Approaches to modular model development

One of the main goals of the International Consortium for Agricultural Systems Applications (ICASA) is to advance the development and application of compatible and complementary models, data and other systems analysis tools. To help reach that goal, it will adopt and recommend modular approaches that facilitate more systematic model development, documentation, maintenance, and sharing. In this paper, we present criteria and guidelines for modules that will enable them to be plugged into existing models to replace an existing component or to add a new one with minimal changes. This will make it possible to accept contributions from a wide group of modellers with specialities in different disciplines. Two approaches to modular model development have emerged from different research groups in ICASA. One approach was developed by extending the programming methods used in the Fortran Simulation Environment developed in The Netherlands. This method is being used in revisions of some of the Decision Support Systems for Agrotechnology Transfer crop models. A simple example of this approach is given in which a plant growth module is linked with a soil water balance module to create a crop model that simulates growth and yield for a uniform area. The second approach has been evolving within the Agricultural Production Systems Research Unit group in Australia. This approach, implemented in software called Agricultural Production Systems Simulator, consists of plug-in/pull-out modules and an infrastructure for inter-module communication. The two approaches have important similarities, but also differ in implementation details. In both cases, avoiding reliance on any particular programming language has been an important design criterion. By comparing features of both approaches, we have started to develop a set of recommendations for module design that will lead to a ‘toolkit’ of modules that can be shared throughout the ICASA network.  

Tomato yield, biomass accumulation, root distribution and irrigation water use efficiency on a sandy soil, as affected by nitrogen rate and irrigation scheduling

Florida is the largest producer of fresh-market tomatoes in the United States. Production areas are typically intensively managed with high inputs of fertilizer and irrigation. The objectives of this 3-year field study were to evaluate the interaction between N-fertilizer rates and irrigation scheduling on yield, irrigation water use efficiency (iWUE) and root distribution of tomato cultivated in a plastic mulched/drip irrigated production systems. Experimental treatments included three irrigation scheduling regimes and three N-rates (176, 220 and 230 kg ha⁻¹). Irrigation treatments included were: (1) SUR (surface drip irrigation) both irrigation and fertigation line placed right underneath the plastic mulch; (2) SDI (subsurface drip irrigation) where the irrigation line was placed 0.15 m below the fertigation line which was located on top of the bed; and (3) TIME (conventional control) with irrigation and fertigation lines placed as in SUR and irrigation being applied once a day. Except for the “TIME” treatment all irrigation treatments were controlled by soil moisture sensor (SMS)-based irrigation set at 10% volumetric water content which was allotted five irrigation windows daily and bypassed events if the soil water content exceeded the established threshold. Average marketable fruit yields were 28, 56 and 79 Mg ha⁻¹ for years 1-3, respectively. The SUR treatment required 15-51% less irrigation water when compared to TIME treatments, while the reductions in irrigation water use for SDI were 7-29%. Tomato yield was 11-80% higher for the SUR and SDI treatments than TIME where as N-rate did not affect yield. Root concentration was greatest in the vicinity of the irrigation and fertigation drip lines for all irrigation treatments. At the beginning of reproductive phase about 70-75% of the total root length density (RLD) was concentrated in the 0-15 cm soil layer while 15-20% of the roots were found in the 15-30 cm layer. Corresponding RLD distribution values during the reproductive phase were 68% and 22%, respectively. Root distribution in the soil profile thus appears to be mainly driven by development stage, soil moisture and nutrient availability. It is concluded that use of SDI and SMS-based systems consistently increased tomato yields while greatly improving irrigation water use efficiency and thereby reduced both irrigation water use and potential N leaching.  

Distribution of rainfall and soil moisture content in the soil profile under citrus tree canopy and at the dripline

