Development of a Spatial Tier 2 Emission Inventory for Agricultural Tractors by Combining Two Large-Scale Datasets

Abstract

Agricultural tractor emissions have a notable influence on total emissions and climate change. This is especially important when developing an agricultural machinery emission inventory since tractors are the most widely used agricultural machinery. However, as with other agricultural machinery, there is a significant lack of data concerning tractors. This study combined two large-scale datasets to develop a Tier 2 tractor emission inventory for the Republic of Croatia. The first dataset, which was collected via a large-scale survey with 8895 respondents, includes data on fuel consumption by machine and engine type. The second dataset, which is official data of the Ministry of the Interior, includes data on machinery age. Since there is no direct link between datasets, data on fuel consumption were calculated for each of 21 Croatian counties and were then linked to the corresponding municipalities’ tractor age and power rating of the vehicle fleet. The result was used to calculate tractor emissions by with a Tier 2 approach using the method listed in the 2019 EMEP/EEA emissions guidebook. The calculations were made for five exhaust emission components: carbon dioxide (CO₂), carbon monoxide (CO), hydrocarbon (HC), nitrogen oxide (NO_X) and particulate matter (PM). The continental counties had the highest emissions for all emission types, especially the Osječko-baranjska and Vukovarsko-srijemska counties. The Tier 2 emission inventory was compared to data calculated using the Tier 1 method on a county level. There were notable differences between the inventories for HC and PM emissions, which is mainly due to Tier 2 approach having different emission factors depending on vehicle age. The end result was compared to the already existing Tier 1 emission inventory on a county level.

1. Introduction

Tractors are extensively used in the agricultural sector, with most work in the field being carried out using tractors. A majority of agricultural tractors release emissions into the air because of their use of fossil fuels, which are usually divided into greenhouse gas emissions and air pollutant emissions. Greenhouse gas emissions greatly contribute to climate change [1] by emitting radiant energy, thus keeping heat in the atmosphere [2]. Air pollutant emissions are substances that have a negative impact on health. Exposure to high levels of air pollutants primarily leads to problems with respiratory and cardiovascular systems [3,4], but can also lead to genetic disorders and negative structural changes in the brain during prenatal exposure [5,6]. Most air pollutant emissions are generated through human activities [7]. Global emissions are rising and accelerating climate change [8], which makes reducing emissions one of the main global goals. An important step to reducing emissions and achieving climate neutrality was the Paris Agreement. The intention of the Paris Agreement is to keep the increase in global temperature to well below 2 °C compared to preindustrial levels. In order to achieve this, greenhouse gas emissions must be lowered by 40% by 2030 compared with 1990 [9], which was confirmed by the Glasgow Climate Pact [10]. The European Union further increased emission reduction to 55% to keep the increase in global temperature below 1.5 °C [11,12]. However, it is possible that residual emissions will require offsets from carbon dioxide removal technologies to make agriculture climate neutral [13], as higher agricultural activity is often tied with larger emission levels on a national scale [14].

The first major step in identifying and reducing emissions is developing an emission inventory. It enables the definition and listing of emissions by source emitter, activity, spatial or temporal distribution, emission type, etc., making it a valuable and reliable tool in developing new policies [15,16] and procedures aimed at reducing emissions. With detailed data, new policies can be implemented at the local level to increase the immediate impact on the local community [17]. Moreover, existing policies can be tested and improved [18], as direct agricultural subsidies can encourage agricultural-related carbon emissions [19], but subsidies directly aimed at reducing emissions can be beneficial to reducing emissions in the agricultural sector [20]. New procedures, such as new standards or better measuring devices, reduce overall emission inventory uncertainty and improve its accuracy [21,22].

Since development of an emission inventory is usually carried out periodically and on a large scale, it is a long-term process that entails constant data collection from a multitude of sources. Consequently, there is a lack of data in some emission inventory areas. Such is the case for nonroad mobile machinery (NRMM). The main reason for this lack of data is that NRMM is not perceived as large a pollutant group as road vehicles. One of the results are more liberal emission limits for NRMM compared to road vehicles worldwide [23], including the EU [24], compared to more rigorous emission limits for road vehicles [25]. Moreover, there is less research concerning NRMM, as well as less data collection from official bodies. Since official data collection is a fundamental step in developing an NRMM emission inventory, there should be an institution that is appointed to collect the data [26]. Resulting inventory could be used to track emission reduction technology impact on a national scale, pace of NRMM replacement and spatial impact of NRMM emissions. Furthermore, the results should be presented in an easily accessible and understandable format. One example of such result format is the Environmental Protection Agency in the USA, which offers NRMM emission data disaggregated by federal states [27]. In the United Kingdom, data on NRMM emissions is available spatially in a visually engaging way [18].

