Daniel C. Yoder

OPTIMIZATION OF FUZZY EVAPOTRANSPIRATION MODEL THROUGH NEURAL TRAINING WITH INPUT–OUTPUT EXAMPLES

In a previous study, we demonstrated that fuzzy evapotranspiration (ET) models can achieve accurate estimation of daily ET comparable to the FAO Penman–Monteith equation, and showed the advantages of the fuzzy approach over other methods. The estimation accuracy of the fuzzy models, however, depended on the shape of the membership functions and the control rules built by trial–and–error methods. This paper shows how the trial and error drawback is eliminated with the application of a fuzzy–neural system, which combines the advantages of fuzzy logic (FL) and artificial neural networks (ANN). The strategy consisted of fusing the FL and ANN on a conceptual and structural basis. The neural component provided supervised learning capabilities for optimizing the membership functions and extracting fuzzy rules from a set of input–output examples selected to cover the data hyperspace of the sites evaluated. The model input parameters were solar irradiance, relative humidity, wind speed, and air temperature difference. The optimized model was applied to estimate reference ET using independent climatic data from the sites, and the estimates were compared with direct ET measurements from grass–covered lysimeters and estimations with the FAO Penman–Monteith equation. The model–estimated ET vs. lysimeter–measured ET gave a coefficient of determination (r 2 ) value of 0.88 and a standard error of the estimate (Syx) of 0.48 mm d –1 . For the same set of independent data, the FAO Penman–Monteith–estimated ET vs. lysimeter–measured ET gave an r 2 value of 0.85 and an Syx value of 0.56 mm d –1 . These results show that the optimized fuzzy–neural–model is reasonably accurate, and is comparable to the FAO Penman–Monteith equation. This approach can provide an easy and efficient means of tuning fuzzy ET models.

The design philosophy behind RUSLE2: evolution of an empirical model.

The RUSLE2 erosion-prediction model is designed specifically as a conservation planning tool, and notnecessarily to provide the most accurate erosion estimates. This paper describes the design philosophy upon whichthe RUSLE2 model is built, including the implications of the specific goals set for the program and of the basicmodeling approach. It describes how RUSLE2 was designed to easily handle erosion science changes, and how theinterface and model architecture accommodate specific needs of conservation planners.

DESIGN AND EVALUATION OF AN IMPROVED FLOW DIVIDER FOR SAMPLING RUNOFF PLOTS

An improved flow divider was designed to simplify and lower the cost of collecting runoff data from research plots. The system was designed around commercially available and inexpensive 5-gal (19-L) plastic buckets with screw top lids. A precision cut sheet-metal divider “crown” is fastened to the lid, allowing it to be easily transferred between buckets. The divider crown can be configured to handle various flow rates by specifying the number of flow divisions. Laboratory evaluation of the design indicated that the system divides runoff with accuracies within .5% over most of the flow range and within .15% at very low and very high flows. These results are similar to those found for the more traditional flow divider designs. Adding sediment to the inflow at three different flow rates yielded sediment division accuracies within 7%. Five field research projects have used the divider system with few problems. The average cost of this system is approximately US

The application of the Revised Universal Soil Loss Equation, Version 2, to evaluate the impacts of alternative climate change scenarios on runoff and sediment yield

500 per plot, in comparison to the US

WOOD CHIPS AS A SOIL COVER FOR CONSTRUCTION SITES WITH STEEP SLOPES

3000 to

SOILWATER SENSOR PERFORMANCE

5000 it often costs to instrument a plot using standard equipment.

Estimation of reference crop evapotranspiration using fuzzy state models

The Revised Universal Soil Loss Equation, Version 2 (RUSLE2), provides robust estimates of average annual sheet and rill erosion for one-dimensional hillslope representations. Extensive databases describing climate, soils, and management options have been developed and are widely used in the United States for conservation planning. Recent RUSLE2 enhancements allow estimation of erosion and runoff from a representative sequence of runoff events that are suitable for linkage with an ephemeral gully model. This paper reviews the sensitivity of RUSLE2 erosion estimates to possible climate change scenarios, demonstrates its ability to evaluate alternative management adaptations, and compares predictions with observations of runoff and sediment yield from a 6.6 ha (16 ac) research watershed located near Treynor, Iowa. When applied to a representative hillslope profile with conventional tillage corn (Zea mays L.), increasing monthly temperature by 0.8°C (1.5°F) and rainfall depth, rainfall erosivity density, and 10-year, 24-hour rainfall depth each by 10% cumulatively increased sheet and rill erosion by 47% and increased runoff by 33%, assuming there was no change in corn yield. If the climate changes decreased corn yield by 10%, the overall effect was to increase soil loss for conservation planning by 63%. These results demonstrate that modest and expected changes in climate will significantly increase the risk of soil erosion, and improved conservation management will be an important part of successful adaptation.