The plant canopy intercepts rain and thus can alter the distribution of water under the canopy as compared to that along the dripline. The effects of a citrus (Citrus sinensis L. Osbeck) tree (25-year-old, Valencia orange) canopy on the distribution of rainfall and soil moisture content within the soil profile either along the dripline (D) or under the canopy near the trunk (inner side; I), and midway between I and Dripline (M) were evaluated, on the east and west sides of trees planted along north-south rows. Results of eleven storm events in 1995 (mean of east and west sides) revealed that the amounts of precipitation at the D, M, and I positions were 97–140, 47–94, and 52–79% of the incident rainfall, respectively. Thus, canopy interception of incident rainfall was quite appreciable. The soil moisture content was greater along the dripline compared to that at the M and I positions, particularly in the deeper (≥60 cm) soil profile. The water flux was significantly greater at the dripline than under the canopy indicating a greater leaching potential of soil-applied fertilizers and other chemicals when placed along the dripline. A substantial reduction in the rainfall and water flux under the canopy as a result of canopy interception suggests that application of fertilizer and chemicals under the canopy could minimize leaching losses. Received: 10 November 1997  

Stability of crop coefficients under different climate and irrigation management practices

Summary The effects to climate and management practices on crop water requirement coefficients were studied for a soybean crop growing on a sandy soil using a mechanistic model that computes evaporation and transpiration in response to soil, crop, and climatic factors. It was found that seasonal errors could the as high as 190 mm when crop coefficients developed under one set of conditions were used under different climate and management conditions. The largest error in ET occurred when vapor pressure was reduced from 26 mb to 14 mb; next in importance were site differences in wind speed, radiation, irrigation interval, temperature and planting date. Correction factors needed to adjust crop coefficients to those site specific conditions ranged from 0.73 to 1.30 depending on the time of season and climate or management variable that was changed. When the overall crop coefficient was divided into a plant specific and a soil specific coefficients, the plant coefficient was relatively stable compared to soil coefficients. The results of this study can help establish a practical range of conditions over which crop coefficients developed at one site can be used to compute the appropriate values for sites where measurements have not been made.Approved for publication as Florida Agricultural Experiment Station Journal Series No. 9514. This research was partially supported by the US AID project, International Benchmark Sites Network for Agrotechnology Transfer, No. DAN-4054-c-00-2071-00  

Influence of likelihood function choice for estimating crop model parameters using the generalized likelihood uncertainty estimation method

Proper estimation of model parameters is required for ensuring accurate model predictions and good model-based decisions. The generalized likelihood uncertainty estimation (GLUE) method is a Bayesian Monte Carlo parameter estimation technique that makes use of a likelihood function to measure the closeness-of-fit of modeled and observed data. Various likelihood functions and methods of combining likelihood values have been used in previous studies. This research was conducted to determine the effects of using previously reported likelihood functions in a GLUE procedure for estimating parameters in a widely-used crop simulation model. A factorial computer experiment was conducted with synthetic measurement data to compare four likelihood functions and three methods of combining likelihood values using the CERES-Maize model of the Decision Support System for Agrotechnology Transfer (DSSAT). The procedure used an arbitrarily-selected parameter set as the known “true parameter set” and the CERES-Maize model to generate true output values. Then synthetic observations of crop variables were randomly generated (four replicates) by using the simulated true output values (dry yield, anthesis date, maturity date, leaf nitrogen concentration, soil nitrate concentration, and soil moisture) and adding a random observation error based on the variances of corresponding field measurements. The environmental conditions were obtained from a sweet corn (Zea mays L.) experiment conducted in 2005 in northern Florida. Results showed that the method of combining likelihood values had a strong influence on parameter estimates. The combination method based on the product of the likelihoods associated with each set of observations reduced the uncertainties in posterior distributions of parameter estimates most significantly. It was also found that the likelihood function based on Gaussian probability density function was the best among those tested. This combination accurately estimated the true parameter values, suggesting that it can be used when estimating CERES-Maize model parameters for real experiments.  

