植被生产力和生物量是表征植被状态的重要指标，在气候变化及碳平衡中有重要作用。滦河流域上游的塞罕坝地区，是守卫京津冀的重要生态屏障，研究该地区地表植被覆盖的变化情况，定量评价该地区不同植被生态系统的生产力，能有效反映该地区生态环境治理成效。已有的生物量产品多以中低分辨率为主，而植被生产力产品高时空分辨率也难以兼得，因此有必要发挥不同分辨率遥感数据的优势，估算滦河流域上游地区高分辨率生物量和高时空分辨率生产力。本研究结合多源遥感数据，基于随机森林算法、遥感数据融合模型以及植被生产力估算模型估算30 m分辨率森林地上生物量和时间序列的植被生产力，并对2001-2017年NPP进行时空变化分析。主要结论如下：

（1）基于GLAS数据、地表分类产品和Landsat数据，利用随机森林模型估算研究区30 m分辨率森林冠层高度，利用地面实测冠高进行检验，结果显示RMSE为3.1 m，R²为0.34。建立冠高、植被指数和实测生物量的回归关系，估算30 m分辨率森林地上生物量，利用地面实测生物量对估算结果进行检验，RMSE为30.33 Mg/ha，R²为0.50。研究区森林地上生物量总体分布在50-200 Mg/ha之间，高值主要分布在中部小滦河附近林区和东南部兴洲河河流西岸、滦河东岸林区。

（2）利用1 km/500 m分辨率GLASS LAI和FPAR产品以及Landsat数据，基于回归树模型和STARFM模型，估算2001-2017年30 m分辨率LAI和FPAR数据。利用地面实测LAI数据进行精度验证，R²为0.72，RMSE为1.1，显示LAI估算结果较好。将降尺度LAI/FPAR与GLASS LAI/FPAR产品年均值变化趋势进行对比，两者总体趋势较为一致，可以用于GPP和NPP的估算。研究区LAI和FPAR多年空间分布总体呈现西北低东南高的趋势，时间上总体呈现波动上升趋势。

（3）基于MuSyQ-NPP模型，利用2001-2017年LAI和FPAR序列数据、气象数据、森林生物量数据等，进行GPP和NPP的计算。根据森林地类的地面实测NPP数据对NPP模拟结果进行检验。结果显示，总体上RMSE为81.70 gC^.m^-2.yr^-1，R²为0.68，NPP估算精度较为可靠。

（4）研究区NPP总体呈现东南向西北递减的分布趋势。东南部和东北部森林、西部农田年均值最高，中部和北部草地NPP年总量均值较低。2001-2017年植被NPP年总量在3.81-5.56 GgC^.yr^-1之间，最低值和最高值分别出现在2002年和2017年，总体呈现波动增加趋势，年增长率为0.1 GgC^.yr^-1。一元线性回归结果显示大部分区域NPP在17年间都有所增加。对NPP季节总量进行统计，结果表明春夏秋季NPP分别占全年总量的12.7%、71.8%、15.4%。对各植被类型分析NPP年际变化趋势，增速最快的为作物，其次是草地、常绿针叶林，落叶阔叶林和落叶针叶林。混交林单位面积NPP最高，草地NPP年总量最高。对NPP和降水、气温的相关性分析显示，NPP和降水的正相关性较强，和气温相关性较弱，多数区域呈现负相关。利用SPI12分析干旱对NPP年际变化的影响，发现在大多数年份NPP是受到降水的影响而变化，但在某些年份温度起主导作用。

Vegetation productivity and biomass are key indicators in characterizing vegetation activity. They both play an important role in climate change and carbon balance. The Saihanba area in the upper Luanhe River Basin is a significant ecological barrier for guarding the Beijing-Tianjin-Hebei region. The quantitative evaluation on productivity of different vegetation ecosystems can effectively reflect the effect of ecological environment management in this area. However, the spatial resolution of existing forest biomass products is too low to meet the needs of forest investigation and dynamic monitoring, and the contradiction between spatial and temporal resolution of vegetation productivity products has also limited their application in ecosystem services. In this condition, making the most of the advantages of remote sensing data in different scales to generate high spatial resolution forest biomass and high spatial-temporal resolution vegetation productivity in upper Luanhe River Basin is of great significance. Based on the random forest model, remote sensing data fusion model and MuSyQ-NPP model, the forest aboveground biomass and time series vegetation productivity of 30m resolution were estimated. The spatial and temporal variation of NPP in the upper Luanhe River Basin from 2001 to 2017 were analyzed. The main conclusions are as follows:

(1) Based on GLAS data, landcover product and Landsat data, the 30 m forest canopy height was generated by using random forest model. Compared with field canopy height, we gained R² of 0.34, RMSE of 3.1 m. The regression relationship between canopy height, vegetation index and field biomass was established to generate 30 m forest aboveground biomass. Compared with field biomass, the model gained R² of 0.50 and RMSE of 30.33 Mg/ha. The forest biomass in the study area was between 50 and 200 Mg/ha. High biomass was mainly distributed around the Xiaoluanhe River, the west bank of Xingzhou River and the east bank of Luanhe River.

(2)Based on 1 km/500 m GLASS LAI/FPAR products and Landsat data, 30 m LAI and FPAR from 2001 to 2017 were generated by using Cubist regression tree algorithm and STARFM. The LAI estimates were validated by field LAI. The R² was 0.72 and RMSE was 1.1. The temporal variation trend of estimated LAI/FPAR and GLASS LAI/FPAR matched well, which demonstrated its utility in generating GPP and NPP. The LAI and FPAR in the study area was higher in the southeast than in the northwest and had a fluctuatly rising trendency from 2001 to 2017.

(3)Based on MuSyQ-NPP model, 30 m GPP and NPP from 2001 to 2017 were generated by using estimated LAI/FPAR, meterological data and forest biomass. Compared with field forest NPP, we gained R² of 0.68, RMSE of 81.70 gC^.m^-2.yr^-1, which proved that estimated NPP were reliable.

(4)The estimated NPP in the upper Luanhe River Basin revealed a decreasing pattern from southeast to northwest. NPP high values distributed mainly in southeast and northeast forest area and west cropland area, and low values distributed extensively in middle and north grassland area. Annual NPP had a fluctuating increase trend from 2001 to 2017 with values ranging between 3.81 and 5.56 GgC^.yr^-1, with an annual increase of 0.1 GgC^.yr^-1. The maximum NPP value presented in 2017 and minimum value in 2002. The linear regression results indicated that NPP presented an increasing tendency in most of the study area. Statistical of seasonal NPP showed that spring, summer and autumn NPP accounted for about 12.7%, 71.8% and 15.4%, respectively. Statistical of NPP for different vegetation types showed that croplands increased most rapidly, followed by grassland, evergreen needle-leaf forest, deciduous broadleaved forest and deciduous needle-leaf forest. Mixed forest was highest in terms of average annual NPP and grassland had highest value in terms of annual NPP. The results of correlation analysis showed that there was a close positive correlation between NPP and precipitation, and negative correlation with temperature in most of the study areas. SPI12 was used to analyze the effect of drought on NPP. It appeared that NPP was affected by precipitation in most years, but temperature played a dominant role in some years.