Peter Enright

APPLICATION OF DECISION TREE TECHNOLOGY FOR IMAGE CLASSIFICATION USING REMOTE SENSING DATA

Abstract Hyperspectral images of plots, cropped with silage or grain corn and cultivated with conventional tillage, reduced tillage, or no till, were classified using the classification and regression tree (C&RT) approach, an innovative intelligent computational algorithm in data mining. Each tillage/cropping combination was replicated three times, for a total of 18 plots. Five hyperspectral reflectance measurements per plot were taken randomly to obtain a total of 90 measurements. Images were taken on June 30, August 5, and August 25, 2000 to reflect three stages of crop development. Each measurement consisted of reflectances in 71 wave bands ranging from 400 to 950 nm. C&RT models were developed separately for the three observation dates, using the 71 reflectances as inputs to classify the image according to: (a) tillage practice, (b) residue level, (c) cropping practices, (d) tillage/cropping (residue) combination. C&RT models could generally distinguish tillage practices with a classification accuracy of 0.89 and residue levels with a classification accuracy of 0.98.

Water Quality Modeling of Two Agricultural Fields in Southern Quebec Using SWAT

To study the dynamics of nutrient transport at the field scale, we collected data from two tile-drained agricultural fields in the Pike River watershed of southern Quebec. A two-year data set was used to calibrate and validate the Soil and Water Assessment Tool (SWAT) for sediment, nitrate, and phosphorus loads exiting the field through surface runoff and tile drainage. We found that SWAT output on water quality required an accurate estimation of the timing and form of field management practices employed. After calibration, the monthly coefficients of performance (Cp) over four site-years varied from 0.23 to 0.89 for sediment loads, from 0.48 to 1.35 for nitrate loads, and from 0.38 to 0.67 for total phosphorus loads. Subsurface nitrate loads accounted for 97.7% and 86.7% of the total nitrate yield, while particulate phosphorus accounted for 61.2% and 87.7% of total phosphorus load on sites 1 and 2, respectively. SWAT underestimated nitrate loads in subsurface drainage during spring snowmelt and large storms. Sediments and particulate phosphorus predictions were most accurate of all simulated parameters, whereas dissolved phosphorus was marginally overestimated year-round. Overall, SWAT satisfactorily reproduced field observations for sediment and nutrient transport and could be used to compare the impacts of implementing different best management practices (BMP) on individual fields for the study site.

PHOSPHORUS LOSSES IN SURFACE RUNOFF AND SUBSURFACE DRAINAGE WATERS ON TWO AGRICULTURAL FIELDS IN QUEBEC

Concentrations of phosphorus (P) above the water quality guideline of 0.03 mg L-1 are commonly found in rivers draining agricultural lands in Quebec. This gives rise to eutrophication, which has become quite problematic in some parts of Quebec. Agricultural nonpoint source pollution is the dominant source of this P. While the links between P losses, surface runoff, erosion and manure management are reasonably well understood, there is far less information available on the movement of P via subsurface tile drainage systems. Since the majority of intensively managed agricultural lands in Quebec are subsurface drained, this represents a major knowledge gap. Two field sites, located in the Pike River Watershed of Southwestern Quebec, which drains into Lake Champlain, have been instrumented to measure and sample surface runoff and tile drainage waters. Monitoring has been ongoing since October 2000. Results are reported here for the six site-years of data. Subsurface drainage was the dominant pathway by which water left the fields. On average, tile drainage accounted for 81% of the total annual drainage. Surface runoff was responsible for majority of the annual P losses. It accounted for 60% of the annual total, which was on average 1.09 kg/ha. The flow weighted average annual P concentration in subsurface drainage waters ranged between 0.06 mg L-1 and 0.37 mg L-1. The flow weighted average annual P concentration in surface runoff waters ranged between 0.20 mg L-1 and 2.15 mg L-1. On average, P concentrations in surface runoff were 10.9 times higher than those found in subsurface drainage waters, indicating that subsurface drainage may reduce annual P losses. These results also reveal that soil test P and percent P saturation are, on their own, inadequate indicators of potential for P pollution.