Precision of soil moisture sensor irrigation controllers under field conditions

New soil moisture sensor systems (SMSs) for irrigation control have been commercialized in recent years. However, limited research has been carried out to evaluate their precision to measure the volumetric soil water content (θ). The objectives of this research were to: (a) determine the relationship between θ and the θ sensed by four commercially available SMSs, (b) quantify the proportion of scheduled irrigation cycles (SICs) that the SMSs bypassed, and (c) determine the θ at which SICs were allowed or bypassed. Sensors from brands Acclima, Rain Bird, Irrometer, and Water Watcher were buried at 7-10 cm depth, on plots with common bermudagrass [Cynodon dactylon (L.) Pers.]. A calibrated ECH₂O probe was also installed in every plot, at the same depth, to monitor θ continuously. When comparing the ECH₂O readings with θ sensed by the SMSs, significant correlations were found for the three Acclima RS500 (AC) systems tested, and for two of the three systems of Irrometer Watermark 200SS/WEM (IM) and Rain Bird MS-100 (RB). Most of the SMS-based treatments bypassed the majority of the SICs during rainy periods, and allowed irrigation during the dry periods. On average, 71% of the SICs were bypassed by the SMS treatments, without detriment to the turfgrass quality. However, most of the SMSs were not found to be precision instruments, because sometimes they bypassed SICs and sometimes they did not, even when reading the same or a lower θ. Considering the average θ range of over which the different SMS treatments always allowed or always bypassed irrigation, brand AC resulted in the significantly narrowest range (1.4%) followed by RB (3.2%), suggesting that they were more consistent and precise in measuring θ than Water Watcher DPS-100 (WW) and IM (7.4 and 7.8%, respectively). These results are consistent with the reported water savings achieved by these SMSs in related studies.  

Irrigation scheduling performance by evapotranspiration-based controllers

Evapotranspiration-based irrigation controllers, also known as ET controllers, use ET information or estimation to schedule irrigation. Previous research has shown that ET controllers could reduce irrigation as much as 42% when compared to a time-based irrigation schedule. The objective of this study was to determine the capability of three brands of ET-based irrigation controllers to schedule irrigation compared to a theoretically derived soil water balance model based on the Irrigation Association Smart Water Application Technologies (SWAT) protocol to determine the effectiveness of irrigation scheduling. Five treatments were established, T1-T5, replicated four times for a total of twenty field plots in a completely randomized block design. The irrigation treatments were as follows: T1, Weathermatic SL1600 with SLW15 weather monitor; T2, Toro Intelli-sense; T3, ETwater Smart Controller 100; T4, a time-based treatment determined by local recommendations; and T5, a reduced time-based treatment 60% of T4. All treatments utilized rain sensors set at a 6 mm threshold. A daily soil water balance model was used to calculate the theoretical irrigation requirements for comparison with actual irrigation water applied. Calculated in 30-day running totals, irrigation adequacy and scheduling efficiency were used to quantify under- and over-irrigation, respectively. The study period, 25 May 2006 through 27 November 2007, was drier than the historical average with a total of 1326 mm of rainfall compared to 1979 mm for the same historical period. It was found that all treatments applied less irrigation than required for all seasons. Additionally, the ET controllers applied only half of the irrigation calculated for the theoretical requirement for each irrigation event, on average. Irrigation adequacy decreased when the ET controllers were allowed to irrigate any day of the week. All treatments had decreased scheduling efficiency averages in the rainy season with the largest decrease of 29 percentile points with a timer and rain sensor (T4) and an average decrease of 20 percentile points for the ET controllers, indicating that site specific rainfall has a significant effect on scheduling efficiency results. Rainfall did not drastically impact the average irrigation adequacy results. For this study, there were two controller program settings that impacted the results. The first setting was the crop coefficients where specific values were chosen for the location of the study when calculating the theoretical requirement whereas the controllers used default values. The second setting was the soil type that defines the soil water holding capacity of the soil. The ET controllers were able to regularly adjust to real-time weather, unlike the conventional irrigation timers. However, the incorporation of site specific rainfall measurements is extremely important to their success at managing landscape water needs and at a minimum a rain sensor should be used.  

Water conservation potential of smart irrigation controllers on St. Augustinegrass

