Relationship between the actual test results of groundwater arsenic level and the predictions of a geographic information system model in four provinces of Huaihe River Basin

Objective To test the accuracy of predicted results by a geographic information system(GIS) model with the actual distribution of groundwater arsenic concentration in four provinces including Shandong,Henan,Anhui and Jiangsu of Huaihe River Basin.Methods The results of groundwater arsenic level of the four provinces in Huaihe River Basin were cited as the data resource; after stratified by arsenic in water,water arsenic ＞ 0.01 mg/L as statistical standards,the number of villages and counties(districts) with arsenic level higher than the standards was calculated.Meanwhile,locations of counties(districts) sampled on the map of GIS predictive model were marked; the gray level of arsenic contaminated probability 0.3 was regarded as the criterion and the consistence of both results predicted by GIS model and detected actually in each province was analyzed.Results A total of 61 824 wells distributed in 2 781 villages around the four provinces of Huaihe River Basin were sampled,and 474 of the 2 781 villages were confirmed as high arsenic villages(arsenic ＞ 0.01 mg/L),with an average detection rate of 17.04％(474/2 781).The detection rates of high arsenic villages in Shandong,Henan,Anhui and Jiangsu were 13.19％ (79/599),23.82％ (101/424),74.25％ (199/268) and 6.38％ (95/1 490),respectively,and the differences of detection rates among the provinces were statistically significant(x2 =820.84,P ＜ 0.05).County(district) as a unit,among all the 103 counties(districts),the number of counties where the probability of high arsenic concentration in groundwater predicted by GIS model that greater than 0.3 was 72,accounting for 69.90％(72/103) of total counties,in which Shandong,Henan,Anhui and Jiangsu were 67.86％(19/28),61.36％ (27/44),85.71％ (12/14) and 82.35％ (14/17),respectively.Among all 103 counties (districts),the number of counties(districts) where the detection rates of high arsenic villages beyond the standards was 62,and among these 62 counties,there were 42 counties where the probability of high arsenic concentration in groundwater predicted by GIS model was greater than 0.3,and the overall sensibility was 67.74％(42/62); among all the 41 counties where the detection rates of high arsenic villages were not beyond the standards,11 counties where the probability of high arsenic concentration in groundwater predicted by GIS model was lower than 0.3,and the overall specificity was 26.83％(11/41); the sensitivities of GIS model in Shandong,Henan,Anhui and Jiangsu were 57.89％ (11/19),59.09％ (13/22),84.62％ (11/13) and 87.50％ (7/8),respectively.Conclusion The results of groundwater arsenic investigated in the four provinces of Huaihe River Basin and predicted by GIS model are consistent,and we have proved that the results of GIS predictive model are accurate. Key words: Huaihe river basin;  Groundwater;  Arsenic;  Geographic information system; Predictive model