Reducing NRMM emissions is also a crucial task in the context of achieving climate neutrality. For example, although there are no official data concerning NRMM emissions on the EU level, a review of Directive 97/68/EC states that 2% of total EU greenhouse gas emissions in 2010 came from NRMM [28]. Concerning pollutant emissions, a report from 2017 stated that NRMM was responsible for 15% of total NO_X emissions and 5% of PM emissions [26]. The first step in achieving emission decrease would be to switch out old machinery with newer models [29], or to retrofit old machinery with new emission reduction technology. Emission reduction technologies, such as diesel particle filters, NO_X adsorbers, selective catalytic reduction, oxidation catalysts and gasoline particle filters, offer solutions aimed at reducing emissions from agricultural tractors [30]. Considering fugitive dust emissions, the light scattering device could be a cost-friendly way of achieving lower dust emissions during fieldwork [31]. Hybrid technologies and electricity as new propulsion technologies are necessary for achieving climate neutrality in the agricultural sector [32,33]. However, this brings additional cost, which can be over 10,000 USD per engine for larger engines with over 450 kW of engine power [34].

Agricultural machinery is a significant area of NRMM research, as well as of NRMM emissions. Since tractors are used in almost all stages of agricultural fieldwork, they represent the major share of agricultural NRMM. An emission inventory would improve the current understanding of agricultural tractor emissions and their influence on total national emissions, as well as on local population. It would also serve as a basis to formulate and update long-term policies aimed at reducing their emissions [35,36]. The Beijing region in China is a good example of policy influence on emission reduction. From 2006 to 2016, emissions of carbon monoxide (CO), hydrocarbons (HC), nitrogen oxides (NO_X), particulate matter (PM) and the total emissions from agricultural machinery decreased by 63.11%, 62.93%, 72.07%, 74.67% and 68.66%, respectively, as a result of policies aimed at emission reduction [37]. In future, China could offset its agricultural emissions by 2060 [38]. Emissions from agricultural work are also temporal in nature, with air quality often being poorer during a heavy agricultural work season [39,40].

A difficulty concerning tractors, as well as agricultural machinery as a group, is a lack of research concerning its activity data and emissions during fieldwork. Tractors can work on different crops, land type, terrain, with different load profiles, etc., which makes it difficult to determine their emission factors with adequate accuracy. Gathering these data is important to lower overall emission uncertainty and for the emission inventory to represent the real situation in the field [41], as the measurement results in the field can vary to as high as 20 times [42]. Emission differences can also be large as a result of terrain. Vineyards are typically on a rougher terrain compared to arable land, and machinery working in vineyards can emit up to 80% more emissions for the same area compared to machinery working on arable land [43].

Various methods of determining agricultural machinery emissions can be used based on available data. The EMEP/EEA Air Pollutant Emission Inventory Guidebook lists three methods [44]. If only basic machinery data are available, i.e., fuel consumption, then the Tier 1 method is used by providing emission factors for basic categories, such as agricultural machinery category. The Tier 2 method is used where activity data are collected on a more detailed basis, based on equipment type and emission reduction technology level. The emission reduction technology levels are divided according to the EU NRMM emission legislation, and the EMEP/EEA Air Pollutant Emission Inventory Guidebook does not prescribe the technological solutions by which these emission limits are achieved. The most detailed is the Tier 3 method, which is used when detailed data are available on the equipment level, e.g., machinery age, annual hours of use, national specific emission factors and engine size. The basic principle is to use as much available data as possible, especially if the considered category is a key emission source. The 2019 EEA EMEP/EEA Air Pollutant Emission Inventory Guidebook states that construction and agriculture sectors are the largest emitters of NO_X and PM emissions [44]. The same sectors are the largest emitters in Switzerland, with the largest proportions in NO_X and PM emissions [45]. Agricultural and construction machinery are also the largest NO_X and PM emitters in China [46].

Because it is challenging to obtain even the most basic data, most countries do not have an NRMM emission inventory, which includes agricultural tractors. Even the existing inventories are mostly developed by using data on machinery sales or machinery imports. In Denmark, the Association of Danish Agricultural Machinery Dealers provides annual sales data, but there are no available activity data [47]. In Germany, agricultural machinery emissions are reported by the Federal German Environment Agency using vehicle stock data. However, for activity data, resale platforms are used [29]. In Switzerland, data is extrapolated based on previous data on fuel sales for agricultural machinery, which are available up to 2004 [45]. A similar situation is present in the Netherlands, where only vehicle sales data are collected, but no activity data are available [48]. Finland and Sweden, which collect agricultural machinery data annually, have much more detailed agricultural machinery emission inventories based on the Tier 3 method. Finland uses the TYKO calculation model with data on engine power, vehicle age, annual hours of work and average load factor. However, these data are not collected on a national level but only partially collected [49]. Sweden also provides detailed emission data on real-world emissions during fieldwork [50]. However, as in Finland, data are not collected on a national level but extrapolated from partially collected data [51]. Another issue concerning agricultural tractors is their historical data. Due to their replacement years often being 30 years or even more [52], there are often even less or no basic data concerning older agricultural tractors.

