JIA-2018-09

2111 WANG Qian-feng et al. Journal of Integrative Agriculture 2018, 17(9): 2107–2117 the robust estimator for the amplitude of trend slopes as proposed by Sen (1968): Slope= Median ( ) x j − x i j − i (1≤ i < j ≤ n ) (6) Where, slope is the monotonic increase or decrease rate, or the linear slope, of the entire data series x k ( k =1, 2, 3, ..., n ) or any segmentation x w ( w = i , i +1, i +2, ..., j ). If this value is positive, the series monotonically increase; and if it is negative, the series monotonically decrease. Median denotes the function to take the median value, and conducted a significance test on the result of the Sen’s trend analysis using the MK approach. 3.3. Spatial characteristics method In this research, the expression of spatial and temporal characteristics is depicted in terms of the spatial pattern and trend of elements in the study area (Wang et al. 2014). For the spatial pattern, the annual mean values of ET during 2000–2014 were computed for each grid to identify the hotspot areas associated with each factor. For the spatial trend, the change of ET trend during 2000–2014 was determined for each grid to identify areas with significant changes. Specifically, spatial pattern and trend are analyzed according to five types of statistical unit with the whole study area and four types of vegetation coverage. In order to quantitatively understand ET responses to environmental factors such as vegetation coverage and precipitation. In order to understand the effects of environmental factors on the ET for 2000–2014, a spatial correlation analysis is conducted. Data on vegetation coverage are calculated based on the NDVI and vegetation types, the detailed calculation process can refer to the previous studies (Mu et al. 2013; Wen et al. 2013). 4. Results and discussion 4.1. Time-series variability characteristics of ET during 2000–2014 Using the ET time-series data, the inter- and intra-annual variations and abnormalities of ET can be effectively identified by the analysis of variability. Fig. 3 shows the monthly mean values and standard deviations of ET during 2000–2014, where the mean values and standard deviations represent water consumption in the region and its dispersion degree, respectively. As illustrated in Fig. 3, ET rates varied seasonally and showed a certain variation in inter-annual ET. ET peaks were above 75 mm in July and August every year, and low ET rates, i.e., below 25 mm, were observed in January, February, March, April, October, November and December due to the lower temperatures. Inter-annual variations indicate that ET rates were above 80 mm in July and August in six years, 2004–2007 and 2012–2013, probably due to higher levels of summer precipitation in those years. The annual maximum value of ET was relatively low during July and August 2009, potentially due to less precipitation during summer. The characteristics of inter- and intra-annual variations of ET for each type of vegetation coverage area were 140 120 100 80 60 40 20 0 12/2000 12/2001 12/2002 12/2003 12/2004 12/2005 12/2006 12/2007 12/2008 12/2009 12/2010 12/2011 12/2012 12/2013 12/2014 Date (mon/yr) ET (mm) Fig. 3 Variability of evapotranspiration in the whole study area in Hebei Province, China during 2000–2014. White line between gray area and black area represents monthly mean value. Gray area represents monthly mean value+standard deviation, back area represents monthly mean value–standard deviation.