农业工程学报
農業工程學報
농업공정학보
2015年
5期
188-193
,共6页
张智韬%兰玉彬%郑永军%陈立平%宋鹏
張智韜%蘭玉彬%鄭永軍%陳立平%宋鵬
장지도%란옥빈%정영군%진립평%송붕
植被%气象%回归分析%NDVI%大豆%岭回归
植被%氣象%迴歸分析%NDVI%大豆%嶺迴歸
식피%기상%회귀분석%NDVI%대두%령회귀
vegetation%meteorology%regression analysis%NDVI%soybean%ridge regression
针对太阳辐射、大气温度、空气湿度和风速等气象因素对大豆归一化植被指数(normalized difference vegetation index, NDVI)在每天不同时间的影响,提高大豆NDVI的监测精度。该研究采用GreenSeeker手持式光谱仪对大豆苗期、花荚期和成熟期3个主要生育阶段的NDVI值以小时为单位进行连续监测,并收集测量时的太阳辐射、大气温度、空气湿度和风速等气象数据,采用偏最小二乘法、逐步回归和岭回归方法,建立不同气象因素对大豆 NDVI 值影响的回归模型,并分析其定量关系。结果表明,影响大豆不同生育期 NDVI 变化的主要气象因素为太阳辐射和大气温度,风速和空气湿度的影响较小,可以忽略不计。经对3种模型进行预测精度评价后得出,岭回归模型的预测精度最佳,其在3个阶段的预测均方根误差(RMSE)分别为0.034、0.018和0.016,决定系数(R2)分别为0.820、0.908和0.934,其次为逐步回归法,偏最小二乘法的预测精度最低。
針對太暘輻射、大氣溫度、空氣濕度和風速等氣象因素對大豆歸一化植被指數(normalized difference vegetation index, NDVI)在每天不同時間的影響,提高大豆NDVI的鑑測精度。該研究採用GreenSeeker手持式光譜儀對大豆苗期、花莢期和成熟期3箇主要生育階段的NDVI值以小時為單位進行連續鑑測,併收集測量時的太暘輻射、大氣溫度、空氣濕度和風速等氣象數據,採用偏最小二乘法、逐步迴歸和嶺迴歸方法,建立不同氣象因素對大豆 NDVI 值影響的迴歸模型,併分析其定量關繫。結果錶明,影響大豆不同生育期 NDVI 變化的主要氣象因素為太暘輻射和大氣溫度,風速和空氣濕度的影響較小,可以忽略不計。經對3種模型進行預測精度評價後得齣,嶺迴歸模型的預測精度最佳,其在3箇階段的預測均方根誤差(RMSE)分彆為0.034、0.018和0.016,決定繫數(R2)分彆為0.820、0.908和0.934,其次為逐步迴歸法,偏最小二乘法的預測精度最低。
침대태양복사、대기온도、공기습도화풍속등기상인소대대두귀일화식피지수(normalized difference vegetation index, NDVI)재매천불동시간적영향,제고대두NDVI적감측정도。해연구채용GreenSeeker수지식광보의대대두묘기、화협기화성숙기3개주요생육계단적NDVI치이소시위단위진행련속감측,병수집측량시적태양복사、대기온도、공기습도화풍속등기상수거,채용편최소이승법、축보회귀화령회귀방법,건립불동기상인소대대두 NDVI 치영향적회귀모형,병분석기정량관계。결과표명,영향대두불동생육기 NDVI 변화적주요기상인소위태양복사화대기온도,풍속화공기습도적영향교소,가이홀략불계。경대3충모형진행예측정도평개후득출,령회귀모형적예측정도최가,기재3개계단적예측균방근오차(RMSE)분별위0.034、0.018화0.016,결정계수(R2)분별위0.820、0.908화0.934,기차위축보회귀법,편최소이승법적예측정도최저。
Normalized difference vegetation index (NDVI) can be used as an ideal indicator of how the crops grow, with which the crops in different growing stages, harmful insects and diseases, water and fertilizer, and the yields can be well predicted. However, the accuracy of NDVI does not remain unchanged due to the ever-changing environmental factors, apart from the impact from the crops growing factors in different stages. The article investigates, using GreenSeeker, the soybean NDVI of its three growing stages of seeding, flowering & podding and maturing in successive hours as the testing unit, for a more accurate monitoring of NDVI affected by air temperature, humidity, solar radiation, and wind speed, etc. in different periods of a day. The results show that the soybean NDVI values, being dynamic in different periods of a day, become smaller from 08:00 or 09:00 am in the morning, reach the valley in 14:00 pm in the afternoon, then give a gradual rise, the whole dynamic process is similar to a quasi-parabola. Moreover, the soybean NDVI values demonstrate different daily variation ranges in different soybean growing stages. From 08:00 am to 18:00pm, the biggest daily variation ranges are respectively among 0.13-0.23, 0.08-0.17 and 0.09-0.19, the biggest relative daily variation ranges are respectively among 20%-26%, 9%-19% and 11%-24%. The correlation study of the soybean NDVI values and the environment meteorological factors show that the changes in soybean NDVI values in its three growing stages are upon great influence by the solar radiation, air temperature, humidity and the wind speed. In four meteorological factors, the solar radiation and air temperature have a negative correlation with the soybean NDVI values at three stages (R2 =0.424, 0.503, 0.631 and 0.602, 0.743, 0.757), humidity shows a positive correlation with the soybean NDVI values(R2 =0.281, 0.435 and 0.654), and the wind speed exerts different influences in different soybean growing stages, specifically, a negative correlation in seeding, flowering & podding (R2 = 0.432, 0.218), and in maturing stage(R2 = 0.127). The regression models were set up to test the impact of four meteorological factors on soybean NDVI and analyze the quantitative relations among them, namely Partial Least Squares (PLS), Stepwise Regression and Ridge Regression. It was found that among four meteorological factors affecting soybean NDVI values, the major factors are the solar radiation and the air temperature, while the minor ones are the wind speed and humidity and their influence on soybean NDVI values can be neglected. In significance tests, predictive accuracy of the three regression models for soybean NDVI in all three growing stages are all statistically significant (P<0.01). By contrast, Ridge Regression has a slightly higher coefficient than partial least squares (PLS) and Stepwise Regression, while the latter two models have almost the same correlation. In contrast of the predictive values and the real test results of the three regression models for soybean NDVI, the Ridge Regression ranks the highest on predictive accuracy, with the Root Mean Square Error (RMSE) of 0.034、0.018 and 0.016 andR2of 0.820, 0.908 and 0.934 in three stages of seeding, flowering & podding and maturating, followed by a less accurate predictive level of Stepwise Regression and the least accuracy of PLS. Regression model can have a better prediction of daily variation trend of soybean NDVI values and a better accuracy of NDVI monitoring.