Contribution of agriculture, forestry, and other land use to climatic phenomena through GHG emissions in Burundi. Empirical evidence from the ARDL approach

In a world where we live in a climatic hazard disaster, understanding human activities' impact on the environment becomes crucial, especially for developing countries. This study addresses this issue through the prism of carbon dioxide emissions (CO2) by examining how the growth of agricultural, forestry, and other land-use activities affects the environment. This study examines the long-term relationship between CO2 emissions and agriculture, forestry, and other land use by introducing other explanatory variables such as income per capita and population for 1970-2016. The econometric analysis results confirm a positive long-term relationship between CO2 emissions and agriculture, forestry, and other land use. The Granger causality tests show that this relationship's direction goes from agriculture, forestry, and other land use to CO2 emissions. For a developing country like Burundi, it is the growth of agricultural, forestry, and other land-use activities that determine GHG emissions level and not the other way around. Moreover, the ARDL model results adopted in this study validate our hypothesis, which assumes that Burundi has not yet reached its turning point in CO2 emissions due to economic growth. A low-carbon policy in favor of the


Introduction
Over the centuries, humanity has constantly faced the rise in earth temperature challenges in the short and long term. Our society, which aims for efficiency, productivity, and optimization, has no more control than expected over the climate's impacts [1]. This efficient, productive, and optimized society can be significantly disrupted by natural climate variability. Nevertheless, this activity contributes to aggravate its vulnerability by adding massively to the atmosphere of greenhouse gases which have already started to disturb significantly a climate that has been in equilibrium for more than 10,000 years [2]. During these 10,000 years, the amount of GHGs present in our atmosphere was relatively constant and allowed the Earth to maintain a relatively stable climate ( Figure 1). These gases' concentration began to rise with the advent of industrialization, increased demand for energy, population growth, and land-use changes [3]. Scientists and researchers have confirmed human activities' role in these changes [4,5,6]. Humanity has increased its needs by continuing natural resources [7]. These activities' induced effects accentuate the natural greenhouse effect and raise the Earth's surface's average temperature [8]. The weather rose to a higher level in 2020, about 14.9 0 C,1.2 0 C, above the pre-industrial (1850-1900) level ( Figure 2). This phenomenon, known as global warming, affects rainfall and other atmospheric and climatic phenomena, 4 endangering human, animal, and plant life due to the potentially catastrophic consequences of climate change [9,10]. Figure2: Annual mean surface air temperature, NASA 2020 Preparations are being made to combat these inevitable effects, and new means are being explored to reduce the gas emissions that are the cause. Zhang and Dong [11] have affirmed that a cleaner green technology and changing consumer habits would reduce harmful emissions and stem global warming. It is too late to avoid some severe climate change incidents in some developing countries [12]. Known as the least developed continent globally, Africa appears to be one of the most vulnerable continents to climate change.
However, it is less responsible for the ozone layer's degradation than other continents [13]. Besada and Werner [14] have demonstrated that a third of the continent's population already lives in areas of drought.
Further, Connolly-Boutin and Smit [15]' research affirmed that an additional 75 to 250 million people could lose their livelihoods due to climate change by the end of the next decade.
One of the consequences that we think of first when we talk about Africa's global warming is the rise in sea levels, with its harmful effects on low-lying coastal areas (floods, vulnerability to storms) [16]. Climate changes such as rising temperatures, reduced water supplies, loss of biodiversity, and ecosystems' degradation have manifested agriculture effects. Current projections suggest that agricultural production's potential will decline in parts of Africa subject to severe climate crises [17]. Warming of around two degrees 5 Celsius would lead to a 10% reduction in total crop yield in sub-Saharan Africa by 2050 [18]. It will pose a new threat to the roughly 65% of African workers whose livelihoods depend on agriculture [19].
Agricultural activities, like other sectors, are the source of air pollutant emissions[ Figure 3, [20]. The farm sector mainly contributes to releases of ammonia and greenhouse gases, which impact the climate [21]. The GHG emissions in Africa are caused by deforestation, soil impoverishment, and erosion, which, in addition to undermining traditional agriculture, releases significant amounts of carbon [22]. Imasiku et al., (2020) IPCC source sector GHG, The World various African countries' adaptation capacities [23]. The continent remains vulnerable to climatic variability because of its multiple political stresses it knew and its ability to respond that remains weak [24].
The Horn of Africa's arid coasts to East Africa's rainforests are among the most vulnerable of the continent, where rainfall is 200% or more than usual during the rainy season [25]. Heavy rains recorded have caused floods, landslides, and damage to infrastructure and livelihoods in Burundi, Kenya, Ethiopia, Rwanda, Somalia, and Tanzania. Burundi, the water levels of Lake Tanganyika are the highest since 1964(latawama.org).

