Abstract

ABSOLUT v1.0 is an adaptive algorithm that uses correlations between time-aggregated weather data and crop yields for yield prediction. At its core, locally (i.e. district-) specific multiple linear regressions are used to predict the annual crop yield based on four weather aggregates and a linear trend in time. In contrast to other statistical yield prediction methods, the input weather features are not predefined or based on a limited number of observed correlations but they are exhaustively tested for maximum explanatory power across all of their possible combinations in all districts of the modelling domain. Principal weather variables (such as temperature, precipitation, or sunshine duration) are aggregated over two to six consecutive months from the 12 months preceding the harvest. This gives 45 potential input features per original weather variable. In a first step, this zoo of possible input features is subset to those very probably holding explanatory power for observed yields. The second, computationally demanding step is making out-of-sample predictions for all districts with all possible combinations of the remaining features. Step three selects the seven combinations of four different weather features that have the highest explanatory power averaged over the districts. Finally, the district-specific best performing regression among these seven is used for district predictions, and the results can be spatially aggregated. To evaluate the new approach, ABSOLUT v1.0 is applied to predict the yields of ten major crops at the district level in Germany based on two decades of yield and weather data from about 300 districts. When aggregated to the national level, the predictions explain 70–90 % of the observed variance between years depending on crop type and time frame considered. District-level performance maps for winter wheat and silage maize show areas with > 40 % variance explanation covering about two thirds of the country.

Highlights

  • Weather-based crop yield predicitons have a long history; correlations between weather variables and agricultural yields had 20 already been studied in the first quarter of the 20th century (Meinardus, 1901; Hooker, 1907; Fisher, 1924), and estimating regional yields by multiple linear regressions from time-aggregated weather data has been around for decades

  • Locally specific multiple linear regressions are used to predict the annual crop yield based on four weather aggregates and a linear trend in time

  • Y(t) is the yield, i. e. the harvested mass per area in dt ha−1, of a certain crop in the year t; the βi are regression parameters; the wj,t are aggregated weather variables with specific time windows associated to t, these 30 are called “weather features” to avoid confusion with weather variables like temperature or precipitation in general; and ε is the estimation error to minimize

Read more

Summary

Introduction

Weather-based crop yield predicitons have a long history; correlations between weather variables and agricultural yields had 20 already been studied in the first quarter of the 20th century (Meinardus, 1901; Hooker, 1907; Fisher, 1924), and estimating regional yields by multiple linear regressions from time-aggregated weather data has been around for decades. E. the harvested mass per area in dt ha−1, of a certain crop in the year t; the βi are regression parameters; the wj,t are aggregated weather variables with specific time windows associated to t (for instance the precipitation sum of December, January, and February preceding the harvest in summer), these 30 are called “weather features” to avoid confusion with weather variables like temperature or precipitation in general; and ε is the estimation error to minimize. This is demonstrated here for Germany and its district-level administrative subunits (Kreise). This study should serve as proof of concept proposing another building block for more accurate predictions in similar setups, e. g. with panel (question 8) or nonlinear regression models

General requirements
Input data
Specifics of the example application
Germany as test bed for agricultural modelling
Primary data from external sources
Preprocessing and actual input data
Methods
Initialization, preparation of weather input features
Naïve exhaustive search and feature selection
Program 300 – “the gold pan”
Programs 400 and 500 – “crucible and mould”
Running program 100
Running programs 200 and 300
Running programs 400 and 500
How restrictive should the final selection be?
The number of climate aggregate features
Selection of principal weather variables
Silage maize 2018 on district level
Error compensation in spatial aggregates
Official in-season yield estimations
14 Sep 16 Sep 17 Sep
Weather input of Gornott and Wechsung (2016)
Findings
Conclusions and outlook
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call