Curve Numbers for Low-Compaction Steep-Sloped Reclaimed Mine Lands in the Southern Appalachians

Wood chips were studied for their efficacy in controlling soil erosion on a steep construction site with disturbed soils. The purpose of the research was twofold: to determine if wood chips could be used to reduce the off–site movement of soil during construction activities, and to find an environmentally sound alternative to the landfill disposal of wood wastes generated in the urban forest. The research was conducted on field plots that received natural precipitation. Twelve erodible plots were established on an embankment with a 55% slope and an elevation change of nearly 12 m. Each plot had a width of 3 m and a horizontal slope length of 10 m. A series of flow dividers was installed at the toe of each plot to measure runoff and sediment. Four treatments were studied: large wood chips, small wood chips, a mixture of wood chip sizes, and a control with no chips. The mixture of wood chip sizes represented the size distribution that was found to occur from chippers. The wood chips were applied at a rate that covered 80% of the soil surface. The erosion rate for the small wood chip treatment was not significantly different from the zero–cover plots. The erosion rates from the large wood chip and mixture of chip sizes were not significantly different from one another, but were significantly different from the zero–cover treatment. Overall, in comparison to the zero–cover treatment, the small wood chip treatment reduced erosion by 22%, the large wood chips reduced erosion by 78%, and the mixture of chip sizes reduced erosion by 86%. The results of this project indicate that wood chips (as produced by a chipper) should be utilized as a soil cover and need not be discarded as solid waste.

Sediment loss and nutrient runoff from three fertilizer application methods

Numerous sensors are currently available to measure soil water content. Although several studies have compared relative sensor performance in the field, there have been no reports of sensor comparisons with carefully controlled soil water contents. Weighed soil columns were used to compare 23 soil water sensors representing eight sensor types. Included in the study were: a Troxler neutron gage, a Troxler Sentry® 200-AP capacitance probe, Aqua- Tel® capacitance sensors, time domain reflectometry (TDR) with two- and four-rod waveguides, gypsum blocks, Watermark® electrical resistance blocks, and Agwatronics® heat dissipation blocks. Measurement errors of the volumetric water content of the soil were determined for each sensor over a range of water contents from the maximum water holding capacity to below 5%. A loam and a sandy loam soil were wetted to the maximum water holding capacity and subsequently drained through four cycles. Sensors were calibrated using data from the first cycle and measurement errors for each sensor were determined using those calibrations in three additional cycles. Measurements outside the range of 0 to 50% volumetric soil water content were discarded. Of 64 possible readings in the test, only the neutron gage and the Aqua-Tel capacitance sensor gave 64 viable readings. The Sentry capacitance probe had the lowest measurement error and yielded 62 of 64 viable readings. Watermark sensors had measurement errors similar to the electrical capacitance sensors, but averaged 57 of 64 viable readings. In order of decreasing performance, the Aqua-tel electrical capacitance sensor, the Sentry electrical capacitance probe, the neutron gage, and the Watermark sensors performed best in this study when accuracy, reliability, durability, and installation factors were considered.

DEVELOPMENT OF A SYSTEM TO MANAGE PESTICIDE–CONTAMINATED WASTEWATER

Microirrigation can potentially “spoon feed” nutrients to a crop. Accurately supplying the crop’s nitrogen (N) needs throughout the season can enhance crop yields and reduce the potential for groundwater contamination from nitrates. A 2–year study (1990–1991) was conducted on a Keith silt loam soil (Aridic Argiustoll) to examine combinations of both preplant surface application (30 cm band in center of furrow) and in–season fertigation of N fertilizer for field corn (Zea mays L.) at three different levels of water application (75%, 100%, and 125% of seasonal evapotranspiration) using a subsurface drip irrigation (SDI) system. The method of N application did not significantly affect corn yields, apparent plant nitrogen uptake, or water use efficiency, but all three factors were generally influenced by the combined total N amount. The N application method did have an effect on the amount and distribution of total soil N and nitrate–N in the soil profile following harvest. In both years, nearly all of the residual nitrate–N after corn harvest was within the upper 0.3 m of the soil profile for the treatments receiving only preplant–applied N, regardless of irrigation regime. In contrast, the nitrate–N concentrations increased with increasing rates of N injected by the SDI system and migrated deeper into the soil profile with increased irrigation. The results suggest that N applied with an SDI system at a depth of 40–45 cm redistributes differently in the soil profile than surface–applied preplant N banded in the furrow.