Modeling Water Table Depth, Drain Outflow, and Nitrogen Losses in a Cold Climate Using DRAINMOD 5.1

The performance of the DRAINMOD 5.1 model was tested for conditions, including freezing and thawing, that prevailed at a 4.2 ha field research facility located at St. Emmanuel, Quebec, Canada. Using two years (1998 and 1999) of data collected from the site, the models ability to predict water table depth (WTD), drain outflow, and nitrate (NO3--N) loads in drain water was tested. The site was arranged in a split-plot design, with two N fertilizer rates (120 and 200 kg ha-1, main plots) factorially combined with two modes of water table management: sub-irrigation (SI) at a WTD of 0.6 m vs. free drainage (FD) at a drain depth of 1.0 m (subplots). The model was calibrated using water table depth and subsurface flow data from 1998 and validated with 1999 data. The models accuracy was evaluated using the coefficient of determination (R2), modeling efficiency (E), and other statistical parameters. DRAINMOD 5.1 performed well at simulating the number and timing of drainage events in both snowmelt and later-season periods. The model underestimated annual drain outflow under free drainage and sub-irrigation by 12 mm and 20 mm, respectively, in both years. The model simulated the pattern of water table fluctuations fairly well, with R2 values ranging from 0.81 to 0.91, indicating good model fit. The model performed well in predicting total NO3--N loads in subsurface flow, even though there was a tendency to underestimate loads under both free drainage and sub-irrigation treatments.

Comparison of experimental and simulated results for nutrients transport on agricultural fields in Quebec.

Phosphorus concentrations exceeding the Ministry of the Environment’s water quality guidelines are common in the waters of the Pike River, situated within the Missisquoi Bay Watershed of southern Quebec. Non-point source pollution from agricultural activities is considered as the principal source of this phosphorus. Historically, surface runoff was considered as the primary cause of P pollution. Consequently, few studies have been conducted to examine the transport of phosphorus through tile drainage systems. Since a majority of agricultural fields in Quebec are subsurface tile drained and because soils in many areas have experienced P enrichment due to concentrations of the livestock industries, it is necessary to better understand the behavior of phosphorus movement through the soil profile and into drainage systems. To better understand the impact of P movement to tile drainage systems, surface and subsurface runoff water quality was monitored on two agricultural fields from 2000 to 2004. The two sites were situated in the Missisquoi Bay Watershed and were instrumented to measure and sample the nutrients content of surface and subsurface runoff exiting the agricultural fields. Also, on each field, rainfall, temperature and solar radiation data was acquired and soil P levels were analyzed to better understand events. While phosphorus concentrations measured in surface runoff were approximately 10 times higher than those measured in tile drainage waters, concentrations in drainage effluent were frequently above the 0.03 mg L-1 guideline recommended by the MENV. Overall, tile drainage effluent accounted for 81% of the total volume of water and of 40% of the total phosphorus exported from the fields, indicating that drainage systems are a pathway for phosphorus transport. Currently, we have identified, a simulation model (FHANTM) which could potentially be used to simulate the transport of phosphorus from the fields. This model will first be calibrated and then validated to simulate the transport of phosphorus on the fields. The calibrated model will be used to identify drainage best management practices that could reduce the load of nutrients reaching river systems. It will permit to use to its maximum the data collected and to work with it as to develop our comprehension of phosphorus transport.

Sediment and Nutrient Removal Efficiencies in a Constructed Wetland in Southern Quebec

Non Point Source (NPS) pollution is considered a major cause of water quality deterioration around the globe. A study was conducted to assess the efficiency of a constructed wetland for sediment and nutrient removal from a riverine source containing NPS pollution in a Nordic climate. The constructed wetland, built near the town of Mystic, Quebec, consists of a sedimentation basin, a sinuous subsurface horizontal flow section and an open water body or pond that continuously receives up to 2.5% of the Walbridge Creek. The plants grown in the wetland are composed of several indigenous emergent aquatic species. Flow into and through the system is controlled by gravity. There is a gate on the intake structure, which allows flow to be adjusted, along with three composite weirs; located at the outlet of each section of the wetland. Water levels at the weirs are monitored continually using submersible pressure transducers and ultrasonic sensors. There are five sampling sites: one located in the river, at the inlet of the wetland, weir 1, weir 2, and weir 3. Grab samples are taken at least once a week and more frequently when storm events occurred. Automated samplers are also used to take composite samples during storm events. Sampling will span a two year time period, winters excluded. The water samples are analyzed for orthophosphates, dissolved phosphorus, organic phosphorus, nitrates, ammonia, dissolved nitrogen, and total suspended solids. A subset of the samples are analyzed in an external laboratory for an additional suite of parameters, including bioavailable phosphorus, total suspended solids, K, Ca, Mg, Na, and pH. From these results mass balances for the sediments and nutrients and nutrient removal efficiencies for the entire system and each of the components will be evaluated. Most of the reduction in 2003 occured during the spring and summer months. From the internal data we observed a 43.8% reduction in the median annual orthophosphate concentrations and a 22.4% reduction in the median annual total P concentrations. From the external data we observed a 47.5% reduction in the median annual orthophosphate, 14% reduction in the median annual total P and 37.8% reduction of the median annual bioavailable P concentrations. From these values (internal and external data sets) the wetland appears to be efficient in reducing mean annual phosphorus concentrations.