UAV-based Visible Light Camera For Monitoring Forage Nutrients And Soil Moisture In Alpine Meadow

UAV visible light sensors are widely used for grassland monitoring and management,but not for forage nutrients and soil moisture estimation.To this end,in the alpine meadow-yak grazing system of Maqu Grassland Agricultural Experiment Station of Lanzhou University,the visible light images of the grazing area were acquired based on the UAV platform,and the quantitative relationships between the nutrient content of forage and soil moisture content and the brightness value of visible light pixels were established to predict and verify the quality and soil moisture content of forage,with a view to providing a basis for a dynamic grassland monitoring technology with multiple locations,large samples,high efficiency and low cost.monitoring technology with multiple locations,large samples,high efficiency and low cost.The main results are as follows.1.UAV-based visible light camera for soil moisture prediction in alpine meadowsThe test sample site was shaded from rain but not the sun,and the UAV was used to take images of the sample site every 2?3 days during soil moisture reduction.Soil moisture measurement,determine the soil moisture content at the time of filming.Modeling and validation,modeling the relationship between brightness values of images and soil moisture,and validating the model with data collected from other grazing areas.The average pixel brightness value(y)of the UAV visible image and the soil moisture(x)from 0 to 10 cm can be fitted by the general linear equation:y=-0.461 x+82.016(R~2=0.6773,P<0.001)for the yak four-season grazing system.For model validation,the absolute values of model efficiency(E,0 to 1,with larger values indicating better model performance),total relative error(|RS|<20%indicating compliance with accuracy requirements)and mean relative error(|RMA|<30%indicating compliance with accuracy requirements),root mean square error(smaller values of RMSE indicating higher model accuracy),coefficient of determination(R~2)between measured and estimated values.The validation data were E=0.366,RS=-0.007%,RMA=1.424%,RMSE=0.652,R~2=0.6118 and Pearson's r=0.782,respectively.UAV for grassland soil moisture determination is related to regional precipitation and requires continuous acquisition of data and images of soil moisture changes at the same location for modeling.In areas with annual precipitation<300 mm,images and soil moisture data during the process of soil moisture change from more to less are obtained by rain shading after irrigation;in areas with annual precipitation>400 mm,images and soil moisture data during the process of soil moisture change from more to less are obtained by rain shading;in areas with 300?400 mm annual precipitation,rain shading or irrigation is selected according to the specific situation.2.UAV-based visible light camera for forage nutrients prediction in alpine meadowsThe UAV carries a camera to acquire images,and takes pictures every 15 days,and uses Image J analysis to derive brightness value parameters.Vegetation samples were collected and analyzed,and forage samples were collected from the shooting area while shooting to analyze the nutrient composition.Model building and validation,using three-quarters of the data to build a quantitative model of the pixel brightness values of the images versus forage quality,and using the remaining one-quarter of the data for model validation.The fitted equations for the brightness values(y)of the drone acquired images with chlorophyll content a,chlorophyll b and total chlorophyll were y=-34.849Chl.a+136.2(R2=0.8001,P<0.001),y=-39.18Chl.b+120.69(R2=0.7578,P<0.001),y=-19.584Chl.t+128.57(R2=0.9123,P<0.001).Chlorophyll content(y)and nutrient content could be fitted by general linear equations,and the general linear models for crude protein(CP)content and chlorophyll a,chlorophyll b and total chlorophyll were Chl.a=0.2565CP(%)-1.0463(R2=0.7957,P<0.001),Chl.b=0.2027CP(%)-1.088(R2=0.6317,P<0.001),Chl.t=0.4644CP(%)-2.3043(R2=0.7391,P<0.001);the general linear models of water soluble carbohydrate(WSC)content with chlorophyll a,chlorophyll b and total chlorophyll were Chl.a=0.1418WSC(%)-0.4763(R2=0.7787,P<0.001),Chl.b=0.1011WSC(%)-0.4885(R2=0.6150,P<0.001),Chl.t=0.2416WSC(%)-1.078(R2=0.7326,P<0.001).Forage quality was mainly positively correlated with CP content,while more than half of the proteins in CP were Rubisco enzymes,and Rubisco enzymes were chlorophyll-protein complexes.Plant leaves with different chlorophyll content can be reflected by UAV visible light images.In the yak all-season grazing system,the mean pixel brightness value(y)of the UAV visible images and the nutritional quality of the forage can be fitted by a general linear equation.Where the brightness are better fitted with CP and WSC,and the fitted equations are y=-8.1831CP(%)+162.43(R~2=0.6647,P<0.001)and y=-4.3113WSC(%)+141.64(R~2=0.5517,P<0.001)respectively.For model validation,the validation data of CP were E=0.707,RS=-2.044%,RMA=7.283%,RMSE=0.746,R~2=0.7574 and Pearson's r=0.870.The validation data of WSC were E=0.707,RS=-2.044%,RMA=7.283%,RMSE=0.746,R~2=0.7574 and Pearson's r=0.870,respectively.