A brief history of AFOLU in Burundi
The relationship between agricultural practices and environmental conditions has raised great interest as long as it is the primary source of life for humans [26]. According to the data compiled by moody's analytics data, Burundi's agricultural sector employs more than 90% of the working population and contributes more than 40% of the GDP. It provides the country with more than 90% of foreign exchange earnings and provides 95% of food intake. Nevertheless, this sector is characterized by subsistence agriculture essentially dominated by food crops that occupy 90% of the land cultivated by about 1.2 million rural families, each less than one hectare of land and whose production is mainly self-consumed by households. Livestock, mainly of the extensive type, enter the gross domestic product for only 5% and continuously decline because of the decrease in grazing areas. The natural and artificial forests occupy about 200,000 hectares. They are devoted to protecting the land heritage against erosion, regulating hydrological systems, wildlife conservation, and the provision of firewood, construction, and timber.

AFOLU and GHG emissions
Taking agricultural practices in the greenhouse gas balance faces uncertainty sources. Pradhan and Limmeechokchai [27] reported that the GHG emission from Thailand agricultural sector in 2015 was estimated to be 45.3 MtCO2eq and projected to reach 51.2 MtCO2eq in 2030 and 63.6 MtCO2eq in 2050.
These agricultural GHGs emissions originate from soil management, enteric fermentation, energy use, and manure management [28]. In descending order of magnitude, methane (CH4), nitrous oxide(N2O), and carbon dioxide(CO2) are the primary GHGs related to AFOLU [20]. Almaraz et al. [29] have shown that biogenic emissions of NOx from agricultural areas are much higher than we used to believe. Natural processes produce this gas, but its production is enhanced by the leaching, volatilization, and runoff of nitrogen fertilizers and the decomposition of crop residues and animal wastes [30]. In the Auvergne-Rhône-Alpes region, in 2015, the agricultural sector was the source of primary fine particle emissions 6,100 tons, 37% of total particles in suspension, and 6% of carbon dioxide emissions of nitrogen NOx (Cadastre of Atmo -Auvergne-Rhône-Alpes emissions 2017 version).
The investigation of HUANG et al. [31] on NH3 emissions from agriculture in Jiangsu Province, which used the emissions factor method, the NH3 emissions from agriculture were mainly contributed by livestock and nitrogen fertilizer application. Livestock produces about 40 percent of the world's gas emissions, mineral fertilizers 16 percent, and biomass combustion and crop residues about 18 percent [32,33].
Ammonia is one of the leading causes of acid rain, which damages trees, acidifies soils, lakes, and rivers, and damages biodiversity [34]. Further, Xu et al. [35] have incorporated the Bi-NH3 module into the process-based Dynamic Land Ecosystem Model to estimate NH3 emission from agricultural systems in SA.
The NH3 is expected to increase due to the increase in population needs.
YUE and GAO [36] have compared the global annual GHG emissions from natural systems with those generated by human activity. Humans were accounted for about 90 percent of biomass combustion, mostly from deliberately ignited forest fires as part of the clearing and burning of pastures. Nevertheless, burning biomass to clear land for agriculture is also a common practice that produces CO2 and other local pollutants [36].
According to Singh and Singh [38], traditional agriculture threatens the environment and releases nitrous oxide, phosphorus, and heavy metals into the atmosphere. The conversion of peatlands to fields, deforestation, fertilizers, cultivation, and animal husbandry produce large amounts of emissions that affect the climate. The expansion of agriculture and animal husbandry has been identified as one of the leading causes of emissions from deforestation [38]. In agriculture, deforestation is carried out directly by transforming forests into agricultural areas [39]. Once the forest cover is removed, GHG is released into the atmosphere, and lands are exposed to erosion, loss of minerals and nutrients [40].
Even though the increasingly recognized importance of emissions from agriculture, forestry, and other land uses in Africa in the global carbon cycle, there is a lack of empirical study on GHG emissions of this sector.
Burundi carries a risk of hurting its environment by exhausting goods supplied by AFOLU if there is a causal link between environment degradation by GHG emissions and AFOLU. It is argued that AFOLU activities lead to CO2 emission sources primarily from agriculture [41].
This paper investigates, therefore, whether there is such a relationship in Burundi. A study on GHGs sources is vital for the definition of public policy instruments. If the links between emissions and sinks are crucial at a given geographic scale, an effective policy will likely cover all sources and sinks to limit unknown effects on the climate. In the case of Burundi, CO2 emissions are poorly documented. This article's contribution considers all the sources of agricultural emissions and the sources of CO2 linked to climatic phenomena.