In order to develop a reliable emission inventory for tractors, data should be collected through a large-scale survey if collecting data on a national level is not possible. A questionnaire with data on machinery type, land type, fuel consumption, fuel type and vehicle age would collect enough data for a Tier 2 emission inventory without requiring detailed data from the respondents. Questionnaires also provide data about fuel consumption during fieldwork, which is beneficial for the emission inventory accuracy [40,53]. The main problem with this approach is that conducting such a survey on an annual basis is difficult without government bodies due to funding demands [54]. The Netherlands [55] and Sweden [56] conduct such surveys, which enable them to develop Tier 3 emission inventories.

There are no official data on tractor emissions in Croatia. Since there is no continuous data collection on any data except the number of registered tractors, it is difficult to develop an initial emission inventory with a satisfactory level of accuracy. A Tier 1 emission inventory for agricultural machinery was developed based on large-scale survey data, with data for 2016 [57]. In this research, tractors were identified as the largest emitters among agricultural machinery. Although data in this research included data on fuel consumption and spatial data, there were no data on vehicle age, which is necessary to develop an inventory using the Tier 2 method. However, since tractors in Croatia need to be registered, there is a large dataset with a list of all registered tractors in Croatia. This list is managed by the Croatian Ministry of the Interior. This paper aims to develop a Tier 2 emission inventory for agricultural tractors in Croatia by combining two large datasets: one with data on fuel consumption and the other with data on vehicle age. Both datasets contain spatial data on the machinery’s county, which means that the emission inventory is spatially distributed by county. Such a method enables the development of an emission inventory with a satisfying level of accuracy without having to use just one dataset, or multiple datasets that need to have a unique identifier for every entry. Tractor emissions calculated in this paper are also compared to tractor emissions calculated using the Tier 1 method for each county.

2. Materials and Methods

2.1. Data Collection and Database Comparison

Since there is no database with both data on fuel consumption and vehicle age, which is needed to develop a Tier 2 emission inventory, two different databases where combined. As there are no official data on fuel consumption of agricultural tractors in Croatia, a previous study was used for fuel consumption data [57]. In this study, data were collected via a questionnaire in collaboration with the Croatian Bureau of Statistics (CBS), and it included 8895 respondents. The respondents filled in data on machinery type, type of fuel used, quantity of fuel used and agricultural land area. In order to ensure high quality of data on fuel consumption, several plausibility control steps were taken. This was enabled by combining survey data with the existing CBS data on agricultural land area and respondent location for each survey respondent. By comparing individual fuel consumption per hectare of arable land with the county average, outliers and illogical inputs could be removed from the final data. Furthermore, a manual plausibility control was conducted for every record of consumption. If consumption differed an order of magnitude, or was more different from the average, the input was removed from the final data. The top and bottom 5% of fuel consumption were also controlled for each county. In order to gather the most reliable data possible, another plausibility control was introduced. A question requiring data on total fuel bills for 2016 was included in the survey. Since in Croatia legal agricultural and marine entities are allowed to use so-called “blue diesel” fuel but must keep all the bills and use the fuel exclusively for business, survey responses regarding fuel consumption could be compared to responses regarding “blue diesel” bills. “Blue diesel” has the same properties as regular diesel fuel (according to EN 540) but is exempted from certain tax duties, which makes it approximately 40% cheaper [57]. The first database contains data on the number of agricultural tractors, spatial data, fuel type (diesel or gasoline) and fuel consumption.

The second database contains data on the number of agricultural tractors, their vehicle age and their spatial data, i.e., county name. This database is operated by the Croatian Ministry of the Interior and contains data on each registered agricultural tractor. Since there is no common unique identifier that could pair the data from the two databases, data were aggregated on a county level. Croatia is divided into twenty-one counties that have large variations in size and population, as well as primary activity due to geographical conditions/differences. The main agricultural activity is present in the continental counties.

2.2. Data Processing and Database Integration

The EMEP/EEA Tier 2 method uses different emission factors based on the technology type of agricultural machinery, i.e., based on machinery age. The main emission categories are divided by stages as they appear in regulations of the European Union. The temporal distribution is different depending on machinery engine power rating and is the same both for diesel and gasoline engines, as shown in Figure 1.

The first database contains data on fuel consumption by county and by fuel type. In terms of determining tractor emissions using the EMEP/EEA Tier 2 method, data on vehicle age were missing. Since there are no data on vehicle age in this database, data from the second database were used. In the second database, although there were no data on fuel consumption, there were data on engine power rating, county and vehicle age for each tractor. These data were aggregated on a county level and for each technology type, as defined in the EMEP/EEA Tier 2 method: pre-1981, 1981–1990, 1991–Stage I, Stage I, Stage II, Stage IIIA, Stage IIIB and Stage IV. They were also aggregated by engine power class, as shown in Figure 1. This way, in the second database, there were data on a number of tractors in each county for every power class and every technology type, which were then combined with data from the first database, i.e., data on fuel consumption for all tractors in every county.

