Ecotoxicological risk assessment of agriculture intensification in Argentinean cropping lands using a fuzzy-logic based model

Contact information Current agricultural models have an increasing dependence on non-renewable resources that have significant implications for the ecosystem health. Pesticides are considered one main impact force that affects both the structure and functions of agroecosystems. We assessed the environmental pesticide risk (1992-2005) of cropping systems in the Inland Pampa (Argentina) using a fuzzy-logic model. Here, the traditional mixed grazing– cropping systems were being replaced by permanent agriculture. However, there is no clear awareness of the ecotoxicological risk associated with these land use changes.


Results: crop yield and risk over mammals and insects
Results: overall risk and some explanatory factors

Contact information
Current agricultural models have an increasing dependence on non-renewable resources that have significant implications for the ecosystem health. Pesticides are considered one main impact force that affects both the structure and functions of agroecosystems. We assessed the environmental pesticide risk (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) of cropping systems in the Inland Pampa (Argentina) using a fuzzy-logic model. Here, the traditional mixed grazingcropping systems were being replaced by permanent agriculture. However, there is no clear awareness of the ecotoxicological risk associated with these land use changes.

Methods -The model
A fuzzy logic-based expert system were developed. Thresholds and membership functions were developed by means of expert opinion. The input variables are: 1) Dose (D) applied 2) Lethal Dose (LD50) for rats (LD50r) and bees (LD50b)

Data analysis
A crop-field scale database (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)  Also, a data-mining technique, classification and regression tree (CART), was applied in order to analyzed the non-linear nature of the data structure. The pesticide index (P) output were clustered in five groups (in 0.2 intervals from 0.0 to 1.0) and these classes were classified using: 1) Number of products applied per crop field (# products) 2) Time period (SET) 3) Crop specie (CROP) 4) Normalized yield (the relative yield of each field with respect to the maximum of each crop and year) Regarding the average number of pesticides applied, all crops showed a significant increment from 1992 to 2000 and a subsequent reduction from 2000 to 2005 (Spearman rank correlation, P < 0.05) and these reductions were higher in summer soybean, from 6.1 to 3.8 application/year in relation to maize, spring soybean, wheat and sunflower.
When inspecting the ecotoxicological risk form mammals (M) and insects (I) both crop and time period had significant effects on final outcome (P <0.05). However, the insect risk (I) dynamic has a distinctive pattern than mammal risk (M). Although almost all crops (except soybean) showed lower index values (i.e. higher risk) by the end of the period studied, the index reduction were higher for mammal (M) index than insect (I) index Higher yields were associated with higher overall ecotoxicological risk (i.e. lower P values). In almost all crops analyzed the fuzzy-logic ecotoxicological indicators showed negative and significant time trend in the period studied (P<0.05). Summer soybean (S2) was the only crop that exhibited better (higher P index) risk values by the end of the period studied. These preliminary results indicated that higher yield in all the crops analyzed were associated to higher ecotoxicological risk of the pesticides used. However, further work has to be conducted to improve the fuzzy-logic ecotoxicological model by including variables such human toxicological risk or other chemical pesticide characteristics (i.e. soil persistence, solubility). Site study and database  Results showed that harvested yield increased significantly in all crops during the period studied (P<0.005). In 1992-base yield, spring soybean (S1) showed the greatest increment and sunflower (Su) the lowest.
Compared with 1992 data, spring soybean (S1) crop yield increased 75 % (the greatest) and sunflower only 12.3% (the lowest) Although the time period had the most significant effect on final P value, some crops (e.g. maize and sunflower) showed also low but significant previous crop effect in the toxicity pattern of the pesticide use. WFC: winter forage crop Soybean (either as S1 or S2) showed the highest ecotoxicological risk value (i.e. lowest P) in each year (P < 0.05).
A noticeable change in land use in the study area was the increase in soybean cultivation from 13% of the area occupied in the period 1992-1994 to 55% in the period 2004-2005 at the expense of maize and sunflower occupied area. Time period showed a slightly higher explanation of overall ecotoxicological risk (P index) than crop type. Soybean was splitted by this tree side, and showed three contrasting terminal nodes.

Classification and Regression tree (CART): the structure
Half of the field was firstly splitted by means of the number of products/crop (splitting point = 3.5) The multivariate CART analysis determines a final model configuration with 10 terminal nodes and a misclassification rate of 0.41 (i.e. a model R 2 = 0.59). The normalized yield (NY) was not retained in the final tree. Time period (SET) was retained twice in final tree configuration using the same splitting point (SET = 2.5, which represent the average value in the splitting node) SET