The spatial patterns in long-term temporal trends of three major crops’ yields in Japan

Abstract Long-term trend of crop yields has been widely studied in global scales to find which crops and which geographic regions offer the best hope of meeting food demands, and which regions needed the most improvements. In this study, a mathematical method was applied to analyze spatial patterns in long-term temporal trends of three major crops’ yields in Japan archipelago. The changes in annual yields of rice, wheat, and soybean over a period of about 60 years in all 47 prefectures of Japan was analyzed by using the data of agricultural records. For all the three crops, the nationwide yields previously improved, but currently were stagnating in Japan. The result suggests that the annual yields were not improving in 53, 85, and 89% of those prefectures in Japan for rice, wheat, and soybean, respectively. The spatial patterns in temporal trends show that the percentage of number of yield-not-improving prefecture was higher in low latitude regions than high latitude regions. These results highlight the increasingly difficult challenge of meeting the growing demands and stagnating supplies in daily staple foods not only for agricultural scientists but also for Japanese society.

Japan is a country having one of the highest levels of crop yields per unit area over the world because only 12% of its land is suitable for cultivation (USDA, 2012). The overall agricultural self-sufficiency rate in Japan is ~50% on fewer than 14 million acres lands cultivated (USDA, 2012). Rice is considered the most important crop for Japan's society. In 2014, Japan dedicated 10.7 million ha to rice cultivation (FAO, 2015), which ranks the 17th in the world. The other two important food staples in Japanese food culture, wheat and soybean, their ranks of production are the 35th and 47th in the world in 2014 (FAO, 2015).

Yield trend analysis
This study was inspired by Ray et al. (2012), who used parsimonious regression models to examine the trends in crop yields for maize, rice, wheat, and soybeans across the globe extending over the period 1961-2008. Yield trends were analyzed using these parsimonious regression models of increasing order for: an intercept-only model (Equation (1)), a linear model (Equation (2)), a quadratic model (Equation (3)), and a cubic model (Equation (4)).
2016; Ray et al., 2012Ray et al., , 2013Tilman et al., 2011). By analyzing spatial patterns in long-term temporal trends of crop yield, the aims of this work are to find which crops and which geographic regions in Japan archipelago offer the best hope of meeting food demands and which regions are improvements most needed. First, I analyze changes in annual yields of rice, wheat, and soybean over a period of ~60 years in all the 47 prefectures of Japan by using the data obtained from the governmental official website. The crop yield trends for each prefecture are classified into four categories including (1) increasing, (2) stagnating, (3) collapsed, and (4) never improved, by using parsimonious regression models of increasing order. Last, I map these different temporal trends in prefectures and discuss their spatial patterns for all the three crops. The results of this research highlight the increasingly difficult challenge of meeting the growing demands and stagnating supplies in daily staple foods for Japanese society.

Data of crop yield
Crop yield data of rice, wheat, and soybean in all the 47 prefectures in Japan ( Figure 1) were downloaded from the official website of Ministry of Agriculture, Forestry and Fisheries, Japan (available online from http://www.maff. Here, Y is the yield (g m −2 ), t is the year, a is the intercept, and b, c, and d are the coefficients of regression.

Choosing the statistical model that best represents production trends
The Akaike Information Criterion (AIC) developed by Akaike (1974) was used to decide which statistical model fitted the observed data best, and computed AIC (Equation (5)) for each of the above four models (Equations (1)-(4)): Here, ss is residual sum of squares, n is the sample size, and p is the number of parameters. The model with the minimum AIC was chosen as the best representation of the production trend for a given prefecture. All calculations and data analyses were performed using R v 3.0.2 (R Development Core Team, 2013).

Classification of production trends
Based on the chosen model parameters, crop yield trends were classified into four main categories: increasing, stagnating, collapsed, and never improved. These classifications are defined as follows. (1) Yield increasing: (i) when the chosen model was linear, with a positive slope; (ii)

Long-term trends of crop yield in Japan
According to Figure 2(a), the annual temporal variation in prefecture average of rice yield in Japan ranged between 346 (in 1958) and 511 (in 2015) g m −2 , with a mean of 455 and standard deviation 48, during the 59-year period between 1958 and 2016. The chosen model for the average rice yield was quadratic with a negative quadratic term, and the yield for the 2010s had not reached the low values in the 1950s. That is, the yield previously improved, but currently was stagnating. As for wheat, the annual temporal variation in prefecture average of yield in Japan ranged between 108 (in 1963) and 306 (in 2000) g m −2 , with a mean of 258 and standard deviation 33, during the 59-year period between 1958 and 2016 (Figure 2(b)). The chosen The yield previously improved, but currently was stagnating. As for soybean, the annual temporal variation in prefecture average of yield in Japan ranged between 94 (in 1949) and 166 (in 1996) g m −2 , with a mean of 133 and standard deviation 18, during the 69-year period between 1948 and 2016 ( Figure  2(c)). The chosen model for the average soybean yield was cubic, and the yield for the 2010s had not reached the low values in the 1950s with the peak beyond 2010. The yield previously improved, but currently was stagnating.