Econometric specifications and estimation methodology
This paper follows the methodology of studies carried out on the relationships between human activities and environmental pollution, which goes through what has come to be called 'the EHRLICH equation: Where e, the random term (possibly including other explanatory variables); b, c, d the ecological elasticity values, respectively for P, A, and T.
This model is tested according to a log-linear equation However, more recent work has given stochastic content to this equation, replacing T with a random term, allowing interesting statistical treatments. In addition to considering component T as an arbitrary term, it is possible to incorporate into the equation other variables supposed to explain the environmental impact.
By going through this modeling, our study attempts to clarify the factors responsible for environmental impacts (particularly CO2 emissions) by integrating AFOLU, population, and per capita income as explanatory variables to test the long-term relationship known as cointegration between CO2 emissions.
The following logarithmic form is proposed: the appropriate offsets will correct problems of correlation and endogeneity of the series. Another reason for using the ARDL approach is that it is more robust and performs better for small sample sizes than other cointegration techniques.
Our model Equation is transformed as follows: Where, 0 is the drift component, and the white noise. The terms with the summation signs represent the error correction model, while those with the the coefficient represents the long-term relationship.
The verification of the cointegration relation is carried out on the Bounds Test. It consists of carrying out an F-test on the hypothesis 1 = 2 = 3 = 4 = 0 against the alternative hypothesis 1 ≠ 2 ≠ 3 ≠ 4 ≠ 0.
To choose an optimal delay for each variable, the ARDL method estimates The data used in this study are taken from the World Development Indicators database of the World Bank

Empirical Results and Discussion
This study uses the Augmented Dickey-Fuller (ADF) and Phillips Perron (PP) stationarity tests to determine the series's integration order. Beforehand, it is necessary to ensure that no string is integrated of order two because the critical values provided by Pesaran et al. (2001) only concern integration levels of 0 and 1. The results of these tests are reported in Table 1. They indicate that all series are not stationary in level but stationary in the first difference. They are therefore integrated of order 1. AIC, SC, LR, and HQ criteria are then used to select the optimal delay number of the vector autoregressive (VAR). The AIC criterion implies a delay of 5, while the LR, SC, and HQ criteria imply an uncertainty of 2. It is this last criterion that is retained for this study.  To determine the long-term relationship between variables, we use ARDL or Black Box approach for cointegration. For this, the Bound Test is used, which calculates an F-statistic (Table 3). It tests the null hypothesis that the lagged variables' coefficients in equation (4) are zero. The F-statistic is equal to 6.843977 compared to the critical values below and above the threshold significance of 5% and 1%. The test statistic is above the upper limit (4.35 and 5.61, respectively). Therefore, we reject the null hypothesis of the absence of cointegration. We conclude the existence of a long-term relationship between the model's variables. The Granger causality test is performed on the variables to determine this relationship's direction (Table 4). The Granger causality test (Table 4) shows a causal relationship between the two main variables of interest (AFOLU and CO2) and that this relationship is in the direction of AFOLU towards CO2 emissions. In other senses, agriculture, forestry, and other land-use granger cause CO2emissions but not vice-versa. The same result is also valid for the GDPC and POP variables; the evolution of CO2 emissions depends on GDP per capita and POP. A low-carbon policy in favor of the environment can be implemented, and this cannot risk any negative repercussions on economic growth.  The coefficient signs of LGDPC and LPOP variables are negative and significant( The error correction model (ECM) mechanism is used to test the short-run relationship between variables ( Table 6). The results show that the error-corrected term ECM(-1) coefficient is significant, which implies that the speed of short-run adjustment to reach equilibrium is substantial. On the other hand, this term equals -0.689295 suggests that when CO2 emissions are above or below their equilibrium value, they would be adjusted by 68% per year. The coefficients of the lagged variables represent the short-run elasticities.
Furthermore, a 1% increase in agriculture, forestry, and other land use would imply a 0.79% increase in CO2 emissions in the short term. We carried out diagnostic tests on residuals of the ARDL regression ( Table 7). The White test confirms the absence of the residuals' heteroskedasticity, while the Jarque-Bera test shows that they follow a normal distribution. On the other hand, Ramsey's test shows no missing variables or functional form issues in the model. repercussions on a country's economic growth.
The CUSUM and CUSUMQ techniques show that the coefficients of the cointegrated ECM model are stable. The results also indicate that agriculture, forestry, and other land use remain significant determinants of CO2 emissions in the short run and the income per capita and population. However, their coefficients are negative, also remain substantial determinants.
In Burundi, populations, as well as the government, wish to develop protected areas. Nevertheless, this is not enough. In rural areas, people engage in systematic plundering of natural resources. Consequences are manifested by runoff to urban areas, and the health of Tanganyika Lake is soiled by pollution.
These results are encouraging, given that Burundi has a REDD + program to reduce greenhouse gas emissions linked to deforestation and forest degradation. Projects to support local communities in reforestation and sustainable agriculture could be established by involving local populations in the ecosystem's preservation and resource exploitation.

Availability of data and materials
The datasets of AFOLU analyzed during the current study are not publicly available due the sharing permission not approved by the Burundian Institute for Statistics and Economic Studies (BISES) but are available from the corresponding author on reasonable request and with permission of BISES.