Contact

Look up in the sky, put your feet in the field and listen to farmers: Collecting rice production data in the TVSEP panel household and panel villages in Thailand self-reported yields, crop cuts and satellite-based measurements

Project profile

Country Thailand
Location About 200 rice plots in TVSEP Villages in Ubon Ratchathani and Nakhon Phanom Province
Project coordinator Prof. Dr. Hermann Waibel
Project participants

Prof. Dr. Eric Strobl, University of Bern, Switzerland

Assistant Prof. Dr. Prof. Sabine Liebenehm, University of Saskatchewan, Canada

MSc Niels R. Wendt, Leibniz University Hannover, Germany

Assistant Professor Noppon Tantisirin, Ubon Ratchathani University, Thailand

Time frame 09/2022 – 12/2023
Total budget Around 550,000 THB
Source of budget Institute of Development and Agricultural Economics, Leibniz University Hannover

 

Background

The Thailand Vietnam Socioeconomic Panel (TVSEP) has been collecting household level data from some 2200 households in Northeast Thailand, since 2007 during a total of nine panel waves. Rice is a staple crop and is still the backbone of food security in the region. Most households in the panel produce wetland rice under rain fed conditions with one crop per year. Until to date, rice production data in TVSEP are recorded based on respondent’s subjective recall assessment at plot level. Plot characteristics like size, water supply and distance from homestead are also collected in addition to data of production inputs and random events like pest outbreaks, drought of flood. The panel data, in principle, offer an excellent base to study trends in agricultural productivity, considering major issues like climate change, for example. However, yield data based on farmer subjective assessments are often inaccurate due to two types of errors, namely (a) incorrect field size measurements and (b) memory/ estimation bias of respondents.

This project is a first step towards a possible future replacement of collecting farm land and crop production data via questionnaires with satellite data. Technological advances in survey data collection such as GPS and satellite data have brought forward new options to advance data quality in household surveys in developing countries. This has stimulated recent studies to evaluate field data (e.g. Carletto et al, 2017; Lobell et al. 2020; Kosmowski et al. 2021) . For the measurement of plot sizes, geo-referencing tools are available with tablet-based questionnaires which to some extent can eliminate the first type of error. For satellite-based yield assessments, dynamic yield growth models, driven by real-time satellite data are now more widely available. The latter however requires ground testing by comparing model predicted yields with actual yield measurements. Hereby, the “gold standard” is crop cuts using full plot measurements. Both data sets can then be compared with Farmer estimates of crop yields. Various modelling exercises can be performed and correction factors for past errors in yield data can be derived. This will facilitate a wide range of scientific analysis of agricultural productivity questions using the TVSEP panel data.

Project implementation

During the 2022 TVSEP household panel wave, scheduled from April to June, GPS-based measurements of all plots recorded in the panel were recorded. These plot-specific georeferenced data can be used for modelling rice yields on plot level. These will be compared with farmer yield estimates and with yield measurements by crop cuts as the reference points.
The core of this TVSEP add-on project are crop cuts in some 230 TVSEP and non-TVSEP households in TVSEP villages in the provinces of Ubon Ratchathani and Nakhon Phanom. It is expected that at least 70 % of the plots belong to TVSEP households and the remainder to non-TVSEP households in TVSEP villages.
From each household who is harvesting rice, we select one rice plot (a plot is a contiguous piece of rice land surrounded by earthen dikes or other demarcations). The size of the plots ranges from a minimum of 200 sqm to several ha, depending on the harvesting method. There are three harvesting methods. First, a larger size combine harvester unloads the rice on a truck parked at road size. The truck brings the rice the rice mill and the famer instantly receives a record of his rice production harvested on that day. This method is generally for larger farmers whose rice field has good road access. The second method is a small combine harvester who unloads the rice in sacks of around 30 kg of paddy each which are weighed once the plot is harvested. The third method is harvesting by sickles and subsequent threshing by hand. This method is rare now but occasionally still exists. The production of the chosen plot is weighed using calibrated local weighing scales. In all cases three samples of moisture measurement were taken at random from the sacks or the truck. Plot harvesting is closely coordinated with the respective household heads and performed by students from Ubon Ratchathani University who received three days training. The TVSEP field technicians will supervise the harvesting and will be in charge for weighing the plot-level harvest and take GPS readings of the plot harvested. The shape files of the plots are forwarded to Prof Eric Strobl, University of Bern, Switzerland for identifying the corresponding satellite data.
A short tablet-based questionnaire is used to interview the farmer about the history of the plot during the 2022 rice production season, ask his subjective estimate of the plot size and the rice production of the plot. Likewise, the actual GPS readings of the plots and the respective yield are entered in the tablet program. A total of 7 technicians/enumerators will carry out the field operation under the supervision of Assistant Professor Nopporn Tantisirin, UBU and Prof. Hermann Waibel, LUH. Online monitoring of questionnaires is carried out by Mr. Niels Wendt, LUH and Dr Sabine Liebenehm. Every technician/enumerator will complete at least one farmer per day. Hence the field operation of the project will take about one month.
The project is carried out during the main harvesting period which is November/December 2022. Specific planning and testing activities started in September 2022. A detailed field research protocol is worked out together with Assistant Professor Nopporn Tantisirin.

Data analysis

Data analysis and writing of scientific papers will start in early 2023 and continue throughout the year. Initial descriptive statistics are prepared to provide feedback to the participating farmers in order to inform them about the actual yield measured versus their subjective estimates and hereby reward them for their cooperation.
Furthermore, the data offer the opportunity to carry out several modelling approaches, including various version of rice production functions based on measured yield data, a model explaining differences between satellite yield estimates and ground measurements, and, perhaps most importantly, a range of behavioural models explaining under- and overestimating of rice yields by TVSEP respondents in the past. Above all, these data will allow a reassessment of rice yields reported in past surveys. The GPS data will facilitate the establishment of a plot panel for rice farmers in the TVSEP panel in Thailand.
The uniqueness of this add-on project offers the possibility to produce several high level papers. Furthermore, the lessons learned from this project will contribute to an advancement in survey technology.

Literature

Carletto, C., S. Gourlay, S. Murray, and A. Zezza. 2017. Cheaper, Faster, and More than Good Enough: Is GPS the New Gold Standard in Land Area Measurement. Survey Research Methods 11 (3): 235–65.
Kosmowski, F., Chamberlin, J., Ayalew, H., Sida, T., Abay, K., & Craufurd, P. (2021). How accurate are yield estimates from crop cuts? Evidence from smallholder maize farms in Ethiopia. Food policy, 102, 102122.

Lobell, D.B., Azzari, G., Burke, M., Gourlay, S., Jin, Z., Kilic, T., Murray, S., 2020. Eyes in the Sky, Boots on the Ground: Assessing Satellite- and Ground-Based Approaches to Crop Yield Measurement and Analysis. Am. J. Agric. Econ. 102 (1), 202–219. doi.org/10.1093/ajae/aaz051.