基于多源数据回归分析的糖料蔗产量估计

Estimation of Sugarcane Yield Based on Regression Analysis of Multi-source Data

糖料蔗生产已成为广西经济的主要来源,其种植面积达到全国60%以上。然而,广西糖料蔗区大多分布于山丘坡地,下垫面非均质性和管理非均匀性造成了甘蔗产量具有较强的空间变异性。利用模型估计产量需要大量的观测数据,且运行成本较高。利用无人机等可以较为方便地得到甘蔗不同时期叶面积指数和株高等信息,地面的传感器也可以采集不同深度的土壤含水率。分析了广西崇左60个试验小区2016年采集的叶面积指数、株高、土壤水数据和糖料蔗产量的相关性,发现相比于地面的叶面积指数观测和土壤水分观测,利用无人机平台所提取的株高观测更有利于糖料蔗的产量估计;叶面积指数观测在伸长中期阶段,数据价值最大,而株高观测则在伸长末期与产量的相关性最高;播种后1-2个月根系区土壤水分对甘蔗最终产量的形成起到了至关重要的作用。采用多源数据回归分析的方法对甘蔗产量进行估计,相关性R^2=0.87,是一种较为便捷的甘蔗产量估计方法。

Sugarcane has been the main source of Guangxi economics and its cultivation area reaches more than 60%of the total area of sugarcane in China.However,the cultivated sugarcane is mostly located on the hillslope,and the heterogeneous soil texture,slope,and nonuniform management lead to the spatial variability of sugarcane yield,so it is difficult to estimate by using crop modeling since it requires too much input data.The information of leaf area index and plant height of sugarcane in different periods can be easily obtained by unmanned aerial vehicle(UAV),and the soil moisture at different depths can be obtained via ground-based sensors.In this paper,the correlation between plant height,leaf area index,soil moisture,and sugarcane yield data collected in 2016 from 60 experimental plots at Chongzuo,Guangxi,was analyzed.It was found that:Crop height data from UAV observations were more helpful than other observations;The leaf area index data collected at the middle stage of elongation and crop height at the end of elongation were the most informative;Root-zone soil moisture at one month after sowing was highly correlated with crop yield.By using the multi-variable linear regression analysis to estimate the sugarcane yield,the correlation index R2 can reach 0.87,which indicated that the multi-variable linear regression analysis is a relatively convenient method to estimate sugarcane yield.