By combining data from the two databases, fuel consumption by vehicle age, power category, fuel type and county could be assessed. Fuel consumption for each category was determined according to the following formula:

FC_i,j,k,m = TFC_i,m × (N_i,j/NT_i) × (NC_i,j,k/N_i,j),

(1)

where FC_i,j,k,m is the fuel consumption; TFC_i,m is total fuel consumption on a county level; N_i,j is the number of agricultural tractors of the same power class on a county level; NT_i is the total number of agricultural tractors on a county level; NC_i,j,k is the number of agricultural tractors of the same power class and technology type on a county level; i is county number (i = 1, … , 21); j is power category (j = 1, … , 5); k is technology level (k = 1, … , 8); and m is fuel type (m = 1, 2).

This enabled the disaggregation of fuel consumption from the first database, which contains only data on fuel consumption on a county level, by engine power class and technology level for each county. After disaggregating the data, emissions could be determined and presented spatially by county for different power categories, fuel type and technology level using the Tier 2 method.

2.3. Calculating Agricultural Tractor Emissions

Total emissions from agricultural machinery were calculated for carbon dioxide (CO₂), carbon monoxide (CO), hydrocarbon (HC), nitrogen oxide (NO_X) and particulate matter (PM) emissions using the method listed in the 2019 EMEP/EEA emissions guidebook. The Tier 2 method was used, since data disaggregation by using two databases enabled data on fuel consumption by power class and technology type [44]:

E_i,j,k,m,n = Σ (FC_i,j,k,m × EF_m,n),

(2)

where E_i,j,k,m,n is the emission of the pollutant for each fuel type from a machinery category in a county; FC_i.j,k,m is the fuel consumption for each fuel from a machine category in a county; EF_m,n is the emission factor for a specific pollutant for each fuel type; i is county number (i = 1, …, 21); j is power category (j = 1, … , 5); k is technology level (k = 1, … , 8); m is fuel type (m = 1, 2); and n is pollutant type (j = 1, … , 5). Emission factors were obtained from the EMEP/EEA emissions guidebook and are shown in

Table 1. Emission factors of gasoline fuel for agricultural machinery for Tier 2 method.

Pollutant	Emission Factor (kg/ton of Fuel) 1 per Technology Level
	Pre-1981	1981–1990	1991–Stage I	Stage I	Stage II	Stage IIIA	Stage IIIB	Stage IV
CO2	3197	3197	3197	3197	3197	3197	3197	3197
CO	1214.855	836.966	768.445	774.457	804.157	804.157	804.157	804.157
CH4 2	0.710	0.910	0.672	0.650	0.568	0.568	0.568	0.568
NMVOC 2	20.182	25.852	19.082	18.469	16.126	16.126	16.126	16.126
NOX	2.429	5.743	7.129	7.088	6.676	6.676	6.676	6.676
PM10 3	0.148	0.147	0.157	0.159	0.159	0.159	0.159	0.159
PM2.5 3	0.148	0.147	0.157	0.159	0.159	0.159	0.159	0.159

¹ Values for four-stroke gasoline are used; ² CH₄ and NMVOC emissions are summed and presented as CH emissions in the results; ³ PM₁₀ and PM_2.5 emissions are summed and presented as PM emissions in the results.

and

Table 2. Emission factors of diesel fuel for agricultural machinery for Tier 2 method.

Pollutant	Emission Factor (kg/ton of Fuel) per Technology Level
	Pre-1981	1981–1990	1991–Stage I	Stage I	Stage II	Stage IIIA	Stage IIIB	Stage IV
CO2	3160	3160	3160	3160	3160	3160	3160	3160
CO	19.804	17.566	14.147	6.463	6.104	6.035	6.087	6.024
CH4 1	0.191	0.158	0.11	0.038	0.029	0.029	0.013	0.013
NMVOC 1	7.760	6.439	4.493	1.544	1.181	1.173	0.544	0.530
NOX	29.901	37.383	49.002	30.799	20.612	12.921	9.318	1.587
PM10 2	5.861	4.047	1.974	0.947	0.624	0.550	0.099	0.099
PM2.5 2	5.861	4.047	1.974	0.947	0.624	0.550	0.099	0.099

¹ CH₄ and NMVOC emissions are summed and presented as CH emissions in the results; ² PM₁₀ and PM_2.5 emissions are summed and presented as PM emissions in the results.

[44].

Spatial visualisation of emissions was conducted for 21 counties in Croatia, which enables the identification of areas of more considerable atmospheric impact of agricultural machinery, i.e., tractors. An additional spatial visualisation was made by comparing emissions calculated using the Tier 2 method to ones calculated using the Tier 1 method.

3. Results

3.1. Number of Agricultural Tractors and Fuel Consumption

The number of agricultural tractors per technology level and per county are shown in

Table 3. The number of agricultural machines by technology level per county.