A variety of technologies for reducing residential irrigation water use are available to homeowners. These “Smart Irrigation” technologies include evapotranspiration (ET)-based controllers and soil moisture sensor (SMS) controllers. The purpose of this research was to evaluate the effectiveness of these technologies, along with rain sensors, based on irrigation applied and turfgrass quality measurements on St. Augustinegrass (Stenotaphrum secundatum (Walter) Kuntze). Testing was performed on two types of SMS controllers (LawnLogic LL1004 and Acclima Digital TDT RS500) at three soil moisture threshold settings. Mini-Clik rain sensors (RS) comprised six treatments at two rainfall thresholds (3 mm and 6 mm) and three different irrigation frequencies (1, 2, and 7 d/wk). Two ET controllers were also tested, the Toro Intelli-Sense controller and the Rain Bird ET Manager. A time-based treatment with 2 days of irrigation per week without any type of sensor (WOS) to bypass irrigation was established as a comparison. All irrigation controller programming represented settings that might be used in residential/commercial landscapes. Even though three of the four testing periods were relatively dry, all of the technologies tested managed to reduce water application compared to the WOS treatment, with most treatments also producing acceptable turf quality. Reductions in irrigation applied were as follows: 7–30% for RS-based treatments, 0–74% for SMS-based treatments, and 25–62% for ET-based treatments. The SMS treatments at low threshold settings resulted in high water savings, but reduced turf quality to unacceptable levels. The medium threshold setting (approximately field capacity) SMS-based treatment produced good turfgrass quality while reducing irrigation water use compared to WOS by 11–53%. ET controllers with comparable settings and good turf quality had −20% to 59% savings. Reducing the irrigation schedule (treatment DWRS) by 40% and using a rain sensor produced water savings between 36% and 53% similar to smart controllers. Proper installation and programming of each of the technologies was essential element to balancing water conservation and acceptable turf quality. Water savings with the SMS controllers could have been increased with a reduced time-based irrigation schedule. Efficiency settings of 100% (DWRS) and 95% (TORO) did not reduce turf quality below acceptable limits and resulted in substantial irrigation savings, indicating that efficiency values need not be low in well designed and maintained irrigation systems. For most conditions in Florida, the DWRS schedule (60% of schedule used for SMS treatments) can be used with either rain sensors or soil moisture sensors in bypass control mode as long as the irrigation system has good coverage and is in good repair.  

Landscape irrigation scheduling efficiency and adequacy by various control technologies

Automated residential irrigation systems tend to result in higher water use than non-automated systems. Increasing the scheduling efficiency of an automated irrigation system provides the opportunity to conserve water resources while maintaining good landscape quality. Control technologies available for reducing over-irrigation include evapotranspiration (ET) based controllers, soil moisture sensor (SMS) controllers, and rain sensors (RS). The purpose of this research was to evaluate the capability of these control technologies to schedule irrigation compared to a soil water balance model based on the Irrigation Association (IA) Smart Water Application Technologies (SWAT) testing protocol. Irrigation adequacy and scheduling efficiency were calculated in 30-day running totals to determine the amount of over- or under-irrigation for each control technology based on the IA SWAT testing protocol. A time-based treatment with irrigation 2 days/week and no rain sensor (NRS) was established as a comparison. In general, the irrigation adequacy ratings (measure of under-irrigation) for the treatments were higher during the fall months of testing than the spring months due to lower ET resulting in lower irrigation demand. Scheduling efficiency values (measure of over-irrigation) decreased for all treatments when rainfall increased. During the rainy period of this testing, total rainfall was almost double reference evapotranspiration (ET_o) while in the remaining three testing periods the opposite was true. The 30-day irrigation adequacy values, considering all treatments, varied during the testing periods by 0-68 percentile points. Looking at only one 30-day testing period, as is done in the IA SWAT testing protocol, will not fully capture the performance of an irrigation controller. Scheduling efficiency alone was not a good indicator of controller performance. The amount of water applied and the timing of application were both important to maintaining acceptable turfgrass quality and receiving good irrigation adequacy and scheduling efficiency scores.  

番茄叶片早疫病近红外高光谱成像检测技术

提出了基于格拉姆斯密特(MGS)模型和贝叶斯罗蒂斯克回归(BlogReg)的近红外高光谱成像技术检测番茄叶片早疫病的方法。利用高光谱图像采集系统获取波长874~1 734 nm范围内70个染病和80个健康番茄叶片的高光谱图像,选取染病和健康叶片30像素×30像素感兴趣区域的光谱反射率。建立了番茄叶片早疫病的最小二乘-支持向量机(LS-SVM)识别模型,再通过MGS和BlogReg提取特征波长(EW),分别得到5个(911、1 409、1 511、1609、1 656 nm)和9个(901、905、908、915、918、1 123、1 305、1 460、1 680 nm)特征波长,并建立EW-LS-SVM和EWLDA模型。在所有模型中,建模集的正确识别率为93%~98%,预测集的正确识别率为96%~100%。结果表明,近红外高光谱成像技术检测番茄叶片早疫病是可行的,MGS和BlogReg都是有效的特征波长提取方法。