INTRODUCTION Predicting Greenhouse Gas Emission of SRI Paddy Fields under Different Soil Conditions using Artificial Neural Networks.

Sustainable Water and Environmental Management in Monsoon Asia 30-31 October 2014, Kaohsiung, Taiwan Predicting Greenhouse Gas Emissions of SRI Paddy Fields under Different Soil Conditions using Artificial Neural Networks Chusnul Arif 1 , Budi Indra Setiawan 1 , Yudi Chadirin 1 , I Wayan Budiasa 2 , Masaru Mizoguchi 3 , Junpei Kubota 4 , Hisaaki Kato 4 1 Department of Civil and Environmental Engineering, Bogor Agricultural University, Bogor, Indonesia 2 Faculty of Agriculture, Udayana University, Bali, Indonesia 3 Department of Global Agricultural Sciences, the University of Tokyo, Japan 4 Research Institute for Humanity and Nature, Kyoto, Japan ABSTRACT Conventional paddy field is a major source of greenhouse gas emissions particularly methane and Nitrous Oxide. Increasing CH 4 and N 2 O concentrations in the atmosphere contributes to global warming. However, it is not easy measured in the fields particularly in Indonesia when the instrumentation is limited. The current study proposes the model to predict CH 4 and N 2 O emissions using artificial neural network ANN with easily measurable inputs such as soil moisture, soil temperature and soil electrical conductivity. To verify the model, two experiments were conducted in the pot and paddy field. The pot experiment was conducted in the greenhouse of Meiji University, Kanagawa Prefecture, Japan from 4 June to 21 September 2012, while the paddy field experiment was conducted in Umejero village, Buleleng district, Bali, Indonesia during first rice season in 2014. In the pot experiment, three irrigation regimes called wet, medium and dry regimes were applied for different pots, while in the paddy field experiment, there were two rice cultivation practices, i.e., conventional and system of rice intensification SRI with three different irrigation regimes, i.e., continuous flooding for conventional cultivation, wet and dry regimes for SRI management. The results showed that CH 4 and N 2 O emissions were fluctuated with different soil conditions. Better prediction with higher R 2 was obtained in the paddy field experiment. Since the field experiments were conducted under certain soil and climatic conditions, so they cannot necessarily be generalized. In addition, more field measurements are needed that represented any soil and climatic conditions to rich the observed data, so the model can be trained well under wider interval of soil and climatic conditions. Keywords: Greenhouse gas emissions, artificial neural networks, paddy fields, soil moisture, soil temperature, soil electrical conductivity

I. INTRODUCTION

Water management of rice production influences the dynamic changes of soil moisture and temperature that implied on enhancingreducing soil microbial activities in which their activities producing greenhouse gases emitted into the atmosphere. There are three main Sustainable Water and Environmental Management in Monsoon Asia 30-31 October 2014, Kaohsiung, Taiwan gases i.e. CO 2 , CH 4 and N 2 O, which are commonly emitted from paddy fields. However, only two of them, CH 4 and N 2 O, are attracted considerable attention during the last decades because of their contribution to global warming Bouwman, 1990; Neue et al., 1990. Methane CH 4 and nitrous oxide N 2 O gases have potential contributing to global warming at 23 and 296 times greater than carbon dioxide, respectively Snyder et al. 2007. Many research findings have been published regarding CH 4 and N 2 O emission from paddy fields over the past 25 years both under conventional rice farming with continuous flooding irrigation and alternative rice cultivation such as System of Rice Intensification SRI with intermittent or non-flooded irrigations e.g. Akiyama et al., 2005; Cai et al., 1997; Dong et al., 2011; Husin et al., 1995; Keiser et al., 2002; Li et al., 2011; Minamikawa and Sakai, 2005; Nugroho et al., 1994; Setiawan et al., 2014; Setiawan et al., 2013; Towprayoon et al., 2005; Tyagi et al., 2010; Zou et al., 2005. There is clear findings that CH 4 emission enhance when anaerobic soil condition is developed under flooded water, conversely, N 2 O emission dramatically increase under aerobic condition with non-flooded water in the fields. Recent studies showed that SRI paddy field with the intermittent wetting-drying irrigation reduced CH 4 emission up to 32 Rajkishore, et.al, 2013, but N 2 O emission increased by an insignicantly 1.5 Dill, et al., 2013. Hence, SRI paddy fields can used as mitigation option for rice production. CH 4 gas is produced by methanogens during organic matter decomposition, under an environment where the oxygen and sulfate are scare. Meanwhile, N 2 O gas is primarily produced from aerobic microbial processes, nitrification and denitrification in soil Mosier et al., 1996. In fact, CH 4 and N 2 O emissions are not only influenced by water availability soil moisture in the fields but also rice varieties Husin et al., 1995; Setyanto et al., 2004 and fertilizer applications Cai et al.,1997; Nishimura et al., 2004. Different application of water management and fertilizer affected on soil parameters level such as soil moisture, temperature, pH, redox potential Eh and electrical conductivity EC varies at particular time. CH 4 flux varies diurnally with its maximum value occurring in the afternoon when soil temperature reach peak value Miyata et al., 2000; Purkait et al., 2007. Higher soil pH was also observed releasing higher CH 4 emission Babu et al., 2005, but its values reduces as the soil Eh becomes more negative Lee et al., 2005; Setyanto and Bakar, 2005; Tyagi et al., 2010. Therefore, those parameters are associated with dynamic change of CH 4 and N 2 O Sustainable Water and Environmental Management in Monsoon Asia 30-31 October 2014, Kaohsiung, Taiwan emissions. However, the relationship between those parameters and CH 4 and N 2 O emissions are very complex and it’s difficult to characterize into deterministic mathematical model. Commonly, CH 4 and N 2 O emissions are measured manually using closed chamber that is placed over singlesome paddies rice. Then, the gas sample is taken from the chamber periodically and the gas sampling is analyzed using a gas chromatograph in the lab. However, even if this method given the results accurately, the method is time consuming and complicated with more expensive equipments. Setiawan et al. 2013 have developed interrelationship model between soil moisture volumetric water content, soil pH and soil temperature with CH 4 and N 2 O emissions using an artificial neural network ANN model, that producing accuracy with an R 2 about 0.70. However, those three soil parameters were measured separately with different sensors. In addition, one of those parameters, soil pH, is usually measured discontinuously by handheld sensor. The current study propose ANN model to estimate CH 4 and N 2 O emissions with easily measurable inputs i.e. soil moisture, soil temperature and soil EC that are measured by single sensor. The specific objectives of this study were 1 to identify different water irrigation regimes on CH 4 and N 2 O emissions with SRI management, 2 to develop the model in estimating CH 4 and N 2 O emissions using ANN model, 3 to characterize greenhouse gas emissions under different soil parameters.

II. MATERIALS AND METHODS 2.1 Field Experiments