County	Number of Agricultural Tractors per Technology Level
	Pre-1981	1981–1990	1991–Stage I	Stage I	Stage II	Stage IIIA	Stage IIIB	Stage IV	% Pre-Stage I
Bjelovarsko-bilogorska	4068	4259	1304	383	407	447	42	35	88%
Brodsko-posavska	1212	1765	543	203	297	269	48	19	81%
Dubrovačko-neretvanska	64	59	34	8	20	65	10	1	60%
Grad Zagreb	373	314	164	65	107	180	44	28	67%
Istarska	1151	1015	421	86	262	531	54	23	73%
Karlovačka	2935	2890	653	107	266	201	19	17	91%
Koprivničko-križevačka	4457	3860	1465	446	436	468	72	70	87%
Krapinsko-zagorska	2079	2399	826	264	367	350	40	30	83%
Ličko-senjska	900	1061	219	28	86	56	4	5	92%
Međimurska	1809	2214	689	171	220	291	72	58	85%
Osječko-baranjska	2614	3685	1186	514	754	971	338	135	73%
Požeško-slavonska	1558	2227	473	112	236	245	50	31	86%
Primorsko-goranska	772	602	147	22	74	114	15	5	87%
Sisačko-moslavačka	3880	3061	929	203	265	280	35	29	91%
Splitsko-dalmatinska	513	539	122	19	43	68	8	2	89%
Šibensko-kninska	79	83	25	1	9	40	1	1	78%
Varaždinska	4027	3873	1371	356	379	351	54	20	89%
Virovitičko-podravska	1673	2454	893	220	493	518	106	52	78%
Vukovarsko-srijemska	1328	2200	998	380	523	687	228	104	70%
Zadarska	224	288	81	12	52	88	10	6	78%
Zagrebačka	5225	4753	1735	529	595	779	102	58	85%
Total	40,941	43,601	14,278	4129	5891	6999	1352	729	82%

, while the distribution of technology levels by engine power category is shown in Figure 2. Furthermore, data on the percentage of tractors with technology level older than Stage I per county are shown. Zagrebačka county had the largest number of agricultural tractors, with 13,776 tractors, followed by Koprivničko-križevačka county, with 11,274 tractors. All counties, with the exception of Dubrovačko-neretvanska county and Grad Zagreb, have at least 70% of tractors with a technology level older than Stage I. A spatial display of total number of agricultural tractors and of percentage of tractors with a technology level older than Stage I is shown in Figure 3a,b, respectively.

A similar distribution is present when agricultural tractors are aggregated by engine power categories. Tractors with a technology level of Stage I or older make up for more than 50% of each power category for tractors with up to 75 kW of engine power. Most of the agricultural tractors were in the 19–37 kW engine category, with 72,655 tractors, or 60.7% of all tractors. The category with 130–560 kW of engine power had the least agricultural tractors, with 1234 tractors, or 1%.

Concerning fuel consumption, Osječko-baranjska county had the largest fuel consumption, with 23,505 tonnes of fuel and a total of 10,197 agricultural tractors, followed by Vukovarsko-srijemska county with 15,167 tonnes of fuel and a total of 6448 agricultural tractors. Such a result is to be expected because these two counties are in the lowland area where the largest part of agricultural production in the Republic of Croatia takes place. Counties with the largest number of agricultural tractors, Zagrebačka and Korpivničko-križevačka counties, had significantly lower fuel consumptions of 6787 tonnes and 10,149 tonnes, respectively. Osječko-baranjska county had the largest fuel consumption and less tractors than Zagrebačka and Korpivničko-križevačka counties. It also had a similar percentage of tractors across engine power categories, as shown in

Table A1. Number of agricultural tractors by county and by power category.

County	Number of Agricultural Tractors per Engine Power 1 Category
	0–19	19–37	37–56	56–75	75–130	130–560
Bjelovarsko-bilogorska	253	6025	2773	1151	654	89
Brodsko-posavska	84	2512	935	468	274	83
Dubrovačko-neretvanska	27	166	28	25	12	3
Grad Zagreb	114	669	186	192	92	22
Istarska	221	2210	594	391	95	32
Karlovačka	364	5454	841	322	84	23
Koprivničko-križevačka	469	6354	2649	1149	572	81
Krapinsko-zagorska	290	4613	1008	371	69	4
Ličko-senjska	79	1957	222	77	22	2
Međimurska	230	3728	756	504	259	47
Osječko-baranjska	141	4553	2362	1451	1354	336
Požeško-slavonska	89	3203	914	459	217	50
Primorsko-goranska	383	1101	161	76	22	8
Sisačko-moslavačka	313	5768	1620	584	329	68
Splitsko-dalmatinska	132	899	182	66	24	11
Šibensko-kninska	24	160	25	18	6	6
Varaždinska	507	7625	1596	491	195	17
Virovitičko-podravska	113	3772	1232	719	484	89
Vukovarsko-srijemska	107	2458	1748	1105	867	163
Zadarska	62	529	91	51	19	9
Zagrebačka	517	8899	2644	1146	479	91
Total	4519	72,655	22,567	10,816	6129	1234

¹ Engine power ratings are expressed in kilowatts.

in the appendix. This indicates that, in general, agricultural tractors in Osječko-baranjska county have higher work hours per agricultural tractor. Almost all of the used fuel is diesel fuel, with 124,215 tonnes out of the total 124,390 tonnes of fuel used by agricultural tractors. Total fuel consumption and total number of agricultural tractors are shown in

Table 4. The number of agricultural tractors and diesel and gasoline fuel consumption per county.

