Classification Of Agricultural Drought Based On Historical Drought Data

Drought is the severest natural disasters affecting agricultural production in China. Agriculturallosses caused by drought accounts for more than55%a year of the total losses caused by naturaldisasters. The ability to quantitatively evaluate the degree of agro-drought and scientifically definedrought levels is significant for agricultural disaster prevention and reduction. The scope of the studycovered national and region levels and data that were used was historical agricultural drought disasterand crop planting data from1951to2010. Conventional statistical methods were used to analyze thetemporal variations of agricultural drought. To quantitatively grade and classify by index China and itsregions agro-drought information from1951to2010, the hierarchical clustering methods, stepwiseclustering method and other multivariate statistical analysis methods were used.The classification index was then verified by Bayes discriminate analysis and actual statisticalvalidation of agricultural drought was used to instantiate the graded results. Verification and validationwere done in order to establish grading standards and achieve quantitative classification of Chineseagricultural drought disasters.The main conclusions are as follows:(1) In recent60years, the agro-drought influenced&affected areas of China showed aninter-decadal tendency to expand and the amount of grain losses also showed10-year increase. Inaddition, compared with the previous30years, the influenced areas expanded by21.7%and affectedareas by75.0percent of recent30years, while grain losses increased207.0%. These show that thenational agricultural drought disaster is aggravating.(2) Agricultural drought in North China and Southwest China were aggravating and it wasmitigation in middle region. Compared with the previous30years, the influenced&affected areas inthe northeast China, the northwest China and the southwest China showed an upward trend, respectively63.8%,55.4%and32.6%for the drought influenced areas,43.1%,36.3%and19.8%for the droughtaffected areas, respectively. While agricultural drought in Huang-Huai-Hai and the Middle and LowerReaches of Yangtze River region were lightening, their drought influenced areas showed a trend ofshrinking, respectively for14.0%and29.4%, while6.2%and36.6%for the drought affected areas.(3) According to the actual statistical agro-drought events, we can know the classifications ofCI&Cqare better than the other indexes while the rate of crop drought influenced areas (Ia) better thanthe rest.Finally,we choose the classification of CI&Cqand Iaas the standards degree of theagro-drought, which are described in the following:National rate of crop drought influenced areas (Ia), classification category calculated by averageand gravity methods, that is:0＜Ia≤7%，7%＜Ia≤13%，13%＜Ia≤18%，18%＜Ia≤23%，Ia＞23%，Respectively, no drought (D0), light drought (D1), moderate drought (D2), severe drought (D3), andextreme drought (D4).According to the variance contribution rates of the two above indicators, the CI index was constructed. Ultimately, choosing classification results by average and gravity methods, theclassification results for CI are:0＜CI≤6%，6%＜CI≤13%，13%＜CI≤18%，18%＜CI≤23%，CI＞23%, respectively, D0, D1, D2, D3, and D4.Regional agricultural drought classification by variance analysis was used to divide China into twoparts, namely the north part (the Northeast, Northwest and Huang-huai-hai) and the south part (theSouthwest, South China and the Middle and Lower Reaches of Yangtze River). And then, selecting theNortheast and the Southwest as typical regions, the data of the two regions were integrated into a newdata series. The new data series was studied using the stepwise hierarchical cluster analysis and finally,Bayes discriminate analysis methods was used for verification. Criteria values were as follows:Regional rate of crop drought influenced areas (Ia), classification results were: D0:0＜Ia≤6%, D1:6%＜Ia≤13%, D2:13%＜Ia≤22%, D3:22%≤＜Ia≤36%, D4: Ia＞36%.According to the variance contribution rates of the two above indicators, Cqindex was constructed.The classification results for Cqwere D0:0＜Cq≤6%, D1:6%＜Cq≤12%, D2:12%＜Cq≤20%, D3:20%＜Cq≤32%, D4: Cq＞32%.（4）We use average method, gravity methods and stepwise cluster method to cluster the indexes, andthere classifications are significant and can be explained reasonably, which indicate the above clustermethods are convenient and efficient for agro-data of long time series.