Statistical Approach
6.3 Materials and methods .1 Yield data
6.3 Materials and methods
Table 6.3 Composition of cultivars in the Kakajan tea estate
Cultivar type Cultivars
1. Old Seed Jats Monipuri, Betjan, Tingamira, Khowang, Kakajan, Nokhoti, Debrapara, Doolia, Joypuri, Rangamattee
2. Selected vegetative clones TinAli-17, Panitola-126, S3A3 3. Tocklai, TRA released
vegetative clones
TV-1, TV-9, TV-18, TV-20, TV-22, TV-23, TV-25
4. Tocklai, TRA released Bi- clonal Seed Stocks
Stock 203, Stock 462, Stock 463
6.3.2 Climate data
Data related to rainfall and temperature was collected for the period 1995–2010 from the Kakajan tea estate. The data were scrutinized carefully by comparing them with the data recorded at neighbouring Meteorological Observatory of Tocklai Experimental Station (TES) and Assam Agricultural University (AAU), located approximately 10 km west and southwest of Kakajan tea estate respectively. The dataset related to monthly rainfall and number of rainy-days was calculated by averaging the data recorded at TES and AAU.
Accordingly, temperature, and sunshine duration data were also computed by averaging the TES and AAU data and assumed to be representative of the entire estate. Different climate variables were grouped into four seasons according to climatic water balance of the study area (section 6.2.1). The notation and units are given below:
Let R stand for the total rainfall in mm, TX is the maximum temperature in °C per day, TN is the minimum temperature in °C per day, TM the mean temperature in °C per day (average of maximum and minimum temperature), DTR the diurnal temperature range in
°C per day (corresponding difference in temperature), and SSH is the average number of bright sunshine hours per day. Let the subscripts 1, 2, 3 and 4 stands for the periods December (preceding year) to February, March to May, June to September and October to November of the current season. Thus, R1 stands for the rainfall, TX1 the maximum temperature, TN1 the minimum temperature, TM1 the mean temperature, DTR1 the diurnal temperature, SSH1 the average sunshine hours for the period December to February (winter season). In addition to these, variability of each of the climatic variables was also
computed by the coefficient of variation (CV) and was computed as the seasonal ratio of the standard deviation to the mean of each climate variables.
6.3.3 Statistical analysis
The impact of climate change on observed yield trend as well as its projected impact on future yield was based on empirical statistical modeling approach. Being a perennial crop, yield of tea is influenced by the weather conditions of the preceding months or seasons in addition to the plucking season’s weather (Sen et al. 1966). Moreover, there exist short- term variation of weather within a growing season (Fordham 1977) and variation between seasons of the year (Barua 1969). Therefore, considering the monthly production pattern as well as climatic water balance of the study area, monthly tea yield data were grouped into three components: the early crop during March to May, main crop during June to September and late crop during October to December. To reduce the number of predictor variables for a relatively short 20-year yield time series, only seasonal averages of different climatic variables were taken into consideration. Apart from measuring the seasonal climate means, variables capturing intra-seasonal variability in temperature and rainfall were also included in this analysis. The variability was measured by the coefficient of variation (CV) calculated as the seasonal ratio of the standard deviation to the mean of each climate variables (Rowhani et al. 2011). Accordingly, climate variables (mean and variability) of winter and pre-monsoon for early crop; winter, pre-monsoon and monsoon seasons for main crop and monsoon and post-monsoon season for late crop were considered in the analysis. Correlation analysis was also performed between seasonal tea yields (early, main and late) and monthly climate variables and discussed suitably as and when the context came in the result and discussion part. The procedures leading to the development of statistical models for assessing the observed and projected climate on tea yield have been discussed in the following section 6.3.3.1.
6.3.3.1 Model for estimating observed impact of climate on productivity
The impact of climate variables (mean and variability) on observed yield trends of early, main and late crop was assessed by developing multiple regression models between first difference of yields (response variables) and first differences of climate variables as predictor variables (Nichols 1997; Lobell et al. 2005). The importance of first difference
method in removing the influence of technology trends from crop yields have been discussed in the previous chapter (section 5.3.1.3). These first-difference values of yield and climate were used to regress in a linear model for each pair of climate and yield data to derive the responses of yield to different climate variables (Zhang et al. 2010). To estimate the relative contribution of each independent climate variable in determining tea yield variability, backward selection procedures were adopted for the regression analysis (Mather 1976; Rowhani et al. 2011). Each model has been checked for multicollinearity by the computation of Variance Inflation Factor (VIF) following O’Brien and Robert (2007).
6.3.3.2 Model for estimating projected impact of climate on productivity
The use of statistical yield models was necessitated by the lack of process-based models for tea crop. One advantage of statistical models is that they intrinsically account for a wide variety of mechanisms (influence of pests, pathogens) that can influence yields in a changing climate (Lobell et al. 2006), which are omitted from most process-based models.
However, unlike process-based models, statistical model do not allow explicit consideration of management changes and CO2 increases in assessing future impact, which may also alter the effect of climate on yields in the future.
In this study, possible effects of projected changes of climate (mean and variability) on tea yields were assessed by developing multiple linear regression models between actual yields of early-, main- and late crop (dependent variables) and mean climatic parameters and their variability terms (independent variables) for the period 1991–2010 (Equation 5.1). A time variable (year 1991, 1992, ---, 2010) was included in these linear models to capture yield changes related to non-climatic factors and other technological development (Rowhani et al. 2011). The multicollinearity between the explanatory variables in each model has been checked by the computation of Variance Inflation Factor (VIF) following O’Brien and Robert (2007). In order to figure out the most important yield influencing variables among all the explanatory variables, a stepwise backward variable selection criteria based on R2 was adopted (Mather 1976; Rowhani et al. 2011).
As the method of the computation of the yield levels under future climate change scenario was not identical with the method of computation of fitted yields under the observed climatic conditions, yield levels have been calculated by using the baseline weather data.
The baseline scenario of the study area is based on 40 years data (1971-2010) has been discussed in Chapter 5 (section 5.3.2.4). Accordingly, the relative changes in yields were worked out by comparing the generated yields under the baseline scenario with generated yields under the climate change scenarios. The main goal was to quantify the sensitivity of tea crop to expected changes in mean state as well as variability of temperature, rainfall, rainy days and sunshine duration, which can provide a basis for prioritizing adaptation efforts.
6.3.3.3 Building up of climate change scenarios
The adopted seasonal temperature and rainfall scenario for 2030 was based on the recent MoEF (2010) report, details of which have been discussed in section 5.3.2.4.3. The temperature scenario represented an increase in temperature of the order of 2.0°C during winter, 2.2°C during pre-monsoon, 1.6°C during monsoon and 2.3°C during post-monsoon seasons in 2030s over the baseline (1961–1990). The intra-seasonal and inter-seasonal variability of temperatures is likely to increase by 10% in 2030s. An increase in monsoon and post-monsoon rainfall by 7% and 4% and decrease of winter and pre-monsoon rainfall by 20% and 15% respectively over 1961–1990 was projected during 2030s over the baseline (1961–1990). A decrease in number of rainy days by 20% during winter, 10%
during monsoon season and increase by 10% during post-monsoon season is expected.
Coefficient of variation of rainfall during winter and pre-monsoon seasons is likely to decrease by 10% over the baseline due to the projected decline in total rainfall during the two seasons. For sunshine duration scenario, the observed linear trend of bright sunshine duration data during the baseline (1971–2010) of the study area was considered in building up of sunshine duration scenario, assuming that the observed trend will continue more or less consistently in the coming 30 years.