Scholarlink

Purpose: Agriculture sector produces a notable amount of greenhouse gas emission (GHG), which contributes significantly to climate variation and global warming. To deal with the consequences of climate variation on agriculture, vigilant management of resources such as soil, water, and management practices will be required. This study aims to find the influencing soil and climatic parameters for CO2 and CH4 emission from the agriculture sector with two different cultivation patterns. The study also aims to understand that weather the agroclimatic zone plays a role in the emission of CO2 and CH4. Design/Methodology/Approach: Implemented lasso and ridge regression for forecasting the CO2 and CH4 emission from different cultivation farm with regularization which avoids the problem of multicollinearity. We used IoT sensors MQ-4 and MQ-135 for measuring the CO2 and CH4 emission, Agrinex solution for NPK and pH values and Google weather app for climate parameters. The Raspberry Pi will read the data from sensors and send it to ThingSpeak cloud server for storing. Findings: For predicting CO2 and CH4 emission the RMSE value for lasso and ridge are 0.031695, 0.034699, 0.00521 and 0.000528 respectively. Lasso has the lowest RMSE value as compared to ridge which indicates a good prediction model. We compare the emission for different cultivation pattern CO2 and CH4. The CO2 and CH4 emission from farm land and protected farm land is 45.9204 (PPM) and 37.5767 (PPM), 202.1688 (PPM) and 182.8158 (PPM) respectively. The result shows that agro-climatic zone affect the emission of CO2 and CH4. Nitrogen, Moisture, Pressure, Humidity and Temperature are most influencing parameter for CO2 and CH4 emission in farm land and protected farm land. Research Limitations/Implications: N2O emission from the agriculture is not considered in the study. Originality/Value: When the dataset contains multicollinearity (two or more variables are closely correlated), the ordinary least square (OLS) approach of linear regression is ineffective. However, there might be a lot of variation. The usual approach to this problem is to reduce variation at the expense of bias. This method is known as regularization which avoids the problem of multicollinearity to gain the performance of the model. Multicollinearity parameters can adversely influence model forecasting on testing data. Several regularization techniques can detect and fix multicollinearity. In our study we used ridge and lasso regression to solve the issue of multicollinearity.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting Influencing Parameter for CO2 and CH4 Emission from Agriculture Using Lasso and Ridge Regression

  • 章节

  • 发表日期：

    2026年

查看全文
  • Pranali Kosamkar,
  • Vrushali Kulkarni

摘要

Purpose: Agriculture sector produces a notable amount of greenhouse gas emission (GHG), which contributes significantly to climate variation and global warming. To deal with the consequences of climate variation on agriculture, vigilant management of resources such as soil, water, and management practices will be required. This study aims to find the influencing soil and climatic parameters for CO2 and CH4 emission from the agriculture sector with two different cultivation patterns. The study also aims to understand that weather the agroclimatic zone plays a role in the emission of CO2 and CH4. Design/Methodology/Approach: Implemented lasso and ridge regression for forecasting the CO2 and CH4 emission from different cultivation farm with regularization which avoids the problem of multicollinearity. We used IoT sensors MQ-4 and MQ-135 for measuring the CO2 and CH4 emission, Agrinex solution for NPK and pH values and Google weather app for climate parameters. The Raspberry Pi will read the data from sensors and send it to ThingSpeak cloud server for storing. Findings: For predicting CO2 and CH4 emission the RMSE value for lasso and ridge are 0.031695, 0.034699, 0.00521 and 0.000528 respectively. Lasso has the lowest RMSE value as compared to ridge which indicates a good prediction model. We compare the emission for different cultivation pattern CO2 and CH4. The CO2 and CH4 emission from farm land and protected farm land is 45.9204 (PPM) and 37.5767 (PPM), 202.1688 (PPM) and 182.8158 (PPM) respectively. The result shows that agro-climatic zone affect the emission of CO2 and CH4. Nitrogen, Moisture, Pressure, Humidity and Temperature are most influencing parameter for CO2 and CH4 emission in farm land and protected farm land. Research Limitations/Implications: N2O emission from the agriculture is not considered in the study. Originality/Value: When the dataset contains multicollinearity (two or more variables are closely correlated), the ordinary least square (OLS) approach of linear regression is ineffective. However, there might be a lot of variation. The usual approach to this problem is to reduce variation at the expense of bias. This method is known as regularization which avoids the problem of multicollinearity to gain the performance of the model. Multicollinearity parameters can adversely influence model forecasting on testing data. Several regularization techniques can detect and fix multicollinearity. In our study we used ridge and lasso regression to solve the issue of multicollinearity.