Long-term trends of crop yield in each prefecture
The yield trends in Japan's prefectures were divided into four types. Figure 3 illustrates examples for each type: (1) collapsed in 11, 62, and 23% of the prefectures, respectively, and it has never improved in two prefectures ( Figure  4(c) and Table 1). Yields were stagnating in more than half of all prefectures in the country for all the three crops. Rice yield was increasing in near half of the prefectures. On the other hand, the number of yield-decreasing prefecture were more than that of the yield-increasing prefecture for wheat and soybean. The result shows some patterns of spatial differences in temporal trend for the regions located in different latitude (Figure 1). For rice and wheat yields, the percentage of number of yield-increasing prefecture was higher in high latitude regions (northern and eastern regions), but the percentage of number of yield-stagnating prefecture was higher in low latitude regions (southern and western regions, Table 1). For soybean, both of the percentages of number of yield-increasing and -stagnating prefectures were higher in high latitude regions, but the percentage of number of yield-decreasing prefecture was higher in low latitude regions (Table 1).

Discussion
Growing conditions of crops have changed over time due to the changes in the natural environment and cultivation (Craufurd & Wheeler, 2009;Lobell & Burke, 2010; Table S1.

Spatial patterns in long-term trends of crop yield
Within the 59-year period of analysis, rice yield was increasing and stagnating in 47 and 53% of the prefectures in Japan, respectively, and it has not collapsed in any prefecture (Figure 4(a) and Table 1). Wheat yield was increasing, stagnating, and collapsed in 15, 55, and 28% of the prefectures, respectively, and it has never improved in one prefecture (Figure 4(b) and Table 1). Within the 69-year period of analysis, soybean yield was increasing, stagnating, and   Figure 5). For rice, the prefectures which originally had high yield (high during 1958−1967) are Nagano, Niigata, and the prefectures in northern region except Hokkaido. For wheat, the prefectures which originally had high yield (high during 1958−1967) are mostly the prefectures in eastern region, and Miyagi, Yamagata, and Fukushima in northern region. For soybean, the prefectures which originally had high yield (high during 1948−1957) are Kagawa, and some prefectures in eastern region. The yields of rice in those prefectures which originally had high yield were still increasing (Figures 4 and 5). However, for wheat and soybean, yields in those prefectures which originally had high yield were either stagnating or collapsed. Hokkaido is one of the leading producers of crop in Japan (Chen, 2016). The yields of rice and wheat were still increasing in Hokkaido currently (Figure 4). Miyagi and Fukui are the only two yield-increasing prefectures for all the three crops ( Figure 4). In Japan, eating quality has been suggested as one of the most important factors for the production. Koshihikari is a famous rice strain mainly grown in Niigata; it is the most popular and expensive strain in Japan (Ebitani et al., 2005). The fame and the high quality of this strain are due to the ideal growing conditions in Niigata (Ishizaki et al., 2005). Nagano prefecture has the largest yield of rice in Japan due to its rivers and complicated water channels designed to bring nutrient-rich water to the crops. Nagano has a high elevation basin surrounded by mountains; thus, the area experiences large differences in temperature between day and night that provide ideal growing conditions for crops (JMA, 2016). The yield of wheat collapsed in most of the prefectures in Kinki and Chūgoku regions (Shimane, Hiroshima, Kyoto, Osaka, Hyogo, and Wakayama, Figure 4). Because wheat al., 2011;Shimono et al., 2010;Walther et al., 2002). The yield of crop is affected by climatic factors such as annual rainfall (Drury & Tan, 1994;Spiecker, 1995), solar radiation (Lobell et al., 2013;Welch et al., 2010), and air temperature (Lobell & Field, 2007;Matsui et al., 2001). Besides, the long history of cultivation and the geographical variation of climatic conditions, such as the number of rainy days during cropping season (for example, the length and starting day of Asian Rainy Season vary from area to area in different years), also result in large spatial differences in crop yield (Lobell et al., 2009). Furthermore, the impact of global warming can negatively affect crop yields on a global scale (Chen, 2016;Rosenzweig & Parry, 1994). To discuss these issues, a large number of studies have analyzed the contributions of climatic factors for rice (Morita et al., 2016;Peng et al., 2004;Shimono, 2008;Shimono et al., 2010), wheat (Asseng et al., 2015), and soybean (Egli, 2008a(Egli, , 2008b. In addition to the climatic factors, the extent of crop yield variation may vary geographically according to the types of cultivation, nitrogen fertilization, and soil type and fertility (Adams et al., 1998;Aydinalp & Cresser, 2008;Chen et al., 2014;Fuhrer, 2003). Contribution of cultivar differences has also been reported as an important factor on crop yield for rice (Peng et al., 1999;Saitoh et al., 1993;Zhang & Kokubun, 2004), wheat (Zhou et al., 2007;Ziska, 2008), and soybean (Matsuo et al., 2016(Matsuo et al., , 2017Ziska & Bunce, 2000). In order to analyze so many factors accompanied with huge datasets, models of crop growth are required to estimate and to predict how crop yield responds to the natural environment and cultivation.
is a long-day plant and can grow among a wide range of area, sunshine hours and daily radiation may be the deciding factors for the decline of yield (Bannayan et al., 2003). The yield of soybean collapsed in most of the prefectures in Kyushu (Oita, Miyazaki, and Kagoshima) and some in southern Japan (Figure 4). The collapse of yield may be related to the changing temperature in growing stage for soybean (Juang, 1993). The results of this study showed that nationwide yields previously improved, but currently was stagnating for all the three crops in Japan. The annual yields were not improving in more than half of the prefectures in Japan for rice, wheat, and soybean. The result showed that the percentage of number of yield-not-improving prefecture was higher in low latitude regions than high latitude regions for the three crops in Japan. New investments and strategies to increase or maintain production in the high-performing areas are required, while simultaneously preserving a sustainable environment and cultivation for all crops.