针对现有非接触式牛个体图像识别模型体积大、参数多、资源占用较大等问题,提出了一种基于改进YOLOv5 模型的轻量级牛个体图像识别模型(Light YOLO Net,LY-Net).将YOLOv5 模型的主干网络替换为轻量级网络Ghost Net,并采用CARAFE(轻量级通用上采样算子),减少网络参数,实现网络轻量化;采用Focal-EIoU Loss作为损失函数,加速收敛并提高了速度.采用甘肃省张掖市某养殖场的 30 头牛,共 6 775 幅牛个体图像作为样本数据,进行模型的训练、验证、测试.实验结果表明:LY-Net模型对牛个体的识别精确率约为 99.6%,召回率约为 99.5%.该模型能够在对牛个体图像高效且准确识别的同时,实现模型的小型化、轻量化.

以C为驱动的WOFOST作物生长模型是基于作物生理生态过程,综合考虑了CO2、土壤、气候等因素对产量的胁迫作用,因此,对WOFOST模型参数进行本地化和优化便可实现时间连续且高精度的草地生物量监测.为探讨WOFOST参数敏感性分析结果在不同草地类型覆盖区表现出的不确定性问题,在天祝藏族自治县不同草地覆盖区选择了4个站点,利用气象数据、草地实测数据及土壤数据,基于扩展傅里叶幅度敏感性检验法(EFAST)研究潜在水平(指保证营养元素和水分为最佳供应,草地地上生物量仅由辐射、温度和作物特性决定)和水分限制水平(假设营养元素的供给仍然是最佳的,但需考虑土壤有效水分对蒸发和草地生物量的影响)下WOFOST模型在不同草地类型覆盖区的全局敏感性参数和优化模型模拟精度.结果表明潜在生产水平下草地地上生物量(AGB)的敏感参数有比叶面积(SLATB)、单叶片CO2的初始光能利用率(EFFTB)、最大光合速率(AMAXTB)、根相对维持呼吸速率(RMR)、总干物质占根和叶的比例(FRTB和FLTB),水分限制条件下的敏感参数有SLATB、AMAXTB、RMR和FLTB.不同生产水平下叶面积指数(LAI)的敏感参数一致,从出苗到出苗后60 d主要受到SLATB、FLTB和FRTB的影响,出苗后60～200 d的敏感性参数为FLTB、FRTB、SLATB和漫射可见光的消光系数(KDIFTB),LAI开始下降后受到KDIFTB的敏感性增强.其中,山地草甸AGB的模拟值与观测值模拟精度最高,R2=0.94、RMSE=11.71 g·m-2,高寒草甸模拟精度最低,R2=0.83、RMSE=32.68 g·m-2.温性荒漠草原LAI的模拟值与观测值模拟精度最高,R2=0.96、RMSE=0.02,温性草原模拟精度最低,R2=0.66、RMSE=0.38.敏感性分析方法在WOFOST模型中的应用减少了人为主观因素的影响,极大地缩短了调参时间,对获取时间连续的草地生长监测方法选择提供参考.

Drought indicators based on remote sensing include the temperature vegetation dryness index (TVDI) and the crop water stress index (CWSI). The data was processed using TVDI, which was calculated by parameterizing the MODIS EVI and LST data connection. For drought monitoring, we compared the efficiency of TVDI with that of CWSI, which is obtained from the MOD16A2 products. The study's findings revealed that drought conditions measured by TVDI and CWSI had a number of differences and similarities, which indicated that both CWSI and TVDI can be used for drought monitoring, although they had some discrepancies in the spatiotemporal characteristics of drought intensity in this region. High TVDI values were mainly concentrated on the northwestern Sichuan Plateau and mountainous areas of southwestern Sichuan, corresponding to extreme drought. The Panzhihua and the mountainous area of southwestern Sichuan had relatively high CWSI values. Spring had the highest TVDI values, followed by autumn and winter. TVDI and CWSI have different patterns, showing moderate and severe drought conditions in different areas. Overall, CWSI values showed a significant decreasing trend (P < 0.05) from 2001 to 2020. The overall trend change of TVDI was insignificant, mainly based on an insignificant increase and an insignificant decrease. An extremely significant decreasing trend, mainly concentrated in the eastern Sichuan basin plain, accounted for 15.54% of the entire province. Spring, summer, autumn, and winter account for 74.33%, 59.15%, 68.28%, and 64.87% of Sichuan Province, respectively, in total area change. The eastern Sichuan basin plain showed a significant increasing trend, accounting for 1.69% of the province in winter. TVDI correlates positively with Yearly maximum value of daily minimum temperature (TNx), Yearly maximum value of daily maximum temperature (TXx), and Yearly maximum consecutive one-day precipitation (PX1), and negatively with Yearly minimum value of daily maximum temperature (TXn), Yearly minimum value of daily minimum temperature (TNn), and Yearly mean temperature (YMT).