County	Total Number of Agricultural Tractors	Fuel Consumption (Tons of Fuel)
		Diesel	Gasoline
Bjelovarsko-bilogorska	10,945	9923	6
Brodsko-posavska	4356	8269	6
Dubrovačko-neretvanska	261	1393	55
Grad Zagreb	1275	2146	11
Istarska	3543	3108	6
Karlovačka	7088	2757	1
Koprivničko-križevačka	11,274	10,144	5
Krapinsko-zagorska	6355	2275	0
Ličko-senjska	2359	1813	1
Međimurska	5524	3238	4
Osječko-baranjska	10,197	23,500	6
Požeško-slavonska	4932	5266	7
Primorsko-goranska	1751	650	0
Sisačko-moslavačka	8682	6826	9
Splitsko-dalmatinska	1314	3938	11
Šibensko-kninska	239	1324	7
Varaždinska	10,431	3637	14
Virovitičko-podravska	6409	9023	3
Vukovarsko-srijemska	6448	15,167	2
Zadarska	761	3037	14
Zagrebačka	13,776	6782	6
Total	117,920	124,215	175

. Total fuel consumption is also shown spatially in Figure 4a.

Fuel consumption per agricultural tractor is shown in Figure 4b. Šibensko-kninska, Dubrovačko-neretvanska and Zadarska county had the largest fuel consumption per agricultural tractor with 5568, 5550 and 4010 kg of fuel per agricultural tractor, respectively. This could be due to terrain type in the coastal area of the country, which is a rougher terrain with more steep slopes compared to the continental part of the country. Furthermore, boats in mariculture also use blue diesel fuel. This fuel could be falsely represented as blue diesel used in the agricultural sector, which lowers data accuracy. Varaždinska and Krapinsko-zagorska counties had the smallest fuel consumption per agricultural tractor with 349 and 357 kg of fuel per agricultural tractor, respectively.

3.2. Emissions from Agricultural Tractors

After determining fuel consumption of agricultural tractors, Equations (1) and (2) were used to calculate CO₂, CO, HC, NO_X, and PM emissions. The emissions were calculated for each fuel type, engine power class and technology level before being aggregated on a county level. Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8 show a spatial distribution of total emissions and Tier 1 and Tier 2 differences for all emission types by county. The Tier 1 and Tier 2 differences are shown with the Tier 1 method as a baseline. For example, the difference for HC emissions for Šibensko-kninska county is 5%, which means that HC emissions for Šibensko-kninska county calculated using the Tier 2 method are 5% higher compared to those calculated using the Tier 1 method. A negative percentage indicates lower emissions for Tier 2 compared to Tier 1. The numbers used to prepare those figures are provided in Appendix A in

Table A2. Total emissions of agricultural tractors by county and emission type.

County	Emissions (t, kt for CO2)
	CO2	CO	HC	NOX	PM
Bjelovarsko-bilogorska	31,361	171	62	336	81
Brodsko-posavska	26,114	135	47	275	60
Dubrovačko-neretvanska	4579	68	7	40	8
Grad Zagreb	6805	40	11	64	14
Istarska	9839	51	17	95	21
Karlovačka	8713	48	18	93	24
Koprivničko-križevačka	32,069	172	63	341	82
Krapinsko-zagorska	7188	36	13	76	17
Ličko-senjska	5734	33	12	62	16
Međimurska	10,238	56	19	108	25
Osječko-baranjska	74,228	351	124	740	157
Požeško-slavonska	16,641	93	32	176	42
Primorsko-goranska	2055	11	4	21	5
Sisačko-moslavačka	21,601	126	45	229	59
Splitsko-dalmatinska	12,478	77	25	132	33
Šibensko-kninska	4206	27	8	41	10
Varaždinska	11,536	74	23	124	30
Virovitičko-podravska	28,514	140	50	296	63
Vukovarsko-srijemska	47,896	215	75	481	93
Zadarska	9632	60	17	96	22
Zagrebačka	21,443	116	41	226	54
Total	392,871	2100	713	4053	915

and

Table A3. Difference in Tier 1 and Tier 2 emission inventories by county and emission type.

County	Difference in Tier 1 and Tier 2 Emission Inventories (%)
	CO2	CO	HC	NOX	PM
Bjelovarsko-bilogorska	0	45	48	−2	112
Brodsko-posavska	0	36	38	−3	90
Dubrovačko-neretvanska	0	36	−62	−18	37
Grad Zagreb	0	27	−5	−14	63
Istarska	0	31	22	−11	80
Karlovačka	0	50	34	−2	125
Koprivničko-križevačka	0	44	51	−2	111
Krapinsko-zagorska	0	39	39	−3	99
Ličko-senjska	0	51	50	−1	124
Međimurska	0	42	42	−3	102
Osječko-baranjska	0	28	40	−9	75
Požeško-slavonska	0	43	46	−3	106
Primorsko-goranska	0	46	57	−6	120
Sisačko-moslavačka	0	50	47	−2	125
Splitsko-dalmatinska	0	49	38	−3	118
Šibensko-kninska	0	38	5	−10	90
Varaždinska	0	38	5	−1	90
Virovitičko-podravska	0	47	24	−5	113
Vukovarsko-srijemska	0	33	35	−8	82
Zadarska	0	23	30	−8	60
Zagrebačka	0	37	−5	−3	86
Average	0	42	44	−6	107

. Results in the form of bar charts are presented in Appendix A in Figure A1, which shows distribution of total CO₂, CO, HC, NO_X and PM emissions by county, and Figure A2, which shows weighted average of Tier 1 and Tier 2 differences for CO₂, CO, HC, NO_X and PM by county.

