•ha −1 •year −1 , in which N 2 O emission in thionic soils are lowest with an average of 0.723 kg N 2 O•ha −1 •year −1 , while N 2 O emission on fluvisols is highest with an average of 1,957 kg•ha −1 •year −1 .GWP in CO 2 eq area fluctuates from 310 to 355 kg CO 2 eq•ha −1 •year −1 .From these simulated results, we developed a thematic map on the distribution of GHG emissions (CH 4 , N 2 O, GWP) for each map unit representative of meteorological, soil, and land use conditions for the whole delta.
Funding information: The authors state no funding is involved.
Author contributions: The authors contributed equally to this work.

Figure 2 :
Figure 2: Location map of meteorological stations.
Fluvisols, Thai Binh, Fluvisols, Spring rice, Summer rice, and Winter crop from rice and annual upland crops using the DNDC model  5

Figure 3 :
Figure 3: Sensitivity of parameters for CH 4 emissions.

Figure 4 :
Figure 4: Sensitivity of parameters for N 2 O emissions.
−1 , the N 2 O emissions increased sharply.The trends in this research are similar to those of Li et al. and Bouwman et al. (2002) [4,6,7].
Similar to the data of model calibration, this study has calculated the value of the coefficient of determination R 2 and the NSI to compare the trend between the calculated emission values by model (after calibration) and observed values.The results of model verification show the following: (i) comparing CH 4 emissions observed and modeled in the spring crop, NSI = 0.79 and R 2 = 0.95; in the seasonal crop, NSI = 0.88 and R 2 = 0.95; and (ii) comparing the observed and modeled N 2 O emissions in the spring crop, NSI = 0.79 and R 2 = 0.81; in the seasonal crop, NSI = 0.73 and R 2 = 0.87.Thus, a good correlation between the field-measured values and the calculated values by the model was demonstrated.The model had a relatively high correlation (shown by the R 2 and NSI asymptotically up to 1).The results are presented in Figures7 and 8 .

3. 5
Input data for the model 3.5.1 Meteorological data The coordinates of meteorological stations in the region are shown in Figure 2. The climate characteristics of each station are shown by daily data, and the average meteorological parameters are shown in Table

Figure 5 :
Figure 5: Simulated and field-measured CH 4 emissions in the (a) spring season and (b) summer season.

Figure 6 :
Figure 6: Simulated and field-measured N 2 O emissions in the (a) spring season and (b) summer season.

Figure 7 :
Figure 7: Simulated and field-measured CH 4 emissions in spring and summer seasons.

Figure 8 :
Figure 8: Simulated and field-measured N 2 O emissions in spring and summer seasons.

Figure 9 :
Figure 9: Complex map of meteorology-soil-land use in the Red River Delta region.

Figure 11 :
Figure 11: Map of N 2 O emissions from rice soil in the Red River Delta.

Figure 10 :
Figure 10: Map of CH 4 emissions from rice soil in the Red River Delta.

3. 6 . 1
GHG emissions from rice cultivationBased on the results of the DNDC simulation for each unit of the complex map of meteorological-soil-land use, thematic maps showing the distribution of CH 4 emissions from rice soil (Figure10) and N 2 O emissions distribution from rice soil (Figure11) were established.The results show a high range of CH 4 emissions from 600 to 700 kg•ha −1 •year −1 .In coastal areas, CH 4 emissions range from 755 to 859 kg•ha −1 •year −1 .The CH 4 emissions were lowest in gray soils with an average of 250.57kg•ha −1 •year −1 and highest in thionic soils with an average of 802.74 kg•ha −1 •year −1 .Comparing the results between the measured and the DNDC model, we found a correlation between the content of soil organic matter and the soil and emissions.For instance, gray soil has lower organic matter content and lower emissions than other soil types, while thionic soil has the highest organic matter content and higher CH 4 emissions.This trend agrees well with the measurement trend and previous studies by Pandey et al., Tariq et al., and Mai et al. [35-37].N 2 O emissions range widely from 0.55 to 2.0 kg•ha −1 •year −1 .N 2 O emissions differ by the soil type; the gray soil has the lowest N 2 O emission with a value of 0.667 kg•ha −1 •year −1 , and saline soil has the highest N 2 O emission with a value of 1,389 kg•ha −1 •year −1 .There is a correlation between the soil nitrogen content and N 2 O emissions.For instance, gray soil had lower nitrogen content and lower N 2 O emissions, while thionic soil had high nitrogen content and the highest N 2 O emission.This finding agrees with field measurement results and previous studies of Arjun et al. [38], Wassmann et al.[35], and Azeem et al.[37].

Figure 12 :
Figure 12: Map of N 2 O emissions from annual upland crop soil in the Red River Delta.

3. 6 . 2
GHG emissions from annual upland crops in the Red River Delta Based on the results of the DNDC simulation results for each unit on the complex map of meteorological-soilland use, thematic maps showing N 2 O emissions distribution from the annual upland crops soil (Figure 12) were established.The N 2 O emission values on the map range from 1.2 to 1.35 kg N 2 O•ha −1 •year −1 .The highest N 2 O emission is concentrated around the Ha Noi center, with emission rates higher than 1.3 kg N 2 O•ha −1 •year −1 , while the lowest emission points are distributed around the Hai Phong city and Thai Binh province with about 0.7-0.9kg N 2 O•ha −1 •year −1 .Thionic soils had the lowest N 2 O emissions (average: 0.723 kg N 2 O•ha −1 •year −1 ); in contrast, fluvisols witnessed the highest N 2 O emissions (average: 1,957 kg•ha −1 •year −1 ).The values ranged from 0.716 to 2,728 kg N 2 O•ha −1 •year −1 .

Figure 13 :
Figure 13: Map of GWP from rice in Red River Delta.

Figure 14 :
Figure 14: Map of GWP from annual upland crops in Red River Delta.

Appendix 3 :
Figure A1: Example of climate data that is one of input data of DNDC model.

Figure A2 :
Figure A2: Example of soil data that is one of input data of DNDC model.

Figure A3 : 29 Figure A4 :
Figure A3: Example of crop data that is one of input data of DNDC model.

Figure A5 :
Figure A5: Example of irrigation that is one of input data of DNDC model.

Figure A6 :
Figure A6: Example of flooding that is one of input data of DNDC model.

Figure A9 :
Figure A9: Example of step 3 -Run the model.