Predicting emergence of western Predicting emergence of western Predicting emergence of western Predicting emergence of western Predicting emergence of western Predicting emergence of western corn rootworm, corn rootworm, Diabrotica virgifera Diabrotica virgifera (L C t) d lt ith d (L C t) d lt ith d d Predicting emergence of western Predicting emergence of western corn rootworm, corn rootworm, Diabrotica virgifera Diabrotica virgifera (L C t) d lt ith d (L C t) d lt ith d d (LeConte), adults with a degree (LeConte), adults with a degree-day day phenology model closely tied to phenology model closely tied to (LeConte), adults with a degree (LeConte), adults with a degree-day day phenology model closely tied to phenology model closely tied to the development of corn the development of corn the development of corn the development of corn Douglass E. Stevenson, Gerald J. Michels, Douglass E. Stevenson, Gerald J. Michels, John B. Bible, John A. Jackman, Marvin K. Harris John B. Bible, John A. Jackman, Marvin K. Harris
17
Embed
Model Predicting Emergence of Western Corn Rootwormamarillo.tamu.edu/...of...Diabrotica-virgifera.pdf · Model Selection Results of NLIN & OLS RegressionsResults of NLIN & OLS Regressions
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Predicting emergence of westernPredicting emergence of westernPredicting emergence of westernPredicting emergence of westernPredicting emergence of western Predicting emergence of western corn rootworm, corn rootworm, Diabrotica virgiferaDiabrotica virgifera(L C t ) d lt ith d(L C t ) d lt ith d dd
Predicting emergence of western Predicting emergence of western corn rootworm, corn rootworm, Diabrotica virgiferaDiabrotica virgifera(L C t ) d lt ith d(L C t ) d lt ith d dd(LeConte), adults with a degree(LeConte), adults with a degree--day day phenology model closely tied to phenology model closely tied to (LeConte), adults with a degree(LeConte), adults with a degree--day day phenology model closely tied to phenology model closely tied to the development of cornthe development of cornthe development of cornthe development of corn
Douglass E. Stevenson, Gerald J. Michels,Douglass E. Stevenson, Gerald J. Michels,John B. Bible, John A. Jackman, Marvin K. HarrisJohn B. Bible, John A. Jackman, Marvin K. Harris
Spread of Spread of D. virgiferaD. virgifera since 1955since 1955
195519551955195519751975197519751990199019901990
Materials & MethodsMaterials & MethodsMaterials & MethodsMaterials & Methods++ The study areaThe study area
66 Fields selectedFields selected!! Ett (N HP)Ett (N HP)!! Etter (No. HP)Etter (No. HP)!! Dalhart (No. HP)Dalhart (No. HP)!! Black (So. HP)Black (So. HP)!! Dimmitt (So. HP)Dimmitt (So. HP)
Materials & MethodsMaterials & MethodsMaterials & MethodsMaterials & Methods++ Model development stepsModel development steps
66 Insect SamplingInsect Sampling!! S t ti d lS t ti d l!! Systematic random sampleSystematic random sample!! Emergence & Pherocon AMEmergence & Pherocon AM®® TrapsTraps!! 4 transects of 5 traps per field4 transects of 5 traps per field!! Checked weeklyChecked weekly
66 DegreeDegree--day computation methodday computation method!! Root orm Temp a e (no c toff)Root orm Temp a e (no c toff) Stress degreeStress degree da s (SDD)da s (SDD)!! Rootworm: Temp. ave. (no cutoff) Rootworm: Temp. ave. (no cutoff) –– Stress degreeStress degree--days (SDD)days (SDD)!! Corn: Barger (1969) method (horiz. cutoff = 86ºF)Corn: Barger (1969) method (horiz. cutoff = 86ºF)
