TEMPORAL SIMILARITY OF MECHANIZATION LEVEL AND CROP YIELD IN ERITREA

全文:

Background:

Agriculture is a system that necessitates a multitude of diverse input resources for successful and efficient production, in addition to the must-available cultivation environments, including power sources (humans, animals, and/or system of machines) (1). The dynamics of agricultural production have marched towards a highly mechanized state, aiming to produce more with less (2,3). The extent to which the agricultural sector uses mechanical power could be a measure of the level of mechanization (LOM). The amount of horse power (hp) per thousand hectares of cultivated land is one of the common indices of LOM (3,4). 

This metric fluctuates over years, especially in low-income countries, indicating its response to prevailing socioeconomic, governmental, and other factors. Despite the technological advancements that could offer machines that are cost-effective and efficient, agricultural sectors can be sensitive, in periods of economic fluctuations (growth or recession), to invest in mechanization (5–8). The pattern of the progress of the LOM could be periodic, influenced by a complex interplay of socioeconomic, environmental, and policy factors that can fluctuate over time (9), which could lead to corresponding variations in the production and productivity of an agricultural system. This can be handled as a time series problem, and ordinary least square (OLS) and the dynamic time warping (DTW) methods can be applied to evaluate the similarities in pattern of the fluctuations between the two sequences.

OLS is a versatile technique applied in a wide range of disciplines for establishing statistical relations between parameters. According to Sharma et al. (2011), Larrabee et al. (2014), and Sharma et al. (2013), the OLS was used in comprehending the relationships between crop yields and influencing factors such as average precipitation (10), soil properties, management practices (11)  and product quality and influencing factors (11). OLS can also be effectively applied for economic factor effect analysis (12), quantify the effect of farm machinery, on crop yield, identify the determinants of adoption of farm machinery, examines connection between machinery utilization and input resources, and assesses farm machinery investment economic feasibility and profitability. Enhances input management tactics and assesses the combined impact of equipment adoption and off-farm employment on farm performance(13). However, it fails to precisely analyze the temporal patterns of timeseries datasets.

The DTW can particularly be used in comparing sequences of similar patterns that might differ in timing, alignment, or length and finds an optimal alignment between them (14–17). In 1978, Sakoe used the DTW optimization algorithm for spoken words recognition (16). Automatic speech-to-lip alignment (18,19); finger print matching validation (20); and detecting faces and eyes ; comparison, alignment, and combining time series sequences living in comparable (17,19) or incomparable space, in multivariate time series classification (21); similarity analysis between two sequences of datasets that have temporal variability with or without shift, scaling, and global invariance (15); data mining for determining averages and indexing (22) can be performed with the DTW algorithm.

In agriculture, DTW has applications in measuring similarities between two temporal sequences to analyze in the classification of areas based on land use and land cover and making maps (23), and the weighted derivative modification of DTW performed better in crops mapping (24–26); and in determining the spatial and temporal characteristics of agricultural non-point source pollution loads and identifying the dominant processes and factors responsible.

It is widely recognized that the LOM greatly affects production and productivity (8). In Eritrean agriculture, particularly in cereal production, the fluctuating pattern associated with the insignificant growth trend over the years could potentially be caused by the modest growth and fluctuating nature of the LOM. To study the effect of the LOM on production, similarity analysis of time series records of the LOM and yield can be performed. However, there is insufficient research analyzing the similarity between time series datasets of cereal production and the LOM, both globally and in Eritrea, specifically using DTW. The Eritrean cereal production shows a fluctuating pattern over years with mild growth, and so does the LOM. Nevertheless, there is no scientific evidence that shows the relationship, effect, dependence, or similarity in pattern between these datasets.

Aim of the Study: This study aims to analyze the similarity between cereal production (yield) and the Level of Mechanization (LOM) in Eritrean agriculture. The objective is to assess the impact of LOM on crop yield by employing Ordinary Least Squares (OLS) regression for trend analysis and Dynamic Time Warping (DTW) for pattern similarity analysis. This analysis seeks to understand the temporal pattern similarity between yearly crop yields and changes in LOM, as well as to evaluate the influence of LOM on agricultural productivity.

MATERIALS AND METHODS

Assessment of Level of Mechanization (LOM) 

Power availability per unit area, kW/ha (hp/ha), as a measure of LOM, is given by equation 1 (4).

                                                               [1]

Where: P_i –Machinery power (kW or hp); and L_i –Total cultivated area (ha).

According to the USDA, the LOM of the countries in the world, from 1961 to 2020, is computed and made available online in an interactive mode (27), enabling inspection of the LOM for each year of the range. A freely available CSV file containing diverse, hefty data, a section of which is the total horse power (in multiples of 1000 hp) of farm machinery in a respective country for the years, is also made available. This data was utilized as a source for determining the LOM in this research, and the corresponding year's cultivated area (in multiples of 1000 ha) was received from the Eritrean ministry of agriculture.

Similarity Analysis

A comprehensive multi-step analysis was conducted to analyze the similarity between the LOM and yield. 1) preliminary time-series trend similarity and statistical analyses (the Ordinary Least Squares (OLS) regression with an interaction term); and (2) a subsequent, more detailed analysis using DTW for robust similarity metrics. This approach offers a nuanced understanding of the temporal relationship between LOM and yield. The entire analysis was conducted in a Python environment, employing appropriate libraries and packages. Since the datasets were at different scales, the yield dataset was standardized and scaled simultaneously to match the LOM using equation 2 (28):

                                                  [2]

Where: Y_s –standardized yield;  –mean of the yield; σ_Y –yield standard deviation; σ_x –LOM standard deviation;  –LOM mean.

Ordinary Least Squares (OLS) regression

The dependent variable and independent variables were modeled using OLS regression with an interaction term as: Interaction = Year_Centered× (Standardized Yield not NA) and combine the 'LOM' and 'Standardized Yield' using combine first and configure the regression model as (29):

                                              [3]

Dynamic Time Warping Distance Similarity Analysis

DTW aligns two datasets using dynamic programming, estimating local costs. Given two multivariate time series sequences, the standard form of DTW (equation 4) of time series  and , where T_x and

作者简介

Tesfit A.

编辑信件的主要联系方式.
Email: noahtesas@gmail.com
ORCID iD: 0009-0002-4371-1323
厄立特里亚

Alexander G.

Email: alevshin@rgau-msha.ru
ORCID iD: 0000-0001-8010-4448

Simon G.

Email: gtsimon1994@gmail.com
ORCID iD: 0009-0002-3336-0523

参考

补充文件