Figure 5a,b show the spatial distribution of total CO₂ emissions and Tier 1 and Tier 2 differences for CO₂, respectively. Osječko-baranjska county had the most significant total CO₂ emissions, with 74.2 kt, followed by Vukovarsko-srijemska county, with 47.9 kt of CO₂ emissions. Concerning Tier 1 and tier 2 differences for CO₂, all the counties have minimal differences compared to the Tier 1 results, with the largest difference of −0.17% in Grad Zagreb. The main reason for this is that CO₂ emission factors are the same for all technology levels, i.e., the emission factor is almost the same as the one for the Tier 1 method. The spatial distribution of total CO emissions and the spatial distribution for Tier 1 and Tier 2 differences for emissions CO are shown in Figure 6a,b, respectively. Osječko-baranjska county had the most significant total CO emissions, with 351 t, followed by Vukovarsko-srijemska county, with 215 t of CO₂ emissions. Ličko-senjska and Sisačko-moslavačka counties had the largest differences between Tier 1 and Tier 2 CO emissions, with differences of 51% and 50%, respectively. Vukovarsko-srijemska county had the lowest difference of 23%. It can be seen that Tier 1 and Tier 2 differences in CO emissions are pronounced with as much as 51% difference, compared to no differences in CO₂ emissions. Similarly, the differences are visible in all other emission types.

Figure 7a,b show the spatial distribution of total HC emissions and the spatial distribution for Tier 1 and Tier 2 differences for HC emissions, respectively. Osječko-baranjska and Vukovarsko-srijemska counties had the most significant total of HC emissions, with 124 t and 75 t, respectively. When considering Tier 1 and Tier 2 differences for HC emissions, Dubrovačko-neretvanska county had the highest difference of −62%, followed by Primorsko-goranska county with a difference of 57%. Much lower Tier 2 HC emissions compared to Tier 1 are mainly due to Dubrovačko-neretvanska county having a large proportion of newer agricultural tractors, i.e., technology level of Stage II or better.

The spatial distribution of total NO_X emissions and the spatial distribution for Tier 1 and Tier 2 differences for NO_X are shown in Figure 8a,b, respectively. Osječko-baranjska county had the most significant total of NO_X emissions, with 740 t. Contrary to CO₂, CO and HC emissions, NO_X emissions differences between Tier 1 and Tier 2 are negative for all counties, which indicates lower Tier 2 values across the whole of Croatia.

Figure 9a,b show the spatial distribution of total PM emissions and the spatial distribution for Tier 1 and Tier 2 differences for PM, respectively. Osječko-baranjska county had the highest total PM emissions, with 157 t. When considering Tier 1 and Tier 2 differences for PM emissions, Karlovačka and Sisačko-moslavačka counties had the largest difference of 125%. As with the NO_X emissions, these differences of more than 100% indicate a large difference in PM emission factors for different technology levels, which are shown in Table 1 and Table 2. Since most agricultural tractors in Croatia in 2016 were on the pre-Stage I technology level, it follows that the total PM emissions were significantly higher for Tier 2 compared to Tier 1.

The spatial distribution for all emission types shows that the continental part of Croatia has much higher emissions for all emission types compared to the coastal counties. The continental part of Croatia is the main region of agricultural activities in Croatia. This is consistent with the Tier 1 emission inventory developed for all agricultural machinery [57]. Tier 1 and Tier 2 differences have higher variations by emission type and by county. Different emission factors for Tier 1 and Tier 2 methods have the largest influence on total differences by emission type and county. A clear example of this can be seen in NO_X and PM emission factors in Table 1 and Table 2, where the differences between the same emission types for different technology levels vary greatly. If Stage IV and pre-Stage I emission factors are compared, there is a clear sharp reduction concerning all emission types. For example, the Stage IV emission limit presents a 98.3% reduction compared to the pre-1981 limit for PM in diesel engines.

Whereas the results show higher Tier 1 CO, HC and PM emissions compared to emissions calculated using the Tier 2 method, NO_X emissions provide an interesting example of how Tier 2 emissions could be higher if calculated using the Tier 2 method. This is mainly due to a large impact of newer emission reduction technology and a sharp reduction in emission limits in newer tractors. Since most of tractors in all counties in Croatia are equipped with pre-Stage I emission reduction technology, the Tier 1 method results in lower NO_X emissions compared to the Tier 2 method. This is mainly due to NO_X emission factors being drastically lower for newer agricultural tractors with emissions levels from Stage II, which can be observed in Table 1 and Table 2. A great example of the influence of emission reduction technology can also be seen in Dubrovačko-neretvanska county. When CO emissions are considered, this county had the highest difference of −62%, whereas other counties had a positive difference. As already mentioned, much lower Tier 2 HC emissions compared to Tier 1 are mainly due to Dubrovačko-neretvanska county having a large proportion of newer agricultural tractors, i.e., technology level of Stage II or better.