66 Determination of base temp. & start dateDetermination of base temp. & start date!! Point prediction model with RingPoint prediction model with Ring Jackman ANOVAJackman ANOVAPoint prediction model with RingPoint prediction model with Ring--Jackman ANOVAJackman ANOVA!! Compares all prospective base temps & start datesCompares all prospective base temps & start dates
Mean & Median dates & degreeMean & Median dates & degree--day sumsday sums 64,000 prospective models64,000 prospective models
!! Selection on the basis of lowest RMSESelection on the basis of lowest RMSE
Materials & MethodsMaterials & MethodsMaterials & MethodsMaterials & Methods++ Model development stepsModel development steps
66 Selection of mathematical functionSelection of mathematical function!! 19 sigmoid functions19 sigmoid functions!! Nonlinear regression: RMSE, F statistic, Prob. > FNonlinear regression: RMSE, F statistic, Prob. > F!! OLS regression: ROLS regression: R22
!! Selection: highest R2 & F stat, lowest RMSE & Prob. > FSelection: highest R2 & F stat, lowest RMSE & Prob. > F
Determination of model coefficientsDetermination of model coefficients66 Determination of model coefficientsDetermination of model coefficients!! Nonlinear regressionNonlinear regression!! 44--parameter modified Gompertz functionparameter modified Gompertz function
66 Synchronization with corn phenologySynchronization with corn phenology!! Corn GDD (Barger 1969)Corn GDD (Barger 1969)
66 Stat. Stat. –– Nonparametric method of (Kutner et al. 2004)Nonparametric method of (Kutner et al. 2004)!! Ratio MSPE / MSERatio MSPE / MSE(model)(model)
66 CalibrationCalibration!! Early / late: predictions shifted Early / late: predictions shifted ±± 10 days early & late10 days early & late!! Natrual variability (location / year): pred Superimposed on obsNatrual variability (location / year): pred Superimposed on obsNatrual variability (location / year): pred. Superimposed on obs.Natrual variability (location / year): pred. Superimposed on obs.
66 ValidationValidation!! Compute ratio & compare to calibrationsCompute ratio & compare to calibrationsCompute ratio & compare to calibrationsCompute ratio & compare to calibrations!! Compute days early & lateCompute days early & late!! Visual: Scatterplot with 4.5Visual: Scatterplot with 4.5--day prediction intervalday prediction interval
Base Temp & Start Day ResultsBase Temp & Start Day ResultsBase Temp & Start Day ResultsBase Temp & Start Day Results
Base = 50ºF, Start day = 112, Median corn VE date = 113Base = 50ºF, Start day = 112, Median corn VE date = 113
Model SelectionModel SelectionResults of NLIN & OLS RegressionsResults of NLIN & OLS Regressions
Model SelectionModel SelectionResults of NLIN & OLS RegressionsResults of NLIN & OLS RegressionsResults of NLIN & OLS RegressionsResults of NLIN & OLS RegressionsResults of NLIN & OLS RegressionsResults of NLIN & OLS RegressionsModel functionModel function RMSERMSE Med. FMed. F Prob > FProb > F RR22
Functions not in top 10:Functions not in top 10: MorganMorgan--MercerMercer--Florin, MichaelisFlorin, Michaelis--Menten, Jolicoeur, Menten, Jolicoeur, Modified Weibull (3Modified Weibull (3--Parameter), Schnute, MitscherlichParameter), Schnute, Mitscherlich--Spillman, Hill,Spillman, Hill,PreecePreece--Baines, ZengBaines, Zeng--Wan.Wan.
Model Coefficient ResultsModel Coefficient ResultsModel Coefficient ResultsModel Coefficient ResultsM d l ffi i tM d l ffi i t ** (( (( ((bb ** (( dd ))))))++ Model coefficients: Model coefficients: y = a * expy = a * exp((-- ((expexp((b b –– c * c * ((xx--d d ))))))
66 aa = 96.5 = 96.5 66
!! Vert. scale parameter (controls shape of curve on Vert. scale parameter (controls shape of curve on yy--axis)axis)!! upper reliability limitupper reliability limit
66 bb = 6.0= 6.0!! Horiz. scale parameter (controls shape of curve on Horiz. scale parameter (controls shape of curve on xx--axis)axis)!! determines inflection points and slope of curvedetermines inflection points and slope of curve
66 cc = 0.00404= 0.00404!! shape parameter (controls assymetry of curve)shape parameter (controls assymetry of curve)!! small values indicate short early tail of the curvesmall values indicate short early tail of the curve
66 dd = 4.0= 4.0!! Shift parameterShift parameter!! Determines starting point along Determines starting point along xx--axisaxis
Valid predictions within Valid predictions within ±± 4.5 days will have MSPE/MSE4.5 days will have MSPE/MSE(model)(model) ratios < 5.1819ratios < 5.1819
ValidationValidationCalibration for Natural VariabilityCalibration for Natural Variability
ValidationValidationCalibration for Natural VariabilityCalibration for Natural VariabilityCalibration for Natural VariabilityCalibration for Natural VariabilityCalibration for Natural VariabilityCalibration for Natural Variability
Unbiased predictions will have MSPE/MSEUnbiased predictions will have MSPE/MSE(model)(model) ratios < 5.0941ratios < 5.0941
Model ValidationModel ValidationMSPE/MSE Ratios for Locations & YearsMSPE/MSE Ratios for Locations & Years
Model ValidationModel ValidationMSPE/MSE Ratios for Locations & YearsMSPE/MSE Ratios for Locations & YearsMSPE/MSE Ratios for Locations & Years MSPE/MSE Ratios for Locations & Years MSPE/MSE Ratios for Locations & Years MSPE/MSE Ratios for Locations & Years
All location / year MSPE/MSEAll location / year MSPE/MSE(model)(model) ratios are within acceptable ratios are within acceptable range for prediction interval range for prediction interval
Model ValidationModel ValidationPrediction Error in DaysPrediction Error in Days
Model ValidationModel ValidationPrediction Error in DaysPrediction Error in DaysPrediction Error in DaysPrediction Error in DaysPrediction Error in DaysPrediction Error in Days
All location / year predictions are within prediction interval (All location / year predictions are within prediction interval (±± 4.5 days) 4.5 days)
Start = corn VE date, DegreeStart = corn VE date, Degree--days = (DD50ºF days = (DD50ºF –– SDD) SDD)
116 Obs hidden by prediction line
ConclusionsConclusionsConclusionsConclusions++ Model predictions give more time for decisions to corn producersModel predictions give more time for decisions to corn producers++ Model predictions give more time for decisions to corn producers Model predictions give more time for decisions to corn producers
in the Texas High Plains than scouting or trapping alone.in the Texas High Plains than scouting or trapping alone.
++ Synchronization with corn permits integration of pestSynchronization with corn permits integration of pest++ Synchronization with corn permits integration of pest Synchronization with corn permits integration of pest management with crop management decisions.management with crop management decisions.
++ The model is applicable to adjacent regions of the So GreatThe model is applicable to adjacent regions of the So Great++ The model is applicable to adjacent regions of the So. Great The model is applicable to adjacent regions of the So. Great Plains.Plains.
++ Use in areas outside Texas High Plains will require fieldUse in areas outside Texas High Plains will require field++ Use in areas outside Texas High Plains will require field Use in areas outside Texas High Plains will require field validation.validation.
++ Shows the fungibility of this approach to developing prediction Shows the fungibility of this approach to developing prediction g y pp p g pg y pp p g pmodels for other insectsmodels for other insects
++ Phenology models developed in this way can provide very close Phenology models developed in this way can provide very close gy y ygy y yestimates of physiological base temperatures and predictions of estimates of physiological base temperatures and predictions of phenological events of interest in the life cycles of insects.phenological events of interest in the life cycles of insects.