These examples show how data on vehicle age hold important information that can greatly influence the final emission inventory. This is why emission control technology and activity data are important information when researching agricultural emissions and the best policy solutions for emission reductions; further gathering of data would be necessary in order to conduct an analysis of this issue. In addition, various variables impacting the overall emissions inventory, e.g., more detailed activity data or data on terrain type, require further research in order to fully understand and contextualise existing results.

As there was no unified database necessary for developing a Tier 2 emission inventory, using two databases was a way to incorporate data on vehicle age without having to conduct a new large-scale survey, which would be costly and time consuming. Since scarcity of data for agricultural machinery is a problem in other countries, such as Germany [59] or Denmark [47], an approach that would enable combining different datasets would be beneficial to other countries and other NRMM sectors as well. In this way, more accurate emission inventories can be developed if the right databases and criteria for data inclusion are used. More accurate emission inventories offer a better understanding of the main emission sources, which enables decision makers to develop optimal emission reduction policies.

In Croatia, there is a comprehensive list of agricultural tractors, since all agricultural tractors must be registered by law. However, a list of agricultural tractors alone is not enough for developing a high-quality emission inventory or for the temporal tracking of emissions. As mentioned in the introduction, other countries, e.g., Switzerland, Finland, Germany, the Netherlands, Denmark, etc., also have a lack of data concerning agricultural machinery data, and their emission inventories contain many assumptions that raise uncertainty regarding the overall emission inventory. Since agricultural machinery represents a large proportion of overall NRMM emissions, it is necessary to improve the current data collection systems, especially when considering Europe’s aim to become climate-neutral by 2050. National statistical bureaus are key institutions due to their knowledge on conducting large-scale data collection research.

If possible, data collection should be structured in such a way that a designated institution has the infrastructure and the funds to conduct annual data gathering. This leads not only to accurate emission inventories for a given year, but, equally importantly, to temporal and spatial emission tracking. Combined with the tracking of policy instruments, it could provide for a more rapid emission reduction process. A basic but effective approach would be to encourage additional emission reduction technologies in new machinery or to retrofit older machinery with newer emission reduction technology, such as NO_X adsorbers, oxidation catalysts and diesel particle filters [30]. Another way of reducing emissions is by using hybrid, electric and hydrogen propulsions, which are compatible with renewable energy sources and can be indispensable in achieving climate neutrality in the agricultural sector [32,33].

4. Conclusions and Recommendations

Although agricultural tractors represent a large proportion of total agricultural machinery emissions, there is still lack of data and of a more structured approach towards their emission tracking and reduction. Data collection, handling and interpretation have a basic role in emission reduction. In this paper, a Tier 2 spatial emission inventory for agricultural tractors in Croatia was developed using two different databases. After determining the necessary data needed to develop a Tier 2 inventory, agricultural tractor emissions were calculated using the EEA Tier 2 method. Osječko-baranjska county had the most emissions for all the calculated emission types, followed by Vukovarsko-srijemska and Koprivničko-križevačka county. This is mainly due to the largest number of tractors and the largest fuel consumption being in these counties. Zadarska and Šibensko-kninska counties had the largest fuel consumption per tractor although their total fuel consumption amounted to only 3051 and 1331 tons of fuel, respectively. This is most likely due to terrain configuration, as this part of Croatia has rocky terrain with steep hills.

Observed differences between Tier 1 and Tier 2 emission inventories vary by county and by emission type. For CO₂ emissions, there were no differences due to Tier 1 and Tier 2 having the same emission factors for all tractor equipment levels. Differences for CO and NO_X emissions varied from 23% to 51% for CO emissions and from −18% to −1% for NO_X emissions. The negative percentages for NO_X emissions, meaning that emissions calculated using the Tier 2 method are larger compared to those calculated using the Tier 1 method, are due to large differences in NO_X emissions factors in different technology levels. The largest emission differences were found for HC and PM emissions, where differences varied from −62% to 57% for HC emissions and from 37% to 125% for PM emissions. As previously stated in Section 3.2, these differences are present mainly due to a combination of a different age profile of agricultural tractors for different counties and a major difference between the emission factors for the same emission type.

Without a designated institution responsible for data collection and emission tracking, it is hardly possible to ensure continuity of results. Knowing the emission reduction or increase trend could be more important than knowing the amount of emissions there were in a given year, because only temporal tracking of emissions can be a basis for emission reduction policy evaluation. If possible, data that would enable a year-by-year Tier 3 inventory, which takes into account load factors, hours of work, agricultural equipment type, etc., should be collected. Annual questionnaires that would target key agricultural stakeholders could be enough to develop such an inventory. Similar action should be taken into account in other NRMM sectors, such as in construction